Attribute detail
status
458 ideas · 7 distinct values
| Value | Distribution | Ideas | Share |
|---|---|---|---|
| done | 399 | 87.1% | |
| incomplete | 29 | 6.3% | |
| in-progress | 19 | 4.1% | |
| draft | 7 | 1.5% | |
| pending | 2 | 0.4% | |
| partial | 1 | 0.2% | |
| todo | 1 | 0.2% |
| Idea | Value | Status | Tags | Summary | Katz |
|---|---|---|---|---|---|
| cli | done | done | Headless twin of extension / MCP analysis: [cli_package] ( `reqlan` / `rq` ), same `<base>/.reqlan` index. [commands]: `init` · `parse <file>` · `analyse` / `analyze` · `search <query>` · `click <target>` · `export` / `export html`. `--json` for machines; `--cwd` / `REQLAN_WORKSPACE`; `REQLAN_INDEX_PATH` overrides index dir. | 0.040 | |
| skills | done | done | rq-* skills ( [chat_skill_naming], [cursor_skills_install] ) teach the graph, not the whole repo. Ontology in one breath — this project as example: - [idea] — named unit of intent ( `cli_package`, `skills` ) - [file] / [ideaset] — containers ( `reqlan rq/cli/cli_package.rq`; file = implicit ideaset ) - [base] — `.reqlan` boundary; one ideas index per base - [reference] — `[idea]` / `[[wiki]]` / `["./path".idea]` to ideas and code - [attribute] / [attribute_body] — `@name …`; first unmarked text is the body Prefer focused search / file context / local graph over full dumps ( [ai_integration] ). | 0.040 | |
| special_attributes | done | done | When creating or updating ideas, skills MUST set: - `@status` — lifecycle ( `draft` | `pending` | `in-progress` | `done` | … ) - `@todo` — open gaps / follow-ups ( bare or with a note ) - `@tests` — quoted test paths that prove the idea ( optional `:test name` suffix ) Do not bury status, todos, or test links in body prose. | 0.046 | |
| active_base_switch | done | done | Changing the active base is a **pointer swap** onto a per-base `WorkspaceIndex` / store. - Do **not** await soft-sync, hard-rebuild, or a full `.rq` walk on the switch call itself. - Host surfaces rebind views to the newly active store ( status snapshot, tables, context panes ) immediately. - If the selected base is not yet ready, show its readiness / progress and **schedule** the normal activate catch-up ( open if needed, then one soft sync ) fire-and-forget — same path as ["../indexer/indexer.rq".triggers] / ["../extension/index-host/graph-analysers.rq".indexing_trigger_open], not a special switcher rebuild. - Idle staleness checks only touch bases that are already open; they must not reopen every discovered base ( ["../extension/index-host/sqlite-artifact-lifecycle.rq".release_when_idle] ). - Open failures stay isolated per base so one bad base cannot leave siblings stranded in `uninitialized`. - Editor-follow must not overwrite a user pin on the same refresh that applies the selection. - Inactive bases release SQLite artifacts when the active pointer moves away ( ["../extension/index-host/sqlite-artifact-lifecycle.rq".sqlite_artifact_lifecycle] ). Surfaces: activity bar ["../extension/local-context/activitybar-panels/workspace.rq".workspace_base_picker], Ideas Summary ["../extension/workspace-summary/ideas_summary/base-scope.rq".ideas_summary_base_scope]. Implemented by [editor_index_adapter] ( `setActiveBaseId` / `scheduleBaseCatchUp` ); engine: [base_registry] `ensureBaseReady` / `releaseArtifacts`, [workspace_index] `ensureReady` / `deactivate`. | 0.093 | |
| base_initialisation_ignore | done | done | New bases seed `<base>/.reqlan/.gitignore` so Git does not track SQLite application-memory files. Patterns cover the ideas index ( [base_ownership] ), the diagnostics store ( [app_memory_doc.index_diagnostics_store] ), and SQLite sidecars ( `*.sqlite`, `*.sqlite-journal`, `*.sqlite-shm`, `*.sqlite-wal` ). This file is Git ignore for `.reqlan/`. Path filters for discovery stay in `.rqignore` ( [extension_configuration.configuration_rqignore] / [app_memory_doc.rqignore] ). Written by [create_base_impl] as part of [base_initialisation] / ["../extension/host/installation.rq".base_installation]. | 0.182 | |
| refresh_rediscovers_bases | done | done | Manual Refresh ( command `reqlan.refreshIndex`, activity-bar / Ideas Summary Refresh buttons ) and Clear & rebuild must **always rediscover** `.reqlan` markers under workspace roots before soft-sync or rebuild. Do not skip discovery when bases are already registered — a new nested or sibling base created on disk must appear after Refresh. Discovery and registration live in [base_registry] ( `refresh` → `rediscover` → optional `ensureBaseReady` for the active base ). The extension adapter ( [editor_index_adapter] ) calls that analytical API; it must not invent a parallel discovery walk. Create Base follows the same refresh path after [create_base_impl] so the new marker is visible immediately ( [create_base_onboarding] ). Related: [multi_base_environment], ["../extension/index-host/graph-analysers.rq".indexing_trigger_manual], ["../extension/index-host/sqlite-artifact-lifecycle.rq".event_driven_base_access]. | 0.075 | |
| architecture | done | done | The cli should be a thin wrapper around `@reqlan/analytical/core` ( Rust engine ). `search` / `analyse` / `export` / `init` / `parse` must not load `@reqlan/language`. `barrel` may dynamically import the Langium barrel helper because page splits stay JS text edits ( ["../core_analysis/rust_port.rq".native_bridge] ). | 0.088 | |
| function_parity | done | done | CLI shares headless function definitions with MCP and other tools through `@reqlan/analytical` `AnalysisApi`, not VS Code host APIs from [extension]. On-disk index storage follows [application_memory] ( same `.reqlan` path as the extension ); optional override via `REQLAN_INDEX_PATH`. | 0.154 | |
| init | done | done | The cli should be able to initialise a new [base]. Creates `<path>/.reqlan/` ( empty directory is a valid marker ), matching [create_base_onboarding]. Headless helper lives in `@reqlan/analytical` so CLI and extension share one create-base path. | 0.162 | |
| pnpm_extra_args | done | done | `pnpm run <script> -- <flags>` inserts a lone `--` between the script body and the extra flags. Clipanion treats `--` as end of options, so `reqlan check -- --skip-target glob` reads `--skip-target` as a command and fails. The CLI drops lone `--` tokens from argv before Clipanion runs, so package-manager extra args stay flags. Related: [commands] ["../core_analysis/check.rq".check_meta_implementation] | 0.142 | |
| windows_esm_entry | done | done | Node's ESM loader on Windows requires `file://` URLs for absolute dynamic `import()`. A native path such as `C:\...\out\main.js` is read as protocol `c:` and crashes every command, including `init`. `bin/cli.js` converts the compiled entry with `fsPathToEsmSpecifier` ( `pathToFileURL` ). The MCP stdio entry ["../../packages/mcp/bin/mcp.js"] uses the same `pathToFileURL` conversion. | 0.083 | |
| agent_advisory | done | done | Agents must pass the [click_session] ID ( `sessionKey` ) to the next click action. This keeps one agent thread on the same session so a revisit can expand connected content and so ranking can use session distance. MCP tool text states this rule and the unique / ambiguous / search paths clearly ( [mcp_server] `click` ). | 0.101 | |
| click | done | done | The [click_ontology] function returns compact local context for one target. Input: an idea name, a `path#idea` ref, a `.rq` file path, or an indexed code file ( [click_input], [click_code_file] ). Resolution uses [click_ambiguity]: no match uses search; more than one match is ranked; one match uses [click_output]. Optional [click_max_detail] flags cap how many names and bodies are listed. Optional `sessionKey` continues a [click_session]. A second click on a centre already clicked in that session returns the connected content ( [click_session] ). Surfaces: CLI `reqlan click` ( [click_cli] ), MCP `click` ( [mcp_server] ), and [native_analysis_api] `click`. MCP search and retrieval use this tool only ( ["../extension/agent/skills-and-mcp.rq".mcp_click_retrieval] ). Engine: [click_rust] on ["./cli_package.rq".architecture] / ["../core_analysis/rust_port.rq".analytical_rust_port]. Related: [function_parity], [commands], [agent_advisory], [click_session], [click_input], [click_max_detail], [click_output]. | 0.200 | |
| click_ambiguity | done | done | Engine function that classifies a name or target as none, unique, or ambiguous. A name is ambiguous when more than one indexed idea or indexed file shares that exact name. Names are unique inside one file ( [idea_name] ); ambiguity is across files. `path#idea` and a full indexed path are unique when that record exists. Click uses this function to choose the [click_input] path and to decide when [click_output] prints a file path next to a name. Outbound [wildcard_reference] targets are ambiguous: print their matching paths. Surfaces: AnalysisApi `checkNameAmbiguity` ( engine ). Click must call it. CLI / MCP exposure is not required. | 0.103 | |
| click_code_file | done | done | An indexed code file is a first-class click target ( [index_code_files], [referenced_files] ). Click accepts the same path forms as [file_search] ( relative path, basename when unique ). Unique output for a code file: - the file path ( no idea body ) - count and names of backlinks: ideas with a [file_reference] to this file; print the idea file when the name is ambiguous - count and names of [comment_reference] targets ( `comment_link` from this file ); print the idea file when the name is ambiguous - siblings do not apply Data is the same indexed edges as [file_related_requirements]. | 0.195 | |
| click_input | done | done | Accept any of: - idea name ( trimmed optional surrounding brackets `[name]` ) - `path#idea` ( indexed file URI + idea name ) — this form is unique - `.rq` file path — the file is one centre, not every idea in the file - indexed code file path ( [click_code_file] ) Resolution uses [click_ambiguity] after the same ref normalisation as ( ["../core_analysis/search.rq".contextual_search] / `resolve_search_context_refs` ). When no centre matches, click uses [fuzzy_search] ( and [file_search] ) with the target as the query and returns those hits. Do not return an empty slice. When more than one centre matches, and the [click_session] already has ideas, rank the matches by undirected hop distance from those ideas ( [idea_graph_distance] / [contextual_search] ). Nearer ranks first. Put the hop count on each candidate. When there is no session key, or the session has no ideas yet, do not calculate hop distance. Do not print hops. Rank by name and file path. On the search path, rank by [fuzzy_search] score only. When exactly one centre matches, return [click_output]. Always return or create a [click_session] key. | 0.156 | |
| click_max_detail | done | done | Flags that cap how many names and bodies click lists. Invalid or missing values use the default. Values less than 1 use 1. Defaults live in `.reqlan/config.json` under `click` ( [configuration_location], [configuration_schema_file] ). CLI and MCP flags override config. - `maxBacklinks` / `--max-backlinks` / `click.maxBacklinks` — first n backlink names in [click_output], then " + m more ". - `maxSiblings` / `--max-siblings` / `click.maxSiblings` — first n sibling names, then " + m more ". - `maxOutbound` / `--max-outbound` / `click.maxOutbound` — first n outbound names or paths, then " + m more ". - `maxCandidates` / `--max-candidates` / `click.maxCandidates` — search hits and ranked ambiguous matches. The default for each list is `click.listLimit` in ["../../packages/analytical/src/core/defaults.json"]. `maxDetail` / `--max-detail` is the old hop-depth for a neighbourhood dump. Unique [click_output] does not use hop depth. The flag is still accepted and ignored on the unique path. | 0.122 | |
| click_output | done | done | The result always includes `sessionKey` ( [agent_advisory] ). The result does not include an `edges` list. Links are implied by the lists below. Shape depends on [click_input]: Unique ( one centre ): - the idea ( or file ) and its content ( indexed unmarked body / summary ) - count of outbound links from that content only: idea refs and [file_reference] that occur in the unmarked body. Do not list refs that occur only in `@tests`, `@implementation`, or `@todo`. Do not list [comment_reference] code files as outbound. List names. When [click_ambiguity] says the name is ambiguous, print the file path with the name. File names are unquoted basenames. - count of backlinks and the first n names ( [click_max_detail] ); print the file when the name is ambiguous or shared; then " + m more " for the remainder - count of siblings ( same [ideaset] / file ), not the ideas named in the content, and the first n names, with file when the name is ambiguous or shared - for a code-file centre, see [click_code_file] Ambiguous ( more than one match ): ranked candidate names with file paths. Include hop distance only when the session has prior ideas. No idea bodies. Cap with `maxCandidates`. Search ( no exact match ): [fuzzy_search] / [file_search] hits. Cap with `maxCandidates`. Revisit: see [click_session]. Show content of the connected nodes, not the centre body again. | 0.117 | |
| click_session | done | done | Click sessions live in SQLite under [application_memory] as `click-sessions.sqlite` ( sibling of `ideas-index.sqlite`; not deleted by ideas-index rebuild ). Each session has a stable `sessionKey`. Sessions are keyed against ideas. Each surfaced idea stores a fingerprint of its name, summary, status, and tags. When that fingerprint changes, the idea may resurface on the next click. Record clicked centres ( unique or revisit targets ) separately from names that were only listed. When the target is a centre already clicked in this session, return the content of the connected nodes ( outbound, backlinks, and siblings that the first click listed ), capped by [click_max_detail]. Do not repeat the centre body. Do not return an empty suppressed slice. Skip a connected body that this session already showed as content, unless its fingerprint changed. Session cache size is [click_session_limit]. Eviction uses last-touched order: when the store exceeds the limit, drop the least recently touched sessions first. Touch the active session on every successful click. Path helper: ["../../packages/analytical/src/core/application-memory.ts"] / [click_sessions_rust]. Layout: ["../indexer/sidecars.rq".reqlan_directory_layout]. | 0.225 | |
| click_session_limit | done | done | Maximum stored click sessions per base is `click.maxSessions` in ["../../packages/analytical/src/core/defaults.json"]. Configure with `click.maxSessions` in `.reqlan/config.json` ( [configuration_location], [configuration_schema_file] ). Invalid or missing values use that default. Values below `click.minimum` use `click.minimum`. Eviction is by session count and last-touched time ( [click_session] ). Count caps for one click result are [click_max_detail], not this limit. | 0.070 | |
| diff | done | done | Compare ideas in two git snapshots. The key is the file path plus the [idea_name]. The command reads `.rq` files from git and parses each snapshot. It does not read or write the ideas index. Line position does not count. The same body and attributes at a different line are unchanged. Content is the idea body plus its attributes. Change kinds: - created — the key exists only on the right snapshot - deleted — the key exists only on the left snapshot - modified — the key exists on both snapshots and the content differs - moved — the name exists in exactly one file on each side, the paths differ, and the content is the same The same name in more than one file on one side stays on the path-plus-name key. A renamed idea is deleted plus created. Unchanged ideas are omitted. A reference-only edit stays modified. No arguments: if any `.rq` path is staged, unstaged, or untracked and not gitignored, the left side is `HEAD` and the right side is that pending snapshot. Otherwise the left side is `HEAD~1` and the right side is `HEAD`. A non- `.rq` change does not select the pending snapshot. One argument: that revision against `HEAD`. Two arguments: the first revision against the second. A target is a branch, a commit, `HEAD`, `~N` ( `HEAD~N` ), or `<rev>~N`. The pending snapshot uses the working-tree file for an unstaged edit, the index blob when the working tree matches `HEAD`, and the working-tree file when a path is both staged and unstaged. An untracked `.rq` file that git does not ignore is included. If `HEAD` has no parent, the left snapshot is empty and every idea on `HEAD` is created. A missing revision or a directory that is not a git repository is an error. Exit status is 2. A successful diff exits 0, including an empty diff. When a side is filled by a default, the human CLI and [diff_webview] print a header `Comparing <from> -> <to>` before the rows. Explicit two-argument runs omit that header. `--json` and MCP stay a row array only. Surfaces: CLI `reqlan diff` ( [commands] ), MCP [skills_and_mcp.mcp_diff], and [diff_webview]. Engine: ["../../crates/reqlan-analytical/src/diff.rs"]. Git spawns stay hidden ( [consumption_silence] ). Related: [architecture] [function_parity] [idea] [idea_reference] [diff_verbosity] [diff_cli_color] | 0.163 | |
| diff_cli_color | done | done | Color the human text when stdout is a terminal and `NO_COLOR` is unset. `--color` forces color. `--no-color` forces plain text. `--json` is never colored. Created and added lines are green. Deleted and removed lines are red. Modified is yellow. Moved is cyan. | 0.061 | |
| diff_verbosity | done | done | Three levels. The default is summary. CLI: `--verbosity summary|references|text`, or `-v` / `-vv`. MCP and [diff_webview] use the same three values. - summary — change kind, idea name, and file. `fromFile` only for moved. - references — summary, plus [idea_reference] targets added on the right and removed from the left. Order each list by name. Omit a target that is on both sides. - text — summary, plus the body and attributes before and after. Created shows the right side. Deleted shows the left side. Modified shows both. Moved has no text change. Do not list reference names at this level. Attributes stay inside modified. The references level does not print attribute text. The text level prints body lines, then attribute lines as `@name` / `@name value`. `--json` omits `references` and `text` at summary. It adds `references` at references only. It adds `text` at text only. | 0.071 | |
| diff_webview | done | done | Command palette title Diff Ideas, category reqlan, id `reqlan.diffIdeas` ( [features_commands.command_palette] ). The panel takes the same two targets and the same verbosity as [diff]. Controls sit in a sticky themed strip: From, To, Detail, Diff. Inputs and the primary button use editor input and button theme tokens. Enter in a control runs Diff. The panel runs once on open with empty targets. Each row uses the editor git decoration colors for created, deleted, and modified, and a distinct color for moved. A row shows kind, idea name, and path as separate fields. Moved shows both paths. The references level lists added and removed reference names. The text level shows monospaced before and after body blocks; modified uses a two-column layout when width allows. Empty results say there are no idea changes. Errors use validation theme colors. Diffing disables the Diff control and shows a loading status. The panel calls the shared engine. It does not parse git output. When a side uses a default, the panel shows the same `Comparing <from> -> <to>` header as the CLI under the strip. | 0.083 | |
| derived_host_bindings | done | done | Components that display host-driven AppState fields must bind through Svelte 5 `$derived` / `$derived.by` ( not legacy `$:` ) so updates from the message listener re-render reliably. Keep ephemeral widget-only UI ( open menus, local filter drafts, selection ) as component-local `$state`. Do not mirror host payloads into a second parallel store inside the component. | 0.040 | |
| host_webview_ownership | done | done | The extension host owns durable data, index / analysis queries, and VS Code APIs. The webview owns presentation state and posts intents ( search, navigate, persist UI prefs, export ). Host session objects ( e. g. activity-bar context session ) mutate on the host and push snapshots to the webview; the webview must not invent a second source of truth for the same domain fields. Typed message modules define the contract; keep request ids / error scopes on async host calls. | 0.050 | |
| imperative_view_controllers | done | done | Heavy imperative widgets ( Cytoscape graph ) keep lifecycle state on a controller class ( `GraphCyController` ), not inside AppState `$state` fields. AppState holds query / slice / loading / error and persisted UI prefs; the component holds the controller instance. See ["../extension/presentation/graph.rq".webview_threading] and ["../extension/workspace-summary/ideas_summary/graphical_graph.rq".graph_loading]. | 0.040 | |
| persistence_tiers | done | done | Three persistence tiers — pick the shallowest that matches lifetime: 1. `acquireVsCodeApi().getState` / `setState` — ephemeral webview session ( pane expand / heights, active tab ). 2. `ExtensionContext.workspaceState` — durable UI prefs shared across reloads ( graph / table UI keys under `reqlan.ideasSummary.*`, optionally per-base ). Normalize + default on read ( `normalizeGraphUiState` / table defaults ); never trust raw JSON shape. 3. `ExtensionContext.globalState` — install / user-level flags ( e. g. onboarding ), not pane view-models. Debounce persist posts ( ~ 200 – 250 ms ) when UI controls churn. Related: [webview_app_state]. | 0.050 | |
| postmessage_clone_safety | done | done | Before `postMessage` from a Svelte webview, JSON-clone payloads ( `JSON.parse(JSON.stringify(...))` ) so `$state` proxies remain structured-clone safe. Shared helpers live in each webview's `lib/vscode.ts`. | 0.040 | |
| svelte_app_context | done | done | Webview components read AppState through Svelte context helpers ( `setAppContext` / `getApp` with a `Symbol` key ) in `state/context.ts` — not by importing the singleton from deep child leaves when a context is available. Root `App.svelte` sets context once after constructing / using `app`. | 0.049 | |
| ui_state_management | done | done | Extension UI state is explicit, typed, and layered — not a single global store shared across host and webview. Choose the layer by lifetime and owner: - Host session / durable prefs live in the extension process ( `ExtensionContext`, plain modules ). - Webview view-models live in Svelte 5 runes ( `$state` / `$derived` / `$effect` ). - Cross-boundary traffic uses typed `postMessage` contracts only. Prefer small state machines ( named phases, request ids, clear transitions ) over ad-hoc flags. Related: [webview_app_state], [persistence_tiers], [zustand_boundary], [host_webview_ownership]. | 0.080 | |
| webview_app_state | done | done | Complex webviews ( activity bar, Ideas Summary ) own a single `AppState` class in `state/app.svelte.ts`. Fields that drive UI are Svelte 5 `$state` properties on that class; side effects and host I / O are methods. Export one module singleton ( `export const app = new AppState()` ) and inject it via [svelte_app_context]. Do not use `svelte/store` writable / readable stores for new UI, and do not put zustand inside webview code ( [zustand_boundary] ). Simple panels ( export form, onboarding, index diagnostics ) may keep `$state` on the root `App.svelte` instead of a separate AppState module when the surface is a single form / flow. Related: ["../extension/workspace-summary/ideas_summary/webview.rq".table_column_filters] `$derived` binding rule; ["../extension/workspace-summary/ideas_summary/graphical_graph.rq".graph_ui_persistence]. | 0.070 | |
| zustand_boundary | done | done | [extension_architecture.state_machine_mindset] still applies: design UI and host flows as state machines. Zustand is no longer used anywhere: the index lifecycle FSM ( `opening` / `syncing` / `ready` / `error` ), counts, file issues, and the document-update ring now live in Rust `WorkspaceIndexRuntime` and are read over napi via `statusSnapshot()` ( ["../core_analysis/rust_port.rq".native_bridge] ); durable diagnostics persist in `index-diagnostics.sqlite`. TypeScript keeps no parallel copy of that state. Do not reintroduce zustand ( or Redux-like globals ) in `packages/extension/webviews/**` or the analytical package. Extension-host TypeScript may keep plain mutable session modules when the surface is small and message-pushed; owned lifecycle / state machines belong in the Rust runtime. | 0.069 | |
| analytical_package | in-progress | in-progress | `packages/analytical` ( `@reqlan/analytical` ) is the shared requirement graph index and analysis layer. It owns the [indexer]: extracts ideas and edges into a rusqlite store via the core native engine ( [workspace_index] ), holds Zustand analytical state, and exposes analysers and an analysis API. Headless CLI / MCP entry is `@reqlan/analytical/core` via `openAnalysisApi` ( native required; no Langium ). Consumers: [extension_package] ( thin editor adapter ), [mcp_package], [cli_package]. Depends on [language_package] for the editor LSP only. Published to npm alongside [cli_package]; version graph role: [package_versioning]. Rust engine port ( lexer / parser-for-index, indexer, search, export ): [rust_port.analytical_rust_port], shipped per [distribution.rust_binary_distribution]. | 0.167 | |
| check | done | done | `reqlan check` should run perform the following checks: - ensure all rq [reference] point to real nodes, unless rq ignored with [comment_reference_ignore]. - this tests rqcomments point to real parts of the namespace, that code references exist, and that rq references exists. - report unresolved imports: [check_unresolved_imports]. - report a comment reference with invalid syntax: [code_comment_syntax_errors]. - report a qualified reference whose file is not on disk: [check_qualified_missing_file]. - report a qualified reference whose `.rq` file does not declare the idea: [check_qualified_missing_idea]. - it should really just be a wrapper of the [index] and return the issues ( as a pipe, terminal output, or json ). Issue order: [check_order_by_target]. Sparse wildcards: [check_wildcard_sparse] via flags [check_wildcard_zero] and [check_wildcard_one]. Path globs: [check_glob_patterns]. Skip missing targets: [check_skip_targets] [check_skip_gitignored_targets]. CLI: ["../cli/cli_package.rq".commands]. MCP: ["../extension/agent/skills-and-mcp.rq".mcp_check]. CI: [check_meta_implementation]. Site demo: ["../site/site.rq".broken_links_showcase]. | 1.000 | |
| check_glob_patterns | done | done | [check] accepts more than one `--glob` pattern. Repeat `--glob` for each pattern. A file is in the recursive scan when it matches any glob. One `--glob` selects the same files as one path glob. A reference target that matches no glob is still checked. That target must resolve. [check] does not walk a directory that matches no glob. A positional path can name a file outside the globs. [check] checks that file. A positional path that is a directory outside the globs is not walked. An empty `--glob` value is an error. API: `AnalysisApi.check({ pathGlobs, paths })`. CLI: ["../cli/cli_package.rq".commands]. MCP: ["../extension/agent/skills-and-mcp.rq".mcp_check]. Related: [check] | 0.308 | |
| check_meta_implementation | done | done | [check] is a step in the sequential GitHub CI job after one workspace build. Pull requests use ["../../.github/workflows/ci.yml"]. Main uses the same workflow through ["../../.github/workflows/release.yml"] ( ["../distribution/distribution.rq".ci_gate] ). The job installs workspace packages once ( `pnpm install --frozen-lockfile` ), asserts `@reqlan/*` resolve under `packages/` ( ["../../scripts/assert-workspace-reqlan.mjs"] ), compiles crates and TypeScript, then runs crate tests, `node packages/cli/bin/cli.js check`, and JS / TS tests against that build. Do not install `@reqlan/cli` or other `@reqlan/*` packages from npm. Platform natives in this workspace come from `packages/analytical-native/` via `packageExtensions` in ["../../pnpm-workspace.yaml"], not from the registry. Do not call a global `reqlan` binary. Do not run a second install before check. GitHub and Azure pass `--skip-gitignored-targets` because gitignored files are not in a CI checkout ( [check_skip_gitignored_targets] ). Azure uses the same workspace CLI after its package build on `ext/v*` tags. Related: [check] [check_skip_targets] [check_skip_gitignored_targets] ["../distribution/distribution.rq".ci_gate] | 0.355 | |
| check_order_by_target | done | done | [check] returns issues ordered by the missing [reference] target, not by the source file. Rows that share the same missing idea, comment target, file path, or import target stay together so you can see every site that points at one missing thing. Inside a target group, order by kind, then source file, then line. JSON and pipe keep a flat list in that order. Terminal output uses one heading per missing target. Related: [check] ["../cli/cli_package.rq".commands] | 0.370 | |
| check_qualified_missing_file | done | done | A qualified [reference] names a file and an idea. Example: `["../../../../my/non-existentfile.ts".term]`. [check] reports an error when that file is not on disk. The issue kind is `file_reference`. The label is the resolved file path. A qualified reference to a file that exists is not this error. A same-file idea reference is not this error. The editor already underlines this target. Related: [check] [check_qualified_missing_idea] ["../language/syntax.rq".reference_file] ["../extension/language/syntax/features-syntax-highlighting.rq".unresolved_reference_diagnostics] | 0.211 | |
| check_qualified_missing_idea | done | done | A qualified [reference] names a file and an idea. Example: `["../extension/index-host/graph-analysers.rq".not_an_idea]`. The same rule applies to a namespace alias: `[alias.idea]` ( [import_namespace] ). [check] reports an error when that file is an `.rq` file, the file does not declare the idea, and the file does not bind that name as an import alias. The issue kind is `references`. The label is the missing idea name. A path with no extension means `.rq` first ( [implicit_file_extension] ). A qualified reference to an idea that the `.rq` file declares is not this error. A qualified reference to an import alias bound in that file is not this error. A namespace import binds its `as` name. A from-import or qualified import binds its `as` name, or the idea name when `as` is absent. A from-import of an import alias stays [check_unresolved_imports]. The alias is not an idea. A qualified reference to a code file is not this error. That form uses [check_qualified_missing_file]. A missing file stays [check_qualified_missing_file]. [check] does not also report a missing idea for that file. [comment_reference_ignore] suppresses this error. Related: [check] [check_qualified_missing_file] ["../language/syntax.rq".reference_qualified] | 0.278 | |
| check_skip_gitignored_targets | done | done | [check] omits a file [reference] issue when Git ignore rules ignore the missing target. The analyser function `check_references` takes `skip_gitignored_targets` ( default false ). Match uses the workspace ["../../.gitignore"] and nested `.gitignore` files on the target path. Match uses the resolved path from the workspace root, not the authored path in the source file. Idea references, comment references, and missing imported idea names stay in the issue list. Missing import paths follow the same skip as file references ( [check_unresolved_imports] ). Use this when a referenced file is build output or editor config that git does not store. Examples: `.cursor/` for ["../extension/agent/chat-participant.rq".mcp], and `packages/extension/media/webviews/` for ["../extension/host/onboarding/page-thanks-for-installing.rq".installation_event]. CLI: `--skip-gitignored-targets` ( ["../cli/cli_package.rq".commands] ). API: `AnalysisApi.check({ skipGitignoredTargets })`. CI: ["../../.github/workflows/ci.yml"] and ["../../azure-pipelines.yml"] pass `--skip-gitignored-targets` because a CI checkout does not contain gitignored files ( [check_meta_implementation] ). Related: [check] [check_skip_targets] [check_unresolved_imports] [check_meta_implementation] | 0.445 | |
| check_skip_targets | done | done | [check] omits an issue when the missing [reference] target matches a skip glob. The analyser function `check_references` takes `skip_targets`: a list of globs. Match uses the same path glob rules as `--glob`. Match is against the issue `label` ( the missing target ), not the source file. An empty skip glob has no effect. Use this when a referenced file is not in git and you name the skip glob. For all Git ignore rules, use [check_skip_gitignored_targets]. CLI: repeatable `--skip-target <glob>` ( ["../cli/cli_package.rq".commands] ). API: `AnalysisApi.check({ skipTargets })`. Related: [check] [check_skip_gitignored_targets] [check_meta_implementation] | 0.445 | |
| check_unresolved_imports | done | done | [check] reports an error when an import does not resolve. A missing import path is an error. A named imported idea that is not in the target file is an error. Do not report imported idea names when the path is missing. An existing folder is a valid import path ( [import_folder_targets] ). A remote URL is not a missing path. Import paths with no extension mean `.rq` first ( [implicit_file_extension] ). Ignore a site with [comment_reference_ignore]. Issue kind is `import`. The `label` is the missing path or the missing idea name. Forms: [import_from], [import_namespace], [import_qualified]. Editor underline: [import_does_not_exist_error]. [check_skip_targets] matches that label. [check_skip_gitignored_targets] omits a missing import path that Git ignore rules ignore. Related: [check] [check_order_by_target] [import_does_not_exist_error] | 0.425 | |
| check_wildcard_one | done | done | `check_references` flag `wildcard_one` ( default `warn` ) controls a ["../reference_types.rq".wildcard_reference] that matches exactly 1 idea. 1 match means the pattern is too narrow — use a concrete [reference] instead. `warn` — emit the row with severity `warning` and match count 1. `error` — emit the row with severity `error` and match count 1. `off` — omit the row. CLI: `--wildcard-one <warn|error|off>` ( ["../cli/cli_package.rq".commands] ). API: `AnalysisApi.check({ wildcardOne })`. Related: [check] [check_wildcard_sparse] [check_wildcard_zero] ["../language/imports.rq".wildcard_references] | 0.507 | |
| check_wildcard_sparse | done | done | A ["../reference_types.rq".wildcard_reference] is for many matches. [check] recounts each wildcard against the full idea catalog, not the stored edges, so files that the indexer visited first still get the true match count. How 0 - match and 1 - match rows are handled is controlled by [check_wildcard_zero] and [check_wildcard_one] ( both default `warn` ). Ignore a site with [comment_reference_ignore]. The language server already warns when a wildcard matches nothing ( ["../language/imports.rq".wildcard_references] ); it also warns when the match count is 1. Related: [check] ["../language/imports.rq".wildcard_references] | 0.417 | |
| check_wildcard_zero | done | done | `check_references` flag `wildcard_zero` ( default `warn` ) controls a ["../reference_types.rq".wildcard_reference] that matches 0 ideas. 0 matches means the pattern hits nothing. `warn` — emit the row with severity `warning` and match count 0. `error` — emit the row with severity `error` and match count 0. `off` — omit the row. CLI: `--wildcard-zero <warn|error|off>` ( ["../cli/cli_package.rq".commands] ). API: `AnalysisApi.check({ wildcardZero })`. Related: [check] [check_wildcard_sparse] [check_wildcard_one] | 0.507 | |
| code_comment_syntax_errors | done | done | A comment reference uses `rq:[idea]` or `rq:["path".idea]`. A path string uses double quotes or single quotes. A qualified reference may use dotted idea names after the path. Line comments use `//` and `#`. The same forms apply in every comment span that the indexer reads. [check] reads each comment reference in a non- `.rq` source file. A comment reference with invalid syntax is a syntax error. [check] reports the syntax error. The issue kind is `comment_syntax`. The issue label is the error reason and the comment reference text. The issue severity is `error`. A valid comment reference stays a ["../language/syntax.rq".comment_reference] site. An unresolved valid target stays a `comment_link` issue. A source file that contains only a syntax error stays in the code index. [comment_reference_ignore] suppresses the syntax error on the next line. Related: [check] ["../extension/language/comment-references/functional-code-comment-references.rq".references_in_functional_code_comments] | 0.174 | |
| consumption_silence | done | done | Hosts that consume the native core ( and helper CLI tools such as `git` / `node` ) must not open a visible console window. On Windows, a GUI editor host that spawns a console-subsystem process without `CREATE_NO_WINDOW` / `windowsHide` flashes a CLI window. Git-dates fill in the core engine and Activity Bar git history were the repeating flashes after `.rq` change; the language-server Node probe can flash once at start on old hosts. Index parse itself stays in-process ( napi `.node` / `.dll` ) and does not spawn a console. Related: ["../indexer/indexer.rq".git_dates], ["../extension/index-host/git-codelens.rq".git_dates_background_indexing], ["../extension/index-host/git-codelens.rq".git_dates_rate_cap], ["../extension/local-context/context-scope.rq".git_context], ["../extension/index-host/index.rq".git_information_capture]. | 0.131 | |
| test_references | done | done | The indexer lists broken references in a [index] base. A broken reference is an unresolved idea or wildcard reference. You can give an optional path glob so that the list covers only a subset of the base. You can optionally include unresolved [comment_reference] sites. The function is in the core analysis module. The CLI exposes it on [commands] ( `analyse --broken-refs` ). Related: [analysers] [runtime] [indexer.index_code_files] ["../language/syntax.rq".comment_reference_resolution_error] | 0.201 | |
| csv_export | done | done | CSV export writes `ideas.csv` with tags and non-status / tag attributes flattened into columns ( `attr:<key>` ), plus `references.csv` for inbound / outbound / unresolved edges. Shared base / scope / output / name and file filters apply. Pipeline: `exportCsv` / `writeCsvExport` ( [export_pipeline] ). Extension UI host: ["../extension/export/export.rq".export_webview]. | 0.055 | |
| export_pipeline | done | done | `@reqlan/analytical` owns export execution for every format. Shared contract: `ExportRequest` / progress callbacks / snapshot build, then format writers. Hosts ( extension webview, CLI ) assemble an `ExportRequest` and call `NativeAnalysisApi.exportHtml` / `exportMarkdown` / `exportJson` / `exportCsv` — they do not reimplement writers. Formats: ["./html_export.rq".html_export], [markdown_export], [json_export], [csv_export], [pdf_export]. Progress phases ( snapshot, then write completed / total ) are reported to hosts for loading UI. Consumed under [analysis_scope]. Rust port of snapshot + writers: ["./rust_port.rq".export_rust]. | 0.114 | |
| json_export | done | done | JSON export writes `export.json` for the selected base: ideas ( with attributes and reference lists ), files, clusters, attributes, and status / tag rollups — suitable for tooling and AI consumption. Shared base / scope / output / name and file filters apply. Pipeline: `exportJson` / `writeJsonExport` ( [export_pipeline] ). Extension UI host: ["../extension/export/export.rq".export_webview]. | 0.055 | |
| markdown_export | done | done | Markdown export writes a `README.md` index for the selected base ( counts, status / tag rollups, idea list, file list ) and optional per-idea `.md` pages under `ideas/` with attributes and inbound / outbound references as relative markdown links. When per-idea pages are disabled, idea bodies are inlined in the README. Shared base / scope / output / name and file filters apply; HTML-only page families ( graph, clusters, attributes, print ) are omitted. Pipeline: `exportMarkdown` / `writeMarkdownExport` using the same snapshot builder as HTML ( [export_pipeline] ). Extension UI host: ["../extension/export/export.rq".export_webview]. | 0.055 | |
| pdf_export | partial | partial | PDF export prepares a print-oriented HTML report from the selected base so the user can Print → Save as PDF in the browser. Direct PDF binary rendering is not required yet; preserve links in the printable HTML ( ["./html_export.rq".html_export_printable_pages] ). PDF-specific options include file filters and print entry filename; shared base / scope / output / name apply. Extension UI host: ["../extension/export/export.rq".export_webview] ( not a separate QuickPick chain ). | 0.073 | |
| html_export | done | done | `@reqlan/analytical` builds documentation as HTML from the requirement graph ( ["./export.rq".export_pipeline], [analysis_scope] ). Html export should render ideas, attributes, references, and navigable structure suitable for sharing. Pipeline: `NativeAnalysisApi.exportHtml` / `AnalysisRuntime.export` → `reqlan-export` snapshot ( `build_export_snapshot` ) → multi-file static site ( `write_html_export` ). Single-file mode remains optional via [html_export_runtime_modes]. Interactive graph / search JS stays bundled static assets ( [html_export_graph_layout] ), not a second writer ( ["./rust_port.rq".export_rust] ). Thin hosts: extension UI ( ["../extension/export/export.rq".export_webview], ["../extension/export/html-export.rq".html_export_form], ["../extension/export/export.rq".export_html] ) and CLI ( `reqlan export` / ["../cli/cli_package.rq".commands] ). Export is scoped to a selected ["../bases/base.rq".base] ( not a merged multi-base workspace graph ). [html_export_multi_file] [html_export_internal_links] [html_export_idea_tooltips] [html_export_layout_cards] [html_export_heading_tooltips] [html_export_table_columns] [html_export_search] [html_export_list_views] [html_export_overview_links] [html_export_idea_pages] [html_export_file_pages] [html_export_code_reference_pages] [html_export_cluster_pages] [html_export_graph_page] [html_export_attributes_index] [html_export_attribute_pages] [html_export_printable_pages] [html_export_print_idea_cards] [html_export_print_body_whitespace] [html_export_print_in_page_anchors] [html_export_print_all_attributes] [html_export_print_button] [html_export_print_order_by] [html_export_runtime_modes] [html_export_scrollable_lists] [html_export_header_link] [html_export_url_base] [html_export_katz_column] | 0.201 | |
| html_export_graph_background_load | done | done | Interactive HTML export graph boot and layout settle must not monopolise the main thread on large graphs. Page chrome ( nav, filters, search, other controls ) should stay responsive while the graph initialises. Graph wiring should start after the first paint / idle opportunity rather than synchronously blocking `DOMContentLoaded`. When Live physics is off, the initial batch settle should run in yielded chunks ( animation frames or idle callbacks ) so the browser can paint and handle input between steps; keep the boot spinner until the first usable canvas frame, then show progressive settle status. A new filter / refresh or remount must cancel any in-flight background settle before starting another. Cross-refs [html_export_graph_page], [html_export_graph_animation], [html_export_graph_interaction]. | 0.093 | |
| html_export_graph_label_auto | done | done | Export-surface wiring for [graph_label_auto]: canvas paint applies continuous zoom opacity ( GRAPH_LABEL_FADE_START → GRAPH_LABEL_FADE_END ); hover / drag stays fully opaque. Labels stay fully invisible while the graph is initialising or batch-settling so text paint does not lag those phases; zoom fade applies only after settle completes. Zoom and Fit from [html_export_graph_interaction] / [manual_reframe] update opacity each paint. | 0.081 | |
| html_export_graph_label_modes | done | done | Export-surface wiring for [graph_label_modes]: Labels control on every interactive export graph cycles auto / on / off; Reset restores auto. | 0.070 | |
| html_export_header_link | done | done | Html export should accept an optional header link ( href + label ) rendered in the topbar ahead of section navigation. Hosts such as the marketing site build can set this so readers can return to the parent site from any exported page. When omitted, the topbar shows only the export section nav. Site embed consumption: ["../site/site.rq".spec_html_export]. Often paired with [html_export_url_base] when the export is mounted under a parent static site. | 0.222 | |
| html_export_heading_tooltips | done | done | Section headings on layout panels show a short descriptive tip via the styled tooltip panel when pointer or keyboard focus rests on the heading. Entity-card title headings tip with that card's short summary text. Idea title headings tip with the idea summary when present. Heading tips must not use the browser title text. Reuses the tooltip surface from [html_export_idea_tooltips]. Cross-refs [html_export_layout_cards]. | 0.092 | |
| html_export_idea_tooltips | done | done | A link to an idea must show a tooltip when the pointer or keyboard focus rests on the link. The tooltip must show the idea name, the source file, and the idea body text. The tooltip must show the status when the idea has a status. The tooltip must be a styled panel. The tooltip must not use the browser title text. The tooltip must stay inside the viewport. An unresolved reference must not show an idea tooltip. Section and card heading tips follow [html_export_heading_tooltips]. | 0.099 | |
| html_export_ideas_ref_columns | done | done | The ideas index table shows inbound reference counts and outbound reference counts in two columns. The first count column is In. The second count column is Out. Each cell shows one integer. Each column has its own header sort and column filter. Filter keys are `in` and `out`. | 0.056 | |
| html_export_katz_column | done | done | HTML export shows the inbound Katz score ( ["../modelling/advanced_analysis.rq".katz_centrality] ) for each idea. The ideas index adds a Katz column after Out. File idea tables, code-file idea tables, cluster member tables, and attribute idea tables add a Katz column. The printable file sheet, code-file sheet, and cluster sheet add the same column. The idea page and the printable idea sheet show the same score as a metric. The cell text uses 3 decimal places. The sort value is the full score. The filter key is `katz`. File, cluster, and attribute index tables do not show this column. | 0.075 | |
| html_export_layout_cards | done | done | In interactive runtime mode, detail-page panels ( panel, table-shell, graph-shell ) and entity-card tiles are drag-reorderable within their layout group and collapsible. In interactive mode, cards that are direct children of a page layout group and use panel, table-shell, or graph-shell chrome are width-resizable by column span. Entity-card tiles inside an entity-list layout group are drag-reorderable and collapsible, but they are not width-resizable. Column hide and column width on tables stay separate from card width. Width-resizable cards sit in a responsive multi-column grid and snap to column spans so a card can take one column, several columns, or the full row. Double-click on the card width handle fills the row. A second double-click restores the last non-full span. Collapsed cards ignore fill. Narrow viewports collapse spans so cards stack without horizontal overflow. The full card header is the drag handle. Links, buttons, and inputs in the header still activate on click. The collapse caret stays hidden until the card is hovered or has focus within. Collapse hides the card body and shrinks the card chrome so the layout does not keep empty body space. Card-width and header drag must not lose control to text selection: pointer capture, select lock, and preventDefault on resize pointerdown. Order, collapse, column-span, and fill-restore state persist in localStorage for that page path. Document and print runtime modes keep a static layout with no rearrange, resize, or collapse behaviour. Printable-report idea cards stay outside this idea ( [html_export_print_idea_cards], [html_export_print_order_by] ). Table column chrome follows [html_export_table_columns]. Cross-refs [html_export_runtime_modes], [html_export_heading_tooltips]. | 0.092 | |
| html_export_overview_links | done | done | The interactive overview page status rollup, tags rollup, and highlighted clusters must be interactive navigation, not plain labels. Status and tag rollup entries should open the matching status / tag cluster page when [html_export_cluster_pages] is enabled; otherwise they should open the ideas list with the Status or Tags column filter seeded from the rollup key. Highlighted clusters should open the matching cluster page ( or the clusters index when cluster pages are disabled ). List pages must honour URL query parameters that seed per-column filters ( for example `ideas.html?status=todo` or `ideas.html?tags=export` ) so overview and in-table facet links land with the filter already applied. Query keys match stable `data-filter-key` values on table headers; filter values use the same display labels as the rollup ( including Not present / Empty ). Cross-refs [html_export_list_views], [html_export_cluster_pages], [html_export_internal_links]. | 0.103 | |
| html_export_print_all_attributes | done | done | Printable report idea cards and per-idea printable sheets must show every declared attribute for the idea, not only status ( or status + tags ). Attribute values that contain idea refs follow the link rules of the hosting print surface ( [html_export_print_in_page_anchors] on the report page ). Cross-refs [html_export_printable_pages], [html_export_idea_pages]. | 0.110 | |
| html_export_print_body_whitespace | done | done | The printable report must show the idea body text that the parser stores. Each per-idea printable sheet must show the stored idea body text. The parser removes the base indent. The base indent is the indent of the first body line. Extra indent on a later line is part of the body text. A later line with 4 more spaces than the first body line is stored with those 4 spaces. Print HTML must keep those leading spaces. Print HTML must not collapse those spaces. The body text must still wrap inside the card width. Cross-refs [html_export_printable_pages], [html_export_print_idea_cards]. | 0.110 | |
| html_export_print_button | done | done | Printable report and printable idea sheets include a Print control that opens the browser print dialog ( `window.print` ). The control must work without the interactive javascript bundle ( inline handler is acceptable ) and should be hidden in the printed output. Cross-refs [html_export_printable_pages]. | 0.090 | |
| html_export_print_idea_cards | done | done | The printable report page lists ideas as cards in a single column that must remain within the page width on narrow viewports — no horizontal spill from long summaries, file paths, refs, or attribute values. Cards use a one-column grid ( `minmax(0, 1fr)` ) and wrapping / overflow rules so content reflows instead of overflowing the layout. Stored idea body spaces on these cards follow [html_export_print_body_whitespace]. Cross-refs [html_export_printable_pages], [html_export_print_in_page_anchors], [html_export_print_all_attributes]. | 0.153 | |
| html_export_print_in_page_anchors | done | done | On the printable report page, each idea card has a stable html `id` derived from the idea identity. Cross-idea links rendered on that page ( title links and inline idea refs in summaries / attributes ) must resolve to same-page `#` fragment urls targeting those ids, not separate idea detail pages. When browsed with the sticky site topbar, fragment navigation must leave the target card fully visible below the header ( scroll padding / scroll-margin ), not clipped behind it. Aligns with stable in-page anchors in [html_export_internal_links]. Cross-refs [html_export_printable_pages], [html_export_print_idea_cards]. | 0.145 | |
| html_export_print_order_by | todo | todo | The printable report should support ordering idea cards by a chosen field or attribute ( for example name, status, file path, or a custom attribute key ). Consumers should be able to select or configure the sort so printed documents follow a deliberate sequence rather than only the default export order. Not implemented yet — tracked for a follow-up. Cross-refs [html_export_printable_pages], [html_export_print_idea_cards]. | 0.120 | |
| html_export_table_columns | done | done | In interactive runtime mode, tables inside layout cards let the user hide and resize columns. An eye control on each column header toggles that column. The eye shows when the header cell is hovered or focused. A column-selector control in the table card header opens a multiselect panel with search, select all, clear, and shift-click range selection. Column resize must not lose control to text selection: pointer capture, select lock, and a resize handle above selectable header text. Hidden columns and column widths persist in localStorage for that page path and table. Document and print runtime modes keep all columns visible at default widths with no column chrome. Cross-refs [html_export_layout_cards], [html_export_list_views]. | 0.095 | |
| html_export_url_base | done | done | Html export should accept an optional urlBase mount prefix ( for example `/spec` or `/reqlan/spec` ). When set, page and asset hrefs must be root-relative under that prefix so static hosts resolve correctly whether the directory URL includes a trailing slash. When omitted, exports keep document-relative hrefs suitable for local folder browsing. Applies across [html_export_internal_links] and [html_export_graph_links]; site embed consumption: ["../site/site.rq".spec_html_export]. Often paired with [html_export_header_link] when the export is mounted under a parent static site. | 0.259 | |
| analytical_rust_port | in-progress | in-progress | Port ["../core-architecture.rq".analytical_package] to Rust: lexer, parser ( index path ), extractor, SQLite index, search, and export. `@reqlan/analytical` remains the published JS facade ( `createAnalysisRuntime` / `AnalysisApi` / Zustand host state ). Native work lives in a `crates/` Cargo workspace and ships as per-architecture binaries per [distribution.rust_binary_distribution]. The editor language server is [lsp_rust.language_server_rust]. This analytical port does not replace that process. Component plans: [rust_crate_layout], [lexer_rust], [parser_rust], [parser_align], [indexer.indexer_rust], [search_rust], [export_rust], [native_bridge], [cutover]. | 0.243 | |
| comment_span_align | done | done | On every frozen [golden_corpus] `.rq` file, `//` and `/* … */` comment character ranges must be identical across: - the Langium lexer ( hidden `SL_COMMENT` / `ML_COMMENT` ) - `findCommentSpansInText` C-style spans ( host `#` / triple-quote spans are not `.rq` comments ) - TextMate `comment.line.reqlan` and `comment.block.reqlan` scopes Same-line fences, glob `/**/`, URL `://`, and fenced ` ` ` bodies must not open comments. Mid-line triple backticks are prose. They must not open a fence or hide later ideas. Example: `@reqlan/*` in inline code must not open a block comment. Related: [parser_align], ["../language/syntax.rq".code_snippets], ["../language/syntax.rq".comments], ["../language/syntax-edge-cases.rq".fencing_comments], ["../language/syntax.rq".inline_code]. | 0.141 | |
| cutover | done | done | Native engine is required for CLI, MCP, and the extension IndexService path ( no `REQLAN_ANALYTICAL_ENGINE=js` escape hatch ). Order: golden parse / extract → headless CLI `analyse` / `search` / `export` → MCP → extension via platform VSIX ( host-matching napi `.node` required ). Shared [bases.base] `.reqlan` DB is schema v4 rusqlite; fixtures verify reopen compatibility. The default editor LSP is [lsp_rust.language_server_rust]. Langium remains only as the transitional `node` engine fallback and dual-parser oracle ( [parser_align] ). | 0.118 | |
| export_rust | done | done | Port [export_reqs.export_pipeline]: `buildExportSnapshot` then HTML / markdown / JSON / CSV writers in `reqlan-export`. Interactive HTML JS ( graph canvas, list filters, search bars ) stays static assets bundled into the export — not rewritten in Rust ( [html_export.html_export], [html_export.html_export_graph_animation] ). Hosts still call `NativeAnalysisApi.exportHtml` / `exportMarkdown` / `exportJson` / `exportCsv`; they do not grow format writers. PDF remains print-HTML ( [export_reqs.pdf_export] ); [html_export.html_export_print_order_by] stays a product follow-up, not a port blocker. | 0.137 | |
| golden_corpus | done | done | Parser / extractor alignment uses a frozen `.rq` tree under `testdata/golden-corpus/`, not live product requirements in `reqlan rq/`. Production idea lists may change without regenerating the dump. Copy a representative file into the corpus and refresh `crates/reqlan-index/tests/golden/langium-corpus-names.json` only when grammar or extract behaviour needs a new sample. Langium and Rust both extract name + kind from that tree and must match the committed dump. Live parse alignment ( elements and refs, both engines in one test ): [parser_align]. Comment character ranges on the same tree: [comment_span_align]. | 0.104 | |
| lexer_rust | done | done | Hand-written Rust lexer that ports ["../../packages/language/src/reqlan-token-builder.ts"]: structural vs prose `{` / `}`, fence opacity, quote-aware comments, `@` only at line start, glob `/**/` not a block comment. Linear in document size with sparse brace-depth lookup ( [syntax_edges.context_sensitive_lexer_scaling] ). Owned by `reqlan-parse`; consumed by [parser_rust] and by [lsp_rust.language_server_rust]. The Langium token builder remains only for the transitional `node` engine fallback and dual-parser tests. | 0.105 | |
| native_bridge | in-progress | in-progress | The Rust engine is the analytical core ( parse, index, search, export, typed queries, git-dates, overview coverage, barrel plan ) — not an optional addon. `@reqlan/analytical` is a typed napi facade ( specta `generated.ts` ) plus thin Node helpers ( `createBase`, path utils, `SqliteIndexStore` wrappers over typed napi ). It does not depend on Langium, `@reqlan/language`, or Zustand. Webview-only helpers ( search highlight, filter labels, file-treatment, context signals / model, graph physics ) live in the extension package. Interactive HTML JS stays in `crates/reqlan-export/assets`. Native owns: parse-for-index, persist, query, search, export execution, IndexService / WorkspaceIndex sync, and the index lifecycle FSM + status snapshot via `NativeWorkspaceIndex` + `NativeSqlDb`. The FSM ( `opening` / `syncing` / `ready` / `error` ), counts, in-memory sync progress / cancel, last error, file-index issues ( cap 500 ), and the document-update / workspace-change rings ( cap 50 ) live in `WorkspaceIndexRuntime`; TS reads them through `statusSnapshot()`. Durable diagnostics persist in `index-diagnostics.sqlite`. Consumers: [cli.cli_package] and MCP import `@reqlan/analytical/core` so they never load Langium; the extension host uses the per-target VSIX ( [distribution.vsix_export], [distribution.extension_host_target] ). The host-matching `.node` is required at runtime; there is no sql. js fallback. | 0.290 | |
| parser_align | done | done | Langium and `reqlan-parse` stay two engines as a dual-parser **test oracle**. Grammar truth is [syntax] / ["../../packages/language/src/reqlan.langium"]. The same source must yield the same parse snapshot: top-level elements, live bracket and wiki refs, inline-code count, and fence count. One Node test runs Langium and native `parseReqlanSource` ( `parseAlignSnapshot` ) together. A committed name dump is not the only check. The default editor LSP already uses Rust ( [lsp_rust.language_server_rust] ); this idea does not gate that cutover. Related: [golden_corpus], [parser_rust], [comment_span_align]. | 0.150 | |
| parser_rust | done | done | Recursive-descent parser matching ["../../packages/language/src/reqlan.langium"] productions into a Rust AST the extractor can walk. Recovery: nameless `{...}` and local import mistakes still yield a usable document ( [parser_lexer.recovery_vs_budget] ). Wall-clock [parser_lexer.parse_budget_timeout] via native timeout / cancel — no Node worker process required on the index path. Open `.rq` buffers and CLI `parse` use `reqlan-parse` ( [lsp_rust.language_server_rust] ). Langium remains a dual-parser test oracle and the transitional `node` engine fallback — not the default editor LSP. Alignment contract: [parser_align]. | 0.217 | |
| rust_crate_layout | done | done | Cargo workspace at repo-root `crates/` ( not inside `packages/analytical` ), crates: - `reqlan-parse` — lexer, parser, AST, recovery, parse budget - `reqlan-index` — extract, rusqlite store, mtime sync, diagnostics - `reqlan-search` — fuzzy / contextual / semantic ranking - `reqlan-export` — snapshot + html / md / json / csv writers; embed existing HTML JS assets - `reqlan-analytical` — facade matching `AnalysisApi` + `WorkspaceIndex` operations - `reqlan-napi` — Node binding for the core engine ( [native_bridge] ) - `reqlan-lsp` — editor language server ( [lsp_rust.language_server_rust] ) Mature crates only ( rusqlite, napi ): no experimental parser generators for the lexer. Golden tests compare Rust extract against Langium on the frozen [golden_corpus], plus the language-test fixtures ported into `reqlan-parse` / `reqlan-index` tests. Format: ["../development/commit.rq".rust_fmt]. Root `pnpm test` runs `test:js` then `test:rust` from ["../../package.json"]. `test:rust` is `cargo test --workspace --exclude reqlan-napi`. GitHub CI compiles crates once, then runs `cargo test` as its own step so crate tests reuse that build ( ["../distribution/distribution.rq".ci_gate] ). | 0.138 | |
| search_rust | done | done | Port [search_reqs.fuzzy_search] ( separator-insensitive ranking ), [search_reqs.contextual_search] ( hop-distance rerank ), and [search_reqs.semantic_search] into `reqlan-search`. Same `AnalysisApi.searchRequirements` / `--context` contract ( [cli.commands] ). Activity-bar idea search uses `WorkspaceIndexRuntime.fuzzy_search` / napi `NativeWorkspaceIndex.fuzzySearch` so ranking stays next to rusqlite and only capped hits cross into JS. Mixed list includes indexed `.rq` file-name hits ( `kind` `file` ) and pages with `offset` / `limit` ( `truncated` ). Client-side highlight helpers used by webviews may stay in the TS `./fuzzy-search` export until a shared WASM / JS copy is needed. | 0.137 | |
| ts_interface | done | done | The TS interface over the Rust functionality stays fully typed via specta ( `crates/reqlan-analytical` `export-ts` → `packages/analytical/src/native/generated.ts` ). Specta copies Rust doc comments. `export-ts` rewrites those `rq:` paths from crate-src depth to `packages/analytical/src/native` depth. | 0.040 | |
| contextual_search | done | done | Search string plus an optional context list of relative `.rq` paths, `path#idea` refs, or bare idea names. Runs [semantic_search], then re-ranks with min undirected hop distance from any resolved context idea and exponential decay ( [idea_graph_distance] ): `finalScore = textScore * 2^(-minHop)` ( unreachable floor = 8 hops ). Surfaces: [native_analysis_api] `searchRequirements(..., { context })` and CLI `reqlan search --context` ( [search_cli] ). Implements [search_with_context]. PageRank ( [pagerank] ) and full [context_scope_v2] ingestion remain future work. Default ranking does not use Katz. Opt in with [katz_search]. Complements [fuzzy_search] when neighbourhood / context should bias ranking. ["../cli/click.rq".click_input] ranks ambiguous click matches the same way, with session ideas as context. | 0.208 | |
| file_search | done | done | Rank indexed `.rq` files and comment-bearing code files by basename, stem, and path using the same separator-insensitive matching as [fuzzy_search] / [fuzzy_search_whitespace]. Hits use kind `file`; id is the indexed file URI; name is the basename; line start is 0. Empty query does not dump the file catalog. Mixed into [fuzzy_search] so one list ranks ideas and files together. Code files are those persisted by ["../indexer/indexer.rq".index_code_files]. Surfaces: native `fuzzy_search` via [native_workspace_index], ["../extension/local-context/activitybar-panels/search.rq".search_pane_file_hits]. ["../cli/click.rq".click_input] uses this mix when click has no exact match. | 0.244 | |
| fuzzy_search | done | done | Quick ranked idea search by name, summary, and tags. Separator and word-order behaviour: [fuzzy_search_whitespace]. Also supports prefix, substring, and character-subsequence ( fuzzy ) ranking. Skips ideaset container rows; empty query returns all non-ideaset ideas ( callers may require a non-empty query ). Ranking engine: ["./rust_port.rq".search_rust] `reqlan-search` ( `fuzzy_search` / `search_ideas` ). Also returns [file_search] hits mixed into the same ranked list ( kind `file` ). Pagination: [fuzzy_search_pages]. Activity-bar search calls [native_workspace_index] `NativeWorkspaceIndex.fuzzySearch` so the host never copies the idea catalog into JS. Client-side highlight helpers stay in [fuzzy_search_impl]. | 0.268 | |
| fuzzy_search_pages | done | done | [fuzzy_search] accepts `offset` + `limit` over the already-ranked mixed idea + file list. `truncated` is true when more hits exist after the returned page ( callers show `{n}+` rather than a full count ). Used by ["../extension/local-context/activitybar-panels/search.rq".search_pane_load_more]. | 0.097 | |
| fuzzy_search_whitespace | done | done | Matching treats `_`, `-`, `.` / `...` / `…`, and whitespace as interchangeable separators so queries like `cli package`, `cli_package`, and `cli-package` hit the same idea. Prefers query word order, but still matches when words are reordered or when intermediate name words are missing ( e. g. `package cli` → `cli_package`, `search actions` → `search_code_actions` ). Extends [fuzzy_search]; implementation [fuzzy_search_impl]. | 0.117 | |
| katz_search | done | done | Search can bias matches by inbound Katz centrality ( [katz_centrality] ). The bias is off unless the caller sets it. CLI flag: `reqlan search <query> --katz`. API field: `SearchRequirementsOptions.katz`. When the bias is on, the score is `textScore * (1 + katzScore)`. The match reasons include `katz:<score>`. Context hop decay runs first. The Katz multiplier runs second. When the bias is off, match order and scores stay on the text ranking. | 0.082 | |
| extension_host_native_stage | done | done | Local F5 and `pnpm run build` / `build:ext` stage the host-matching core engine `.node` into `packages/extension/native/` so activation loads the same layout as a per-target VSIX ( [distribution.vsix_export], ["../distribution/native_host_binary.rq".native_host_binary_development] ). The same build copies the `.node` into `packages/analytical-native/<napiSuffix>/` so the workspace optionalDependency matches a production npm install ( [distribution.rust_binary_distribution] ). The orchestrator builds `cargo build -p reqlan-napi -p reqlan-lsp --release` when needed, then copies `reqlan_napi.node` beside the extension. It is not a separate manual install. That command writes `reqlan-lsp` into `crates/target` ( ["../extension/language/support/language-server-rust.rq".language_server_rust] ). Azure compiles the [distribution.native_targets] matrix, then stages those binaries with `prepare-native-packages.mjs --binary-dir` before tests and VSIX packaging. Versions come from git at the `ext/v*` tag. Do not fetch platform packages from npm on Azure ( [distribution.extension_distribution] ). Implementation: ["../../scripts/stage-host-native.mjs"], ["../../scripts/ensure-host-native.mjs"], ["../../packages/extension/scripts/build.mjs"], ["../../azure-pipelines.yml"]. | 0.140 | |
| incremental_extension_build | done | done | Extension builds fingerprint inputs and outputs per independently cacheable artifact: - Langium generated sources / TextMate grammar - logo media - generated marketplace README - each of the five Vite webviews separately - extension-host / language-server esbuild outputs Host napi staging ( [extension_host_native_stage] ) always runs: cargo is incremental locally; Azure stages matrix artifacts and skips a second cargo compile in the package job. A step is skipped only when its command / input fingerprint and current output-content fingerprint both match the cache under `packages/extension/out/`. Missing, changed, or manually edited output invalidates that step. Fingerprint walks skip `node_modules` so Vite scratch files under `.vite-temp` and installed deps do not race or inflate hashes; dependency identity stays in `pnpm-lock.yaml` via shared inputs. Webview output fingerprints cover each complete emitted directory, including future lazy chunks. Host-bundle inputs include bundled phonebook data as well as workspace package output. TypeScript retains its native `tsc -b` project-reference incrementality. TextMate syntax copies are cached by content; Cursor-skill sync still runs because it targets workspace-external state. `pnpm run build:force` / `node scripts/build.mjs --force` bypasses all cache entries. `vscode:prepublish` uses forced mode with `--skip-native` so `vsce package --target` cannot replace a staged per-target `native/` with the packaging agent's host binary ( [distribution.vsix_export] ). `clean` removes `out/`, including the cache, so clean builds remain cold. Root `pnpm run build:ext` delegates only to the extension package. The extension build generates Langium artifacts when their fingerprints change, then its `tsc -b` builds language / analytical project references before downstream fingerprints are evaluated. It does not rebuild unrelated CLI, MCP, site, or test projects. | 0.045 | |
| typescript_compile | done | done | Langium writes `packages/language/src/generated` ( ast, grammar, module ). That directory is gitignored. `tsc -b tsconfig.build.json` fails on a clean checkout if those files are missing. Root script `build:tsc` is `langium:generate` then that `tsc -b`. Full `pnpm run build` starts with `build:tsc`. GitHub CI, Azure package build, and npm deploy use `build:tsc` so compile matches a clean local build ( [distribution.prerelease_tests], [distribution.ci_gate], [distribution.npm_distribution] ). GitHub CI compiles the TypeScript workspace once in the sequential job, then reuses that emit for [check.check] and package tests ( [check.check_meta_implementation] ). `langium generate` validates ["../../packages/language/langium-config.json"] against the langium-cli schema. langium-cli 4. 3. 0 does not set schema `$id`. jsonschema 1. 5. 0 then throws `ERR_INVALID_URL` for `#/$defs/languageItem` on Node 24. The workspace pins jsonschema 1. 4. 1 for langium-cli in ["../../pnpm-workspace.yaml"] until a langium-cli release includes that `$id`. | 0.045 | |
| typescript_emit | done | done | Package TypeScript emit goes to each package `out/` via `tsc -b` project references ( ["../../tsconfig.build.json"] ). Shared compiler options live in ["../../tsconfig.base.json"]. ["../../tsconfig.json"] is a solution file ( `files: []` ) so a plain `tsc` / VS Code " tsc: build " does not write `.js`, `.d.ts`, or `.js.map` next to sources. Webview tsconfigs are `noEmit` — Vite bundles to `packages/extension/media/webviews/`. A clean compile must generate Langium sources first ( [typescript_compile] ). | 0.040 | |
| rust_fmt | done | done | CI runs `cargo fmt --all` in write mode on all crates ( [rust_port.rust_crate_layout], [rustfmt_toml] ). Format is a chore, not a gate: an unformatted tree does not fail the job. On same-repo pull requests, if fmt changes files, CI commits `chore: cargo fmt` and pushes to the PR branch. Fork PRs and `workflow_call` ( release ) only format the working tree; they do not push. Recursion stops because: ( 1 ) pushes with the default `GITHUB_TOKEN` do not start new workflow runs; ( 2 ) a second run finds no diff after fmt, so it does not commit again. Pull requests and the release workflow share the same sequential CI job ( ["../distribution/distribution.rq".ci_gate] ). Git pre-commit runs the same formatter in write mode so local commits stay formatted. Skip when no crate Rust is staged, format in place and restage, skip merge / rebase, `SKIP_RUST_FMT=1` escape hatch. [workspace_package] `prepare` sets `core.hooksPath` to `.githooks`. | 0.081 | |
| phonebook_icon_keys | done | done | per [phonebook_reqs.phonebook_icons] A phonebook link or package can omit `icon`. If `icon` is present, it is an object with `set` and `name`. Set ids and lookup URLs are canonical in ["../phonebook.json"]. icons. Use each set's `search` URL to find replacement slugs when a name changes or an icon is removed upstream. Typed helpers live in [phonebook_script]: `formatPhonebookIconKey`, `getPhonebookIconSearchUrl`. | 0.048 | |
| pnpm_val | done | done | Root `pnpm run val` chains `check`, then `test`, then `build` ( ["../../package.json"] ). Use for local / ingestion validation ( ["../ingest/ingest.rq"._file] ). | 0.046 | |
| ci_gate | done | done | Pull-request CI and the main release pipeline share one sequential workflow: ["../../.github/workflows/ci.yml"]. ["./distribution.rq".deployment_flow] must not apply changesets or create tags until that workflow succeeds. Order in one job: install workspace packages, compile crates and TypeScript once, then crate tests, ["../core_analysis/check.rq".check], then JS / TS tests ( ["./distribution.rq".prerelease_tests], ["../core_analysis/check.rq".check_meta_implementation], ["../development/commit.rq".rust_fmt] ). `@reqlan/*` packages come from this git checkout ( `workspace:*` ), not from the npm registry. Platform native packages are linked the same way in this workspace through `packageExtensions` in ["../../pnpm-workspace.yaml"]; they are not listed in git `@reqlan/analytical` `optionalDependencies`. A push to main does not start CI a second time as an independent workflow; ["../../.github/workflows/release.yml"] calls ["../../.github/workflows/ci.yml"] with `workflow_call`. Because ["../../.github/workflows/ci.yml"] declares `permissions.contents: write`, the caller job in ["../../.github/workflows/release.yml"] must grant `contents: write` when invoking it. The version-bump commit ( `chore(release):` ) skips the gate and the version job to prevent a recursive run. | 0.172 | |
| extension_distribution | in-progress | in-progress | Azure DevOps builds, tests, packages, and publishes the VS Code extension only for tags matching `ext/v*`. Tags are created by ["./distribution.rq".deployment_flow]. The extension version comes from packages / extension / package. json at the tag; the pipeline must not overwrite it from the tag suffix. The implementation lives in ["../../azure-pipelines.yml"]. The pipeline builds language, packages, and the extension only; it does not run the site static build or install Playwright ( [site_distribution] owns site CI ). Native core-engine binaries for VSIX come from the Azure `build_native` matrix on this git tag ( [vsix_native_not_from_npm], [rust_binary_distribution], [vsix_export] ). GitHub Actions still builds and publishes platform packages on `analytical/v*` for CLI / MCP ( [npm_distribution] ). Before tests Azure stages matrix artifacts into workspace packages ( `prepare-native-packages.mjs --binary-dir` ) and checks the host optionalDependency ( `scripts/ensure-host-native.mjs --no-build` ). VSIX packaging follows ["./distribution.rq".vsix_export]: one VSIX per [native_targets] tuple; Marketplace selects the VSIX matching the extension-host target ( [extension_host_target] ). Marketplace publish is complemented by Open VSX per ["./distribution.rq".openvsx_registry]. Pipeline cache and publish parallelism: [extension_pipeline_cache], [extension_pipeline_publish]. | 0.243 | |
| extension_host_target | in-progress | in-progress | The extension host must load the native core engine for its machine. The selection strategy is ["./native_host_binary.rq".native_host_binary]. IndexService and WorkspaceIndex use `NativeWorkspaceIndex` ( ["../core_analysis/rust_port.rq".native_bridge] ). If the engine is missing, activation fails. Loader: ["../../packages/analytical/src/native/load-native.ts"]. The extension registers search folders from ["../../packages/extension/src/analytical_submodule/index.ts"]. | 0.268 | |
| extension_package | done | done | The VS Code extension lives in packages / extension with npm package name `reqlan-extension` ( Marketplace / Open VSX id `reqlan.reqlan-extension` ). Its package. json version is the Marketplace and Open VSX release version. That version is independent of `@reqlan/analytical` ( [version_management] ). | 0.088 | |
| extension_pipeline_cache | done | done | Azure `build_native` mirrors GitHub `Swatinem/rust-cache` via `Cache@2`: cache rustup + cargo home under `$(Pipeline.Workspace)/rust-cache` ( `CARGO_HOME` / `RUSTUP_HOME` ) and `crates/target`, keyed by OS, `rustTarget`, and `crates/Cargo.lock`. Install rustup only when `$CARGO_HOME/bin/rustc` is missing after restore; otherwise reuse the cached toolchain ( `--no-modify-path` ). Azure `build_test_package` caches the pnpm store keyed by OS and `pnpm-lock.yaml` ( same role as `cache: pnpm` on GitHub ). Do not cache VSIX or napi pipeline artifacts across tags; git at `ext/v*` stays canonical ( [vsix_native_not_from_npm] ). Implementation: ["../../azure-pipelines.yml"]. Related: [extension_distribution], ["./distribution.rq".ci_gate] ( GitHub rust-cache + pnpm ). | 0.072 | |
| extension_pipeline_publish | done | done | Marketplace and Open VSX publish every per-target VSIX in parallel ( `ForEach-Object -ThrottleLimit 6 -Parallel` ), not one after another. Open VSX uses the workspace `ovsx` package ( `pnpm --filter./packages/extension exec ovsx` ); do not `pnpm dlx ovsx` per file. Local `--openvsx` in ["../../scripts/package-extension-targets.mjs"] uses the same `pnpm exec ovsx` path. Implementation: ["../../azure-pipelines.yml"], ["../../packages/extension/package.json"], ["../../scripts/package-extension-targets.mjs"]. Related: [extension_distribution], [openvsx_registry], [vsix_export]. | 0.094 | |
| grammar_linguist | pending | pending | The grammar should be updated to use https://github.com/github-linguist/linguist Acceptance is dependent on maturity and usage. ( this is what gh uses for language rendering etc ) | 0.040 | |
| native_targets | done | done | Canonical VS Code `targetPlatform` / npm `os` + `cpu` tuples for native distribution: - `win32-x64` - `win32-arm64` - `linux-x64` - `linux-arm64` - `darwin-x64` - `darwin-arm64` napi-rs package suffixes map as: `linux-*-gnu`, `darwin-*`, `win32-*-msvc` ( gnu-only for linux; musl / alpine out of scope for the first matrix ). Azure `build_native` uses the same tuples and `ciImage` from ["../../scripts/native-targets.mjs"]. Used by [rust_binary_distribution], [vsix_export], [extension_host_target], and ["./native_host_binary.rq".native_host_binary]. Implementation: ["../../scripts/native-targets.mjs"]. | 0.317 | |
| npm_distribution | in-progress | in-progress | npm publishing runs only when a `language/v*`, `analytical/v*`, `cli/v*`, or `mcp/v*` tag is pushed. Tags are created by ["./distribution.rq".deployment_flow]; the npm workflow does not publish on ordinary commits. The workflow builds platform binaries when `@reqlan/analytical` was bumped ( `analytical/v*` tag, or `analytical_tag` from the release workflow ). It does not use `github.event_name == 'workflow_call'` for that gate: in a reusable workflow that name is the caller event ( `push` ), and `npm_ref` is the first language / analytical / cli / mcp tag, so a language bump would skip natives. It then publishes each package in sequence with `pnpm --filter <pkg> publish --access public --no-git-checks`. The sequence is language, then all platform packages for [native_targets], then `@reqlan/analytical`, then cli, then mcp. The workflow does not publish `@reqlan/analytical` until every platform package of that `@reqlan/analytical` version is on the npm registry or was published in the same job ( [rust_binary_distribution] ). ["../../scripts/require-natives-published.mjs"] enforces that gate from the analytical package. json version ( not Changesets ). Before each package publishes, ["../../scripts/skip-already-published.mjs"] with `--filter <pkg>` checks the public npm registry and skips that step when the version is already published ( sets `should_publish=false` ). That avoids re-attempting already-published versions under OIDC. Auth uses npm trusted publishing ( OIDC ) from GitHub Actions — no long-lived `NPM_TOKEN`. Workflow requirements: `permissions.id-token: write`, Node ≥ 22. 14 with npm ≥ 11. 5. 1, registry `https://registry.npmjs.org`. Each of `@reqlan/language`, `@reqlan/analytical`, `@reqlan/cli`, `@reqlan/mcp`, and every `@reqlan/analytical-<platform>-<arch>` platform package must have a Trusted Publisher on npmjs. com for GitHub Actions with org / user `littletuna4`, repository `reqlan`, workflow filename `deploy-npm.yml`, allowed action `npm publish`. Package `repository.url` must match `https://github.com/littletuna4/reqlan.git`. Provenance attestations are generated automatically for public packages from this public repo. Bootstrap: first publish of each scoped package can be done locally ( you must be logged in to npm as an `@reqlan` org member ) with `pnpm --filter @reqlan/language publish --access public --no-git-checks`, then platform packages, then analytical, then cli, then mcp — or wait for the trusted-publisher CI path after packages exist and Trusted Publisher is configured. Note: Packages must be `private: true` to skip local `changeset publish` ( `pnpm run release:npm` ), as with `reqlan-extension`. Platform packages are in the Changesets `ignore` list ( `@reqlan/analytical-*` ). Git `@reqlan/analytical` does not list them as `optionalDependencies`. ["../../scripts/prepare-native-packages.mjs"] `--publish-versions` writes pinned optionalDependencies for the npm checkout only. CI publishes platform packages per-package after binaries are built and does not use `changeset publish` for those tarballs. Published packages declare `publishConfig.access: public`. Trusted Publisher is configured on npmjs. com for each `@reqlan/analytical-<platform>-<arch>` package | 0.266 | |
| npm_package_readme | done | done | Publishable npm packages ( `@reqlan/language`, `@reqlan/analytical`, `@reqlan/cli`, `@reqlan/mcp` ) ship README. md generated from each package's `README.template.md`. Pattern matches [extension_readme]: centered logo, package name, description, features, phonebook links, and Changesets changelog body. Generated by ["../../scripts/generate-npm-package-readmes.ts"] via root `pnpm run build:readme`. Contact, Discord, sponsorship, and repository links come from [phonebook_reqs.phonebook] via ["../../scripts/phonebook.ts"] per [phonebook_reqs.npm_package_readme_links]. CLI native-engine install note: ["./native_host_binary.rq".published_host_native_install]. | 0.046 | |
| openvsx_registry | in-progress | in-progress | The extension should be published to the Open VSX registry at https://open-vsx.org so it is installable in editors that use Open VSX instead of the Visual Studio Marketplace. Publishing should use the same tag-driven `ext/v*` releases as ["./distribution.rq".extension_distribution]. Each per-target VSIX from ["./distribution.rq".vsix_export] is published to Open VSX ( Marketplace and Open VSX both select by extension-host `targetPlatform` ). Azure DevOps is the authoritative publish workflow; Open VSX publish belongs in ["../../azure-pipelines.yml"] alongside marketplace deploy. Publish path follows [extension_pipeline_publish]. | 0.166 | |
| prerelease_tests | done | done | JS / TS package tests ( `pnpm run test:js` ) run after the shared TypeScript compile ( `pnpm run build:tsc` ) from this git checkout. GitHub CI runs them in the same job as crate tests and ["../core_analysis/check.rq".check], after that compile ( ["./distribution.rq".ci_gate] ). This job runs on every pull request and on main before versioning. Host native addon is required because analytical and language tests load `reqlan-napi`. The host release build compiles `reqlan-napi` only. Crate tests compile `reqlan-lsp`. That binary ships in the VSIX, not in this host package. Azure runs `pnpm run test:js` before VSIX publish ( [extension_distribution] ). Local `pnpm run test` runs `test:js` then `test:rust` ( ["../core_analysis/rust_port.rq".rust_crate_layout] ). The implementation lives in ["../../.github/workflows/ci.yml"]; ["../../.github/workflows/release.yml"] calls it with `workflow_call` as part of ["./distribution.rq".ci_gate]. Related: ["../core_analysis/check.rq".check_meta_implementation] | 0.069 | |
| rust_binary_distribution | in-progress | in-progress | One platform-specific npm package per [native_targets] tuple: `@reqlan/analytical-<platform>-<arch>` ( napi suffixes: `linux-x64-gnu`, `linux-arm64-gnu`, `darwin-x64`, `darwin-arm64`, `win32-x64-msvc`, `win32-arm64-msvc` ). Each package declares `os` and `cpu`, contains the matching core engine `.node`, and uses the `@reqlan/analytical` package. json version as the single source of truth. ["../../scripts/prepare-native-packages.mjs"] writes that version. Changesets ignores `@reqlan/analytical-*` in ["../../.changeset/config.json"]. Git `@reqlan/analytical` does not list platform packages as `optionalDependencies`; this workspace links them through `packageExtensions` in ["../../pnpm-workspace.yaml"]. The published tarball lists pinned optionalDependencies ( `--publish-versions` ) ( [version_management] ). CLI and MCP install the host-matching binary from those published optionalDependencies. The published tarball runs ["./native_host_binary.rq".published_host_native_install] so a host miss prints a warning. Published on the existing `analytical/v*` channel ( [npm_distribution] / [release_tagging] ); not a separate tag channel. The napi build runs when this channel moved: a direct `analytical/v*` tag, or `analytical_tag` from [deployment_flow]. The first npm tag ( `npm_ref` ) can be `language/v*` in the same release, so that ref is not the native-build gate. Website download listing and GitHub-release postinstall are later follow-ups ( out of scope here ). Engine port: ["../core_analysis/rust_port.rq".analytical_rust_port] / ["../core_analysis/rust_port.rq".native_bridge]. Per-target VSIX packaging uses the same package layout under ["../../packages/analytical-native/"], filled by Azure from this git commit ( [vsix_export], [vsix_native_not_from_npm] ). npm tarballs of those packages are for CLI / MCP only. `collect-napi-binary.mjs` copies `reqlan-lsp` only with `--with-lsp`. The npm workflow does not pass that flag. Azure does, because the VSIX ships `bin/reqlan-lsp` ( [vsix_export] ). Implementation: ["../../packages/analytical-native/"], ["../../scripts/prepare-native-packages.mjs"], ["../../scripts/require-natives-published.mjs"], ["../../packages/analytical/scripts/check-host-native.mjs"], ["../../.changeset/config.json"], ["../../.github/workflows/deploy-npm.yml"] build-native job, ["../../azure-pipelines.yml"] build_native job. | 0.380 | |
| version_management | done | done | Package versions are managed with Changesets via ["../../.changeset/config.json"]. Publishable channels ( each tag-driven ): - Marketplace / Open VSX: ["./distribution.rq".extension_package] ( private npm package; tagged `ext/v*` ) - GitHub Pages: ["./distribution.rq".site_package] ( tagged `site/v*` ) - npm: `@reqlan/language`, `@reqlan/analytical`, `@reqlan/cli`, and `@reqlan/mcp` ( tagged `language/v*`, `analytical/v*`, `cli/v*`, `mcp/v*` ) `reqlan-extension` has its own semver. It is not version-locked to `@reqlan/analytical`. The extension can update independently of the analytical surfaces it uses ( [extension_package] ). The version of `@reqlan/analytical` is the single source of truth for every `@reqlan/analytical-<platform>-<arch>` package ( [rust_binary_distribution] ). Platform packages are in the Changesets `ignore` list ( `@reqlan/analytical-*` ). Git `@reqlan/analytical` does not list them as `optionalDependencies`, so `changeset version` can run. This workspace still links the checkout packages during `pnpm install` via `packageExtensions` in ["../../pnpm-workspace.yaml"] ( `workspace:*`, not npm ). After `changeset version`, ["../../scripts/prepare-native-packages.mjs"] copies the analytical version onto each platform package. The published tarball lists pinned `optionalDependencies` ( `--publish-versions` ) so CLI and MCP customers get the host binary. Platform native packages are not a separate tag channel; they ride `analytical/v*` ( [rust_binary_distribution] ). VSIX packaging uses the package. json versions in that git commit ( [vsix_native_not_from_npm] ). `@reqlan/language`, `@reqlan/cli`, `@reqlan/mcp`, and the site keep their own versions. Contributors record release intent by adding a changeset file when a publishable package changes. | 0.233 | |
| vsix_export | in-progress | in-progress | One VSIX is built per [native_targets] tuple via `vsce package --target <tuple> --no-dependencies`. Each VSIX keeps today's esbuild-bundled JS ( TS dependency graph stays bundled ); only the matching platform native package is staged into the VSIX so other targets' binaries are excluded. The same VSIX packs `bin/reqlan-lsp` for that target ( ["../extension/language/support/language-server-rust.rq".language_server_rust] ). Azure `collect-napi-binary.mjs --with-lsp` copies that binary into the native artifact. The pack script checks that zip entry by `file_size`. It does not read the binary through a pipe. The napi addon stays for the index host. npm platform packages do not contain it ( [rust_binary_distribution] ). Stamp `targetPlatform` so Marketplace / Open VSX route by extension host ( [extension_host_target], ["./native_host_binary.rq".native_host_binary_distributed] ). Artifact name shape: `reqlan-extension-<version>-<target>.vsix`. The VSIX includes logo assets from the extension media build: PNG for marketplace icon, language file icon, activity bar, and README; WebP as a raster fallback. Implementation: ["../../scripts/package-extension-targets.mjs"], ["../../scripts/prepare-native-packages.mjs"], ["../../scripts/collect-napi-binary.mjs"], ["../../packages/extension/.vscodeignore"] ( `native/`, `bin/` ), ["../../azure-pipelines.yml"]. | 0.446 | |
| vsix_native_not_from_npm | done | done | Azure must not use the npm deploy path as an input when it packs VSIXes. Git at the `ext/v*` tag is canonical. Azure compiles the [native_targets] matrix from that commit ( [vsix_export], [extension_distribution] ). npm stays the install path for CLI and MCP only ( [npm_distribution], [rust_binary_distribution] ). | 0.241 | |
| repo_license | done | done | The repository root must declare the same licence as the publishable packages and crates. SPDX id: `AGPL-3.0-only`. The root must hold a `LICENSE` file with the GNU Affero General Public License version 3 text. The extension package `LICENSE` must use the same text as the root `LICENSE`. Package metadata must set `"license": "AGPL-3.0-only"` for the workspace root and for every publishable package under `packages/`. The Rust workspace must set `license = "AGPL-3.0-only"` in ["../../crates/Cargo.toml"]. | 0.040 | |
| native_host_binary | done | done | Select the native core-engine binary from the extension host. The host is the Node process that runs the extension. Read `process.platform` and `process.arch` from that process. Do not use the platform of the VS Code UI client. Examples: - UI on Windows and host on Linux SSH: use a linux tuple. - UI on Windows and host on WSL: use a linux tuple. - UI and host on macOS: use a darwin tuple. Map the platform and the arch to one tuple in ["./distribution.rq".native_targets]. Load only the binary that matches that tuple. Do not load a binary for a different tuple. If the platform and the arch are not in ["./distribution.rq".native_targets], fail. If the binary is missing, fail. Do not use sql. js. Keep the loader map the same as ["../../scripts/native-targets.mjs"]. Local development: [native_host_binary_development]. Production distribution: [native_host_binary_distributed]. Implementation: ["../../packages/analytical/src/native/load-native.ts"]. | 0.144 | |
| native_host_binary_development | done | done | Use this procedure on a developer machine ( F5 and `pnpm run build` ). Identify the host from `process.platform` and `process.arch` of the build process. The build process and the extension host must be the same machine. Copy the matching `.node` file to `packages/extension/native/reqlan_napi.node`. Write `packages/extension/native/target.json` with the host tuple. If the binary is not present, build it with `cargo build -p reqlan-napi -p reqlan-lsp --release`. At run time, search in this order: 1. The `native/` folder of the extension, if `target.json` records the host tuple. 2. The resolved `main` of the host npm package ( `createRequire` from `@reqlan/analytical/package.json` ). 3. The host file in `packages/analytical-native/<napiSuffix>/` ( workspace package directory, same layout as a published package ). 4. The host npm package name. 5. `crates/target/<rustTarget>/release` and `crates/target/release`. 6. The debug folders with the same names. Do not search packages for other platforms. If `target.json` records a different tuple, do not load `native/reqlan_napi.node`. Windows cargo emits `reqlan_napi.dll`; the loader uses `process.dlopen` for `.dll` and `.node` ( Node `require` only loads `.node` ). Windows UI with a WSL extension host uses the linux tuple ( [native_host_binary] ). Windows UI with a Windows extension host cannot load a `.node` staged from WSL; restage on Windows or open the folder in a WSL remote window. Before tests, the workspace host package must contain the `.node` so loaders behave like production ( [native_host_binary_distributed] ). `scripts/ensure-host-native.mjs` uses an existing matrix binary, or cargo locally. Azure compiles the ["./distribution.rq".native_targets] matrix from git and stages those artifacts; it does not fetch binaries from npm. Detail: [development_build.extension_host_native_stage]. Implementation: ["../../scripts/stage-host-native.mjs"], ["../../scripts/ensure-host-native.mjs"], ["../../packages/extension/src/analytical_submodule/index.ts"]. | 0.078 | |
| native_host_binary_distributed | in-progress | in-progress | Use this procedure for a published VSIX and for published npm packages. VS Code extension: The registry installs the VSIX that has `targetPlatform` equal to the extension host. Each VSIX contains one `.node` file for one ["./distribution.rq".native_targets] tuple. The file name is `native/reqlan_napi.node`. The VSIX also contains `native/target.json`. At run time, read `process.platform` and `process.arch` on the extension host. If `target.json` records a different tuple, do not load `native/reqlan_napi.node`. Do not put binaries for other platforms in that VSIX. The Azure pipeline compiles each platform binary from this git commit ( `build_native` + `collect-napi-binary.mjs` + `prepare-native-packages.mjs --binary-dir` ). Package versions in git are canonical. The pipeline does not wait for npm. Related: ["./distribution.rq".vsix_export], ["./distribution.rq".extension_distribution], ["./distribution.rq".vsix_native_not_from_npm]. CLI and MCP: npm installs the host package from `optionalDependencies` on the published `@reqlan/analytical` tarball. Git does not list those entries. ["../../scripts/prepare-native-packages.mjs"] `--publish-versions` writes them at npm publish. The loader resolves that package from `@reqlan/analytical/package.json` ( `createRequire` ), which is where the optional dependency is attached after `npm install`. Published installs also run [published_host_native_install]. Related: ["./distribution.rq".rust_binary_distribution], [rust_port.native_bridge]. Implementation: ["../../scripts/package-extension-targets.mjs"], ["../../scripts/prepare-native-packages.mjs"], ["../../packages/analytical/src/native/load-native.ts"], ["../../packages/analytical/scripts/check-host-native.mjs"], ["../../azure-pipelines.yml"]. | 0.129 | |
| published_host_native_install | done | done | After a published install of `@reqlan/analytical`, the host platform package must resolve. The host is `process.platform` and `process.arch` of the install process ( [native_host_binary] ). Git `@reqlan/analytical` does not list `optionalDependencies`. The check exits 0 in this workspace. The published tarball lists pinned optionalDependencies ( ["./distribution.rq".rust_binary_distribution] ). That tarball runs ["../../packages/analytical/scripts/check-host-native.mjs"] as `postinstall`. If the host package name is missing from `optionalDependencies`, or `createRequire` from `@reqlan/analytical/package.json` cannot resolve that package to a file, the script prints a warning and exits 0. The warning text tells the user how to enable optional dependencies and how to limit pnpm to the host OS and CPU. Do not use `--force`. `--force` installs every optional platform package. Do not skip the script with `--ignore-scripts` on a published install. CLI install documentation: ["../../packages/cli/README.template.md"]. Implementation: ["../../packages/analytical/scripts/check-host-native.mjs"], ["../../packages/analytical/package.json"], ["../../packages/cli/README.template.md"], ["../../packages/analytical/README.template.md"]. | 0.114 | |
| docs_image_generation | done | done | Images of HTML components for docs, READMEs, and slide packs are generated at build time from production UI. Prefer live HTML imported from production code when embedding is practical; otherwise capture PNGs from the same bundles. Deck asset keys are owned by ["../marketing_and_media/tutorials.rq".presentation_assets]. Pipeline: [docs_webview_build] → [docs_playwright_capture] → [site_reqs.sync_presentations]. | 0.040 | |
| docs_playwright_capture | done | done | Playwright loads each built webview via a harness page with a mocked `acquireVsCodeApi` host, posts demo index / context payloads, and writes PNGs into ["../../presentations/assets/"]. Generated assets replace the former coloured SVG stubs ( activity-bar, montage-ide, chat-search, file-link ). Wait for document load, then the shot ready selector. The Ideas Summary montage shot waits longer because the bundle is large. On a ready-selector timeout, print page errors and `#app` HTML. Runs as a step in ["../../site/scripts/build-static.mts"] before presentation sync. CI installs Chromium via `pnpm exec playwright install --with-deps chromium`. | 0.045 | |
| docs_webview_build | done | done | The site static build builds the extension webviews needed for docs captures before Playwright runs. Builds only the webviews listed by the capture catalog ( activity-bar, ideas-summary, onboarding ) — not a full extension VSIX. Uses the same Vite configs as [build_reqs.webview_build]. | 0.045 | |
| add_to_chat | done | done | A set of ["../host/features-commands.rq".command_palette] functions that add the selected idea, ideaset, or file to the chat. Commands: Add Idea / Ideaset / File to **Current** Chat ( primary ) and to **New** Chat ( secondary ); category Reqlan. Uses the idea / ideaset under the cursor ( or ideas overlapping the editor selection ) when in a. rq file; otherwise QuickPick. Primary: inject `#requirement` / `#file` context into the **current** chat input without submitting. Secondary: open a **new** chat / agent session, then seed the same payload into its input. Discoverability: editor context uses a **Reqlan: Add to Chat** submenu ( current + new for idea / ideaset / file ). Palette titles name Current vs New explicitly. Host APIs ( Cursor vs VS Code ) — see open-chat helper: - Cursor: `workbench.action.chat.open` always creates a **new** tab ( `openInNewTab: true` ); do **not** use it for current-chat. Current = `composer.focusComposer` + paste; New = `composer.createNew({ openInNewTab: true, partialState: { text, richText } })`. - VS Code Copilot: current = `workbench.action.chat.open` + `isPartialQuery: true`; new = `workbench.action.chat.newChat` then open. Clipboard fallback when no chat command is available. Host open helper: ["../../../packages/extension/src/agent_module/ai_commands/open-chat.ts"] ( `target: 'current' | 'new'` ). Activity bar search hits: ["../local-context/activitybar-panels/search.rq".search_add_to_chat] — split button with ▾ menu. | 0.061 | |
| ai_fix_stale_references | done | done | A chat skill `/rq-fix-stale-references` finds and repairs unresolved idea, comment, file, import, and wildcard references. The issue list comes from ["../../core_analysis/check.rq".check] ( `reqlan check --json` ). Repair choices follow ["../language/support/features-imports.rq".import_error] and ["../mutation/refactor_support.rq".comment_reference_refactor_support]. Slash-command name uses [skills_and_mcp.chat_skill_naming]. | 0.048 | |
| mcp_check | done | done | MCP `check` wraps [check] for AI clients ( [mcp_server] ). The tool returns JSON issue rows in [check_order_by_target] order. An empty list means that there are no issues. Optional arguments match the CLI [check] flags: glob, wildcard-zero, wildcard-one, skip-targets, and skip-gitignored-targets. `glob` is one path pattern or a list of path patterns ( [check_glob_patterns] ). A file that matches any pattern is in the scan. Related: [mcp_tools] [check] ["../../core_analysis/check.rq".check_unresolved_imports] | 0.243 | |
| mcp_click_retrieval | done | done | MCP `click` is the canonical search and retrieval tool ( [click], [mcp_server] ). One tool returns compact unique context, ranked ambiguous matches, or search hits. A second click on the same centre in the session returns connected content ( [agent_advisory] ). This surface maximises the token-to-insight ratio: names, short bodies, and caps — not edge dumps or full-graph lists. Keyword search, file context, local graph, and subtree expansion are click paths, not extra MCP tools. Agents must pass `sessionKey` on the next click. Related: [mcp_keyword_search], [mcp_tree_interrogation], [mcp_tree_summarisation]. | 0.144 | |
| mcp_diff | done | done | MCP `diff` wraps [diff] for AI clients ( [mcp_server] ). Arguments are optional `from`, optional `to`, and optional `verbosity` ( [diff_verbosity] ). The tool returns the same JSON array as `reqlan diff --json`. An empty array means that no idea changed. A missing revision is a tool error. Related: [mcp_tools] [diff] | 0.074 | |
| mcp_interaction_discovery | done | done | MCP clients discover tools through the MCP protocol tool list. Do not expose a list_interactions MCP tool. | 0.040 | |
| mcp_keyword_search | done | done | Keyword search across idea names, bodies, and attributes is a [click] path when no centre matches. Do not expose a separate MCP search tool. | 0.059 | |
| mcp_tools | done | done | The MCP server ( `@reqlan/mcp`, [mcp_server] ) exposes a small tool set for AI clients. Search and retrieval use [mcp_click_retrieval] only. Reference checks use [mcp_check]. Idea diffs use [mcp_diff]. Do not expose parse, search, list, file context, local graph, summarise subtree, requirement_reference, file_reference, or list_interactions as MCP tools. Those operations stay on AnalysisApi and CLI when they exist. Headless index storage follows [application_memory] ( `<base>/.reqlan` ), shared with the extension and CLI. | 0.097 | |
| mcp_tools_prompt | done | done | An mcp tool may provide a prompt-oriented entry point for working with the requirement graph. That tool must call [mcp_click_retrieval] for search and retrieval. It must not call the removed tools. | 0.040 | |
| mcp_tree_interrogation | done | done | File and idea neighbourhoods are a [click] unique result ( outbound, backlinks, siblings ). Do not expose a separate MCP file-context or local-graph tool. | 0.059 | |
| mcp_tree_summarisation | done | done | Compact subtree context is a [click] revisit: connected content for a centre already clicked in the session. Do not expose a separate MCP summarise-subtree tool. | 0.059 | |
| agent_surface | done | done | `@reqlan` chat, LM tools, Cursor skills, and MCP install / wrap ( MCP package stays separate ). [chat_participant_doc.chat_participant], [features_ai.ai_integration], [skills_and_mcp]. | 0.046 | |
| export_surface | done | done | One export webview over analytical writers ( HTML, Markdown, JSON, CSV, PDF ). [export_doc.export_webview]. | 0.046 | |
| extension_product | done | done | The VS Code / Cursor extension is the editor product for the reqlan requirement graph. It is not the language definition ( [core_architecture.language_package] ), the index engine ( [core_architecture.analytical_package] ), or the headless MCP / CLI. Host entry: [extension_main]. Product bound: [host_scope.vscode_extension], [host_scope.lsp_support]. Event-driven host mindset: [host_architecture.event_driven]. Domains nest under [extension_domains]. Surfaces for [core_architecture.extension_surfaces] point at these domain hubs — not a flat bag of features-* files. Composition root: [extension_main] registers domain modules after the index adapter activates. | 0.053 | |
| graph_mutation | done | done | Create / split / merge / copy, file-move hooks, and idea refactor. [mutation_hooks], [refactor_support.refactor_changes]. | 0.046 | |
| host_lifecycle | done | done | Install vs base, activation, first paint, onboarding, and configuration schema contribution. [installation.installation], [activation.activation], [startup.extension_startup_reliability], [onboarding.installation_event], [configuration.configuration_location], [command_palette_doc.command_palette]. | 0.046 | |
| index_host | done | done | Thin VS Code adapter over `@reqlan/analytical`: watchers, Uri helpers, multi-base routing, idle stale checks, git-date fill, and index diagnostics webview. Engine ownership stays in the indexer; see [index_host_doc.application_memory], [graph_analysers], [index_diagnostics.index_diagnostics], [git_codelens]. | 0.046 | |
| language_surface | done | done | Edit-time intelligence in the text editor: LSP client glue, syntax, imports, comment `rq:[…]`, CodeLens, inlays, and code actions. [language_syntax.syntax_features], [language_imports], [language_primitives.code_actions], [comment_refs.references_in_functional_code_comments]. | 0.046 | |
| local_context | done | done | Editor-adjacent Context sidebar and the composed context model. Chrome and pane catalog: [activitybar.activity_bar]. Model: [context_scope_doc.context_scope]; evolution: [context_scope_v2.context_scope_v2] ( not a sibling product surface ). | 0.046 | |
| presentation | done | done | Shared webview discipline and the cytoscape graph library. Not a user-facing product feature. [webview_doc.webview], [graph_lib.graph_library]. | 0.046 | |
| workspace_summary | done | done | Ideas Summary — workspace-wide ontology tabs, graph, timeline, and index health. [ideas_summary_doc.ideas_summary]. | 0.046 | |
| first_paint_startup | done | done | The activity-bar HTML shell is assigned by ["../../../packages/extension/src/activity_bar_module/activity-bar-webview-provider.ts"] after one event-loop turn ( retry once on invalid-state ). Host `postMessage` starts only after the webview `ready` handshake; disposed / invalid-state errors are swallowed by ["../../../packages/extension/src/shared/safe-webview-post.ts"]. Its Svelte app installs the host message listener before mount, then sends the one-shot `ready` handshake from a task queued after its first animation frame. ["../../../packages/extension/src/extension/startup-gate.ts"] resolves that handshake once; `IndexService.activate` then discovers bases and syncs. If no view opens, a bounded three-second fallback starts the index without any visible first paint to contend with. Index open / sync requires the host-matching core engine ( rusqlite ); missing `.node` fails activate. Windows claim / dispose races: ["./startup-performance.rq".webview_claim_race]. | 0.148 | |
| configuration_rqignore | done | done | Each base may define path ignore rules in `.reqlan/.rqignore` using **gitignore syntax**. Patterns are relative to the base root ( parent of `.reqlan` ), not the `.reqlan` directory. Owned by `@reqlan/analytical` so the CLI, MCP, and extension share one filter for discovery and indexing walks. Built-in defaults always apply ( dependencies, venvs, build output, VCS / editor stores, `*.db3` / sqlite DBs and SQLite sidecars, binary globs per ["../index-host/index.rq".binary_ignore], files that have no comments per ["../index-host/index.rq".commentless_file_ignore], `*.secret.rq`, hidden entries, and `.reqlan/` itself ); the on-disk file adds or overrides via the same syntax, including `!` negation. New bases seed `.reqlan/.rqignore` with those defaults ( ["../index-host/index.rq".rqignore] ). Missing or unreadable `.rqignore` still uses built-in defaults. Location convention: [configuration_location]. Application memory: ["../index-host/index.rq".application_memory]. | 0.223 | |
| configuration_schema_file | done | done | The machine-readable JSON Schema for `.reqlan/config.json` is ["../../../packages/extension/schemas/config.schema.json"]. That file is the single authoring source. Its `$id` is the public URL `https://reqlan.com/schemas/v1/config.schema.json`. Path segment `v1` is schema compatibility version: additive keys stay on `v1`; breaking shape changes introduce `v2` and a new seed URL. The site hosts a synced copy under `/schemas/v1/` ( ["../../site/site.rq".sync_config_schema] ). The extension contributes the local schema for `**/.reqlan/config.json` via `jsonValidation` in ["../../../packages/extension/package.json"] ( offline editor path ). New bases seed `$schema` to the public `$id` ( ["./installation.rq".base_config_seed] ). `.reqlan/config.json` is a JSON object. Unknown properties are not part of the schema and are ignored when loading. `$schema` is an allowed property. | 0.105 | |
| barrel_page | done | done | When a `.rq` page grows large, barrel it into a container: create one sibling file per top-level idea, import those files into the current page, and replace the page body with a single container idea whose name the user chooses ( default: the sanitized file basename ). Same-file sibling refs are rewritten to qualified imports in the child files. Existing page imports are copied into each child. Top-level ideasets are preserved in the container file. Behaviour lives in ["../../core-architecture.rq".analytical_package] and is exposed via [cli] / ["../../cli/cli_package.rq".commands] ( `reqlan barrel` ). Extension: file-level source code action **Barrel page into container … ** and command palette **Barrel Page** ( `reqlan.barrelPage` ) prompt for the container name, confirm, then create child files and rewrite the active page via WorkspaceEdit. Related: [file_based_code_actions] [decompose] | 0.184 | |
| code_actions_ignore_error | done | done | A Quick Fix [code_actions] on a line with diagnostics inserts [comment_reference_ignore] ( `//rq-ignore-error` ) on the line before the error. The inserted comment uses the same indent as the error line. One action is offered per line when that line has one or more diagnostics. The action is omitted when that line is already suppressed, and when the client asks only for non-quickfix kinds. Available in `.rq` files and in a comment context in a code file ( ["../language/comment-references/functional-code-comment-references.rq".references_in_functional_code_comments] ). In `#` comments the action inserts `# //rq-ignore-error` so the line stays a comment. | 0.074 | |
| command_palette | done | done | Extension commands should be available via the vscode command palette and configurable keyboard shortcuts. Command titles should not repeat the category prefix; use category " reqlan " with plain titles so the palette shows " Reqlan: List All Ideas ", not " Reqlan: Reqlan:... " or " RQ: RQ:... ". Commands belong to product domains; this file is the palette naming hub under host lifecycle. Agent: [agent_ai.add_to_chat]. Index host: [index_diagnostics.open_index_diagnostics]. Diff: [diff_webview]. Export: [export_functions.export_html] and siblings. Mutation: [refactor_support], [barrel_page], [decompose]. Language: [search_code_actions], [code_actions_ignore_error], [create_idea_below_idea_containing_unresolved_reference_under_cursor]. | 0.087 | |
| create_idea_below_idea_containing_unresolved_reference_under_cursor | done | done | If a local idea reference does not resolve, a Quick Fix [code_actions] creates a new idea below the idea that contains the reference. The action uses the unresolved local idea names in the idea that contains the caret: - one unresolved name: apply `Create '<name>' below` directly - several unresolved names: offer `Create all unresolved ideas below` and one action per name Qualified file / import references are omitted. The new idea is an empty block inserted after the containing idea. | 0.074 | |
| search_code_actions | done | done | when cursor is in a `[reference]`, there shouls be a code action to search for references to that idea. it should also allow wrapping a selected string, or the word at the cursor when in idea-body prose, as a new `[reference]`. it should open up a webview modal that shows a search dialog that allows the user to search for potential references. the search menu should preserve / show the context of the would-be reference: the rest of the containing idea, with styling, and with the selected text pulsing / styled prominently. it should support fuzzy / partial searching via [search_analysis.fuzzy_search]. it should list the closest matches in a paginated view. selecting the idea should fill / replace the reference ( or wrap the selection as `[name]` ) and an import ( if required ) appropiately The lightbulb command ( `reqlan.searchReference` ) is arg-less because VS Code caches arg-bearing code-action commands and may dispose them on refresh ( error text like Actual command not found ). At invoke time the extension resolves the site via LSP `reqlan/referenceSearchSite`, or a local parse of the active editor when the language client is not yet running — it must not fail with a false " language server is not ready " when a local site can be resolved. | 0.116 | |
| base_config_seed | done | done | New bases write `.reqlan/config.json` with `$schema` pointing at the public v1 schema URL via [create_base_impl]. Seed shape: `{"$schema":"https://reqlan.com/schemas/v1/config.schema.json"}`. Semantics and discovery of applying config: [configuration.configuration_location], [configuration.configuration_import_roots], [bases.base_configuration]. Schema / editor validation: [configuration.configuration_schema_file]. A child base does not inherit a parent base's config file. Existing bases are not rewritten ( idempotent create leaves them alone ). | 0.076 | |
| welcome_webview_install | done | done | The **Welcome to Reqlan** onboarding webview is the extension-install UX surface. Opened once on first activation when `onboardingMessageShown` is false ( [welcome.installation_event] / [welcome.installation_event_trigger] / [welcome.onboarding_state] ). Content: welcome copy, resolved phonebook links, syntax-highlighted example `.rq`, open-as-untitled control, activity bar / `@reqlan` tips, sponsor thanks. Must not block [activation.activation_sequence]; errors are logged only. Reopen on demand via [welcome.show_onboarding_command] without resetting shown state unless product policy changes. Implementation: ["../../../packages/extension/src/extension/open-thanks-for-installing.ts"], ["../../../packages/extension/src/extension/onboarding-panel.ts"], ["../../../packages/extension/webviews/onboarding/"]. | 0.069 | |
| installation_event | done | done | On installation of the extension a webview should be opened showing: - A welcome message - Resolved links to site, github, docs etc. ( not template placeholders ) - A syntax-highlighted example `.rq` with real URLs ( clickable ) from ["../../../../packages/extension/templates/thanks-for-installing.template.rq"] after placeholder substitution - A control to open that example as an untitled `.rq` editor beside the webview - Short tips for the activity bar and `@reqlan` chat - A thanks to sponsors Place in overall install sequencing: ["../installation.rq".welcome_webview_install] / ["../installation.rq".extension_installation]. Implemented as a Svelte webview under ["../../../../packages/extension/webviews/onboarding"] bundled to ["../../../../packages/extension/media/webviews/onboarding"] and opened once on first activation by ["../../../../packages/extension/src/extension/open-thanks-for-installing.ts"] via ["../../../../packages/extension/src/extension/onboarding-panel.ts"]. Resource links and template values are resolved at runtime from [phonebook_reqs.phonebook] through ["../../../../packages/extension/src/shared/phonebook.ts"] and posted to the webview; the example block imports ["../../../../packages/extension/templates/thanks-for-installing.template.rq"] via Vite `?raw` and substitutes `{{…}}` placeholders before display. Highlighting uses the onboarding tokenizer in ["../../../../packages/extension/webviews/onboarding/lib/rq-highlight.ts"]. The source template still uses placeholders such as { { SITE_URL } }; those matching braces are body prose per ["../../../language/syntax-edge-cases.rq".nested_curly_braces] and must parse without errors. Activation state is stored in extension global state key `onboarding` with `onboardingMessageShown` and `lastVersion` per ["../../../../packages/extension/src/extension/onboarding-state.ts"]. VS Code persists `globalState` in the extension global storage directory ( `globalStorage/reqlan.reqlan-extension/state.vscdb` under the editor user data folder ). | 0.141 | |
| installation_event_trigger | done | done | The installation event should be checked and triggered at extension [activation.activation]. Verified: ["../../../../packages/extension/src/extension/main.ts"] schedules `openThanksForInstallingIfNeeded` after the activity-bar paint gate or 400 ms ( not in the same tick as sidebar claim ), and [activation.onboarding_check] documents the gating behaviour. | 0.060 | |
| show_onboarding_command | done | done | A [commands.command_palette] action should reopen the onboarding webview on demand. Command: Show Onboarding ( category Reqlan, id `reqlan.showOnboarding` ). Implemented by ["../../../../packages/extension/src/extension/register-onboarding-commands.ts"] calling [open_onboarding_page]. Does not reset [onboarding_state]. | 0.059 | |
| deferred_startup_sequence | done | done | Required order: 1. Load the smallest practical host entry. 2. Synchronously register commands and the activity-bar provider; return from `activate`. 3. Yield one turn, assign the activity-bar shell, and wait for the post-first-frame `ready` handshake. Do not `postMessage` until `ready`. 4. Start index discovery / sync after `ready`; if the sidebar stays closed, start after a bounded three-second fallback so startup indexing and watchers are not lost. 5. Start the language client after `ready`, with a one-second fallback so language features work when the sidebar stays closed. 6. Open first-run onboarding after the same paint signal or 400 ms, so it does not race the sidebar iframe ( [webview_claim_race] ). 7. Report incremental index progress inside the already-painted sidebar. The ready signal and both startup paths are one-shot / idempotent. [activation.background_startup] | 0.046 | |
| development_bundle_freshness | done | done | F5 executes `package.json.main` ( `out/extension/main.cjs` ), not TypeScript source. `out/`, generated webview media, and copied syntaxes are gitignored, so launching without a build can run missing or stale artifacts even when source already contains the fix. `.vscode/launch.json` therefore uses a deterministic pre-launch task that invokes the extension build orchestrator directly; root `pnpm run build:ext` delegates to the same script. The task content-fingerprints Langium and physics generation, each webview, packaging media, and extension-host / language-server output; `tsc -b` retains project-reference incrementality. Unchanged artifacts are verified and skipped before the development host opens. Watch mode remains useful during iteration, but concurrent watchers do not restart an already-running extension host or language-server child; stop and start F5 after host / server changes. ["../../../.vscode/launch.json"] ["../../../.vscode/tasks.json"] | 0.046 | |
| extension_startup_reliability | done | done | Reqlan startup must make the activity-bar shell and contributed commands available before optional indexing, parser construction, database initialization, or language-client startup can monopolize the extension host. The implementation contract is [first_paint_contract], [webview_claim_race], [deferred_startup_sequence], [lazy_runtime_dependencies], and [development_bundle_freshness]. Operational diagnosis follows [startup_diagnostics]. [activation.activation_sequence] [activitybar.loading_state] [indexer.nonblocking_index] [build.extension_bundle] | 0.046 | |
| first_paint_contract | done | done | The platform spinner ends only after VS Code loads the host module, calls `activate`, the provider registers, and `resolveWebviewView` assigns HTML. `resolveWebviewView` is async: it binds the view, yields one turn so the overlay can be claimed, then assigns the lightweight HTML / Svelte shell ( retry once on invalid-state ). It does **not** `postMessage` in that method. The webview installs its host message listener, mounts, waits for its first animation frame, then posts the one-shot `ready` signal from a queued task. Host messages start only after that handshake. Index discovery / sync waits for the same signal so synchronous parser or database setup cannot race the first visible frame. Windows overlay restore can dispose and re-resolve the sidebar; the provider rebinds and waits for a new `ready`. [activation.first_paint_startup] [webview_claim_race] ["../../../packages/extension/webviews/activity-bar/App.svelte"] ["../../../packages/extension/webviews/activity-bar/state/app.svelte.ts"] ["../../../packages/extension/src/extension/startup-gate.ts"] ["../../../packages/extension/src/shared/safe-webview-post.ts"] | 0.058 | |
| invalid_url_activation_failure | done | done | The concrete incident failed while requiring `out/extension/main.cjs`, before `activate()` ran: `TypeError: Invalid URL` in `embedPhysicsCoreSource`. Runtime `readFileSync(new URL(..., import.meta.url))` is unsafe when esbuild emits CommonJS for an ES2017 target: `import.meta.url` is unavailable / empty and source-relative assets do not exist beside the bundled entry. The HTML export used to embed physics via `html-export-assets.ts` `embedPhysicsCoreSource`. That writer is gone; export now inlines assets from ["../../../crates/reqlan-export/assets/app.js"]. Code that still needs a module directory or `createRequire` ( core engine loader and parse worker ) must not mention `import.meta` statically in a way that esbuild CJS empties: prefer CJS `__dirname` / `__filename` when those exist ( extension host bundle ), and `import.meta.url` on the Node ESM path ( CLI / MCP ). The host esbuild step defines `import.meta.url` as `import_meta_url` from this file's `__filename` so the CJS bundle stays valid and does not warn `empty-import-meta`. CLI / MCP keep the TypeScript `import.meta.url` emit. Do not eval `import.meta.url` — Node 24 treats that as a script eval and throws `Cannot use 'import.meta' outside a module`. ["../../../packages/analytical/src/native/load-native.ts"] ["../../../packages/language/src/reqlan-async-parser.ts"] ["../../../packages/extension/esbuild.mjs"] | 0.040 | |
| lazy_runtime_dependencies | done | done | Base discovery must be cheap: - Constructing `WorkspaceIndex` does not construct Langium / Chevrotain services; index sync uses the core native engine ( rusqlite ). - The host-matching `reqlan_napi` `.node` is required at activate; there is no sql. js vendor bundle in the extension host. - Export / physics source is generated at build time rather than read via extension-runtime URLs. ["../../../packages/analytical/src/index-store/workspace-index.ts"] ["../../../packages/analytical/src/index-store/sqlite-store.ts"] ["../../../packages/analytical/src/index-store/index-diagnostics-store.ts"] ["../../../packages/extension/esbuild.mjs"] | 0.046 | |
| verification_contract | done | done | Startup changes are complete only when: - The production extension build succeeds and emits `out/extension/main.cjs`. - The main bundle does not embed a sql. js vendor; index open requires the core native engine. - Extension startup-gate, webview-claim, and analytical lazy-service tests pass. - Activity-bar shell paint is visually checked in a newly restarted F5 development host; activation failures are cached for the lifetime of a failed host session. - Requirement analysis resolves the updated startup graph. | 0.040 | |
| webview_claim_race | done | done | Setting `webview.html` or calling `postMessage` before the workbench has claimed the overlay ( or after it has disposed it ) can fail with `Webview is in an invalid state` / `Webview is disposed`. This is intermittent and more common on Windows during extension init, when the activity-bar view is restored in the same tick as another webview ( first-run onboarding ). Contract: - Activity-bar host posts only after the webview's `ready` signal; `post` is a no-op until then and swallows disposed / invalid-state errors. - `resolveWebviewView` may run more than once: dispose prior view listeners, clear `webviewReady`, rebind, reassign HTML. - First-run onboarding waits for the activity-bar paint gate or 400 ms, whichever first, so it does not register a second webview service worker during sidebar claim. Palette `reqlan.showOnboarding` stays immediate. Index and language-client startup still use the existing paint / timeout fallbacks — a dead sidebar must not strand indexing. [first_paint_contract] [activation.onboarding_check] ["../../../packages/extension/src/activity_bar_module/activity-bar-webview-provider.ts"] ["../../../packages/extension/src/extension/main.ts"] ["../../../packages/extension/src/shared/safe-webview-post.ts"] | 0.091 | |
| git_dates_background_indexing | done | done | Host wiring for [git_idea_timeline_analysis] / [indexer.git_dates]: after the ideas index is ready ( and when the catalog updates ), missing idea git dates must be filled by a mostly silent background process: - small batches - yields between batches so the extension host stays interactive - no progress UI / no blocking of Activity Bar context load or Timeline tab open - session-scoped attempted set so failed lookups are not hammered repeatedly - queue priority: ideas in the active `.rq` editor file are attempted before the global backlog; switching the active editor re-prioritises ( in-flight wave may stop and reschedule ) - catalog / save scheduling follows [git_dates_rate_cap] - git / node helper processes stay hidden per ["../../core_analysis/core.rq".consumption_silence] Consumers only read indexed columns; they must not invoke `git_dates` on the UI path. [git_idea_timeline_analysis] [indexer.git_dates] [git_dates_rate_cap] [ideas_summary_webview.timeline_page] | 0.158 | |
| git_dates_rate_cap | done | done | Git history analysis is expensive ( `git log -L` and follow logs per idea ). Do not start a git_dates wave on every `.rq` save or catalog notify. Catalog and index-status events must debounce ( quiet period after the last notify ) and must wait a minimum gap after a wave that actually filled dates. Editor-switch priority and draining a capped wave may use the short idle delay. Distinct from ["./graph-analysers.rq".indexing_trigger_filesave]: the ideas index still updates on save; only git analysis is rate-capped. Related: [git_dates_background_indexing], [indexer.git_dates], ["./index.rq".idea_index_triggers], ["./index.rq".git_information_capture], ["../../core_analysis/core.rq".consumption_silence]. | 0.123 | |
| git_idea_timeline_analysis | done | done | Idea creation and last-modified dates, plus change count, are derived from git history by the `git_dates` analyser and stored on the ideas index ( `git_created_at` / `git_modified_at` / `git_change_count` ). Calculation is owned by [git_dates_background_indexing] / [indexer.git_dates] — not by on-the-fly editor, Activity Bar focus, Timeline open, or Ideas table load. Active-editor queue priority for fill-in is specified on [git_dates_background_indexing]. Implemented by ["../../../crates/reqlan-index/src/git_dates.rs"] and ["../../../packages/analytical/src/index-store/sqlite-store.ts"]. [git_dates_background_indexing] [git_dates_rate_cap] [indexer.git_dates] [ideas_summary_webview.ideas_list] | 0.091 | |
| indexing_trigger_auto | done | done | When the editor host is idle, run a **background staleness check** — not a reindex and not a soft full sync UI pass. Philosophy: computationally cheap. Load document mtimes in **one** SQLite read, compare each file's FS mtime in memory, and only then index mismatched or new files; drop deleted docs in one batched delete. A single MAX ( mtime ) watermark is not sufficient ( one file can get newer while another gets older ). If nothing is stale: no state transition, no progress UI, no parse. If some files are stale: index **only those files** under [indexer.nonblocking_index]. Schedule after a quiet period with no index activity; prefer running when the window is unfocused; defer or cancel if the user becomes active again. Extension host only ( [editor_index_adapter] ); CLI / MCP have no idle loop ( they sync on activate / command ). | 0.113 | |
| index_diagnostics | done | done | Indexing performance diagnostics: measure and inspect time spent indexing a [indexer.index] base. Timing is persisted in a dedicated **diagnostic index** under application memory ( [app_memory.index_diagnostics_store] ), not in `ideas-index.sqlite`, so history survives Clear & rebuild. Surfaces: [index_diagnostics_webview]. Owned by `@reqlan/analytical` for recording; the extension hosts the webview. | 0.145 | |
| index_diagnostics_metrics | done | done | For each sync / index run on a base, record: - wall-clock **sum / total duration** of the pass - **file count** ( visited ), skipped-by-mtime count, indexed ( parse / persist attempted ) count, error count - **average path depth** of visited files ( segments under the base root ) - per-file **duration** and outcome ( `mtime_skip` | `mtime_refresh` | `hash_skip` | `persisted` | `error` ) Runs are attributable to a trigger ( `soft_sync` | `rebuild` | `enqueue` | `stale` ). Instrumentation wraps [workspace_index] soft sync and [indexer.nonblocking_index] single-file paths via [diagnostics_store]. | 0.120 | |
| index_diagnostics_webview | done | done | An editor webview panel displays index timing diagnostics for the active base. Capabilities: - Base / latest-run summary: total time, file counts, average depth - Drill into recent runs - Rank files by duration ( slowest first ); show outcome and path depth Opened by command `reqlan.openIndexDiagnostics` ( [open_index_diagnostics] ). Optional entry from workspace index health ( [workspace_pane.workspace_pane] ) may link later; command is the primary entry. Implementation: [diagnostics_panel], [DiagnosticsApp]. | 0.145 | |
| open_index_diagnostics | done | done | Command palette action **Open Index Diagnostics** ( `reqlan.openIndexDiagnostics` ) opens [index_diagnostics_webview] for the active base. Shows total index time, run drill-down, and per-file ranking ( duration, depth, outcome ). | 0.071 | |
| application_memory | done | done | Base-local application memory for reqlan tools lives under `<base>/.reqlan/`. The presence of `.reqlan` **marks** a directory as a ["../../bases/base.rq".base]; each base owns its own store. Shared by the VS Code extension, ["../../cli/cli_package.rq".cli_package], and MCP ( [mcp_server] ) — not VS Code `globalStorageUri` / extension-private storage. The ideas graph index ( [indexer.index_ideas] and [indexer.index_technology] ) persists as `ideas-index.sqlite` inside that directory. Schema: ["../../indexer/sqlite-schema.rq".ideas_index_sqlite]. Index timing diagnostics persist separately as `index-diagnostics.sqlite` ( [index_diagnostics_store] / ["./index-diagnostics.rq".index_diagnostics] ) so rebuilds of the ideas index do not wipe history. Click session state persists as `click-sessions.sqlite` ( ["../../cli/click.rq".click_session] ) so agent / CLI click cursors survive ideas-index rebuilds. Diagnostics schema: ["../../indexer/diagnostics-schema.rq".diagnostics_sqlite]. WAL sidecars and JSON siblings: ["../../indexer/sidecars.rq".reqlan_directory_layout]. Export form defaults ( all formats via ["../export/export.rq".export_webview], including HTML options in ["../export/html-export.rq".html_export_form] ) persist as `export_settings.json` in the same directory ( active base ). Path ignore rules live as `.rqignore` in the same directory ( [rqignore] / [configuration.configuration_rqignore] ). Git ignore for SQLite stores lives as `.gitignore` in the same directory ( ["../../bases/base.rq".base_initialisation_ignore] ). Path resolution is owned by [application_memory_impl] ( `resolveApplicationMemoryPath` ); consumers must use it ( or `REQLAN_INDEX_PATH` override for a single store ) so tools read and write the same store for a given base. Multi-base discovery / registry: ["../../bases/base.rq".multi_base_environment]. Index engine ownership: ["../../indexer/indexer.rq".ownership]. Extension wiring: [analytical_submodule]. Aligns with ["../../cli/cli_package.rq".function_parity]: shared analytical memory, not host-only state. | 0.494 | |
| binary_ignore | done | done | Discovery, coverage, and index walks skip binary files by default. Built-in [rqignore] patterns include a broad set of binary globs ( images, archives, packages, native objects, fonts, media, and office files ). Opt in with a gitignore negation in `.reqlan/.rqignore`. Example: `!*.bin` includes files with a `.bin` extension. Opted-in files can receive ["../../indexer/indexer.rq".index_code_files] comment indexing. They are not parsed as `.rq` sources. The editor source watcher stays on allowlisted text extensions. A soft sync indexes opted-in binary paths from the scan list. Idle stats stored paths and does not walk. Same filter as [configuration.configuration_rqignore]. | 0.201 | |
| commentless_file_ignore | done | done | Discovery, coverage, and index walks skip common files that have no comment syntax. The set includes JSON ( `.json`, `.jsonl`, `.ndjson` ), tabular data ( `.csv`, `.tsv` ), lockfiles, checksums, minified scripts, source maps, notebooks, and certificates. These files cannot hold `rq:` comment references. SQLite WAL, SHM, and journal sidecars stay with the database patterns in [rqignore] ( `*.sqlite-wal`, `*.sqlite-shm`, `*.sqlite-journal`, and the same suffixes on `.sqlite3`, `.db`, and `.db3` ). Opt in with a gitignore negation in `.reqlan/.rqignore`. Example: `!*.json`. JSON with comments ( `.jsonc`, `.json5` ) stays eligible for ["../../indexer/indexer.rq".index_code_files]. The seeded file lists the same patterns ( [rqignore_initialisation] ). | 0.188 | |
| index_diagnostics_store | done | done | Per-base diagnostic timing index at `<base>/.reqlan/index-diagnostics.sqlite`. Sibling of `ideas-index.sqlite`; not deleted by Clear & rebuild. Records sync runs and per-file durations for ["./index-diagnostics.rq".index_diagnostics_metrics]. Webview: ["./index-diagnostics.rq".index_diagnostics_webview]. | 0.223 | |
| rqignore | done | done | `.reqlan/.rqignore` filters which paths discovery and indexing crawl, using gitignore syntax relative to the base root. Owned by [rqignore_impl] in `@reqlan/analytical` so CLI and editor share behaviour. Defaults cover `node_modules/`, virtualenvs, build dirs ( `bin/`, `generated/`, `out/` ), VCS / editor stores, database files ( `*.db3`, sqlite ) and SQLite sidecars, binary globs ( [binary_ignore] ), `*.html`, secrets, and hidden entries. Walks also skip common files that have no comments ( [commentless_file_ignore] ). New bases seed the file via [rqignore_initialisation]. Product configuration statement: [configuration.configuration_rqignore]. | 0.296 | |
| rqignore_initialisation | done | done | New bases seed `.reqlan/.rqignore` with comments that explain every feature of [rqignore], [binary_ignore], and [commentless_file_ignore]. Written when a base is created: ["../host/installation.rq".base_rqignore_seed] / ["../../bases/base.rq".base_initialisation]. Comments cover purpose ( discovery and index walks ), base-root relative paths, gitignore syntax, last matching rule, `!` negation, built-in defaults that always apply, binary opt-in, hidden paths, `*.secret.rq`, SQLite sidecars, files that have no comments ( [commentless_file_ignore] ), and `.reqlan/` itself. The seeded file is the editable surface. Built-in rules still apply if a user deletes a line. Existing `.rqignore` files are not rewritten ( ["../host/installation.rq".base_installation] is idempotent ). | 0.125 | |
| text_prefix_skip | done | done | Before any index content read, inspect at most the first 8 KiB. Skip the file when that prefix has a NUL, a UTF-16 BOM ( `FF FE` or `FE FF` ), or invalid UTF-8. A trailing incomplete UTF-8 code unit is not a skip: the rest of the file may finish the character. Skip when the file is larger than 1 MiB. A rejected prefix is not followed by a read of the remainder. `.rq` files and allowlisted code files use the same check. This is in addition to [binary_ignore] extension globs. | 0.092 | |
| analysis_api_dispose | done | done | [analysis_runtime] `openAnalysisApi` dispose must close the native analysis runtime and drop its SQLite connection ( ideas index and click-sessions ). A no-op dispose is a defect. Extension commands use a short-lived `withAnalysisApi` ( open → run → dispose ) mediated by [analysis_runtime]; they must not cache `NativeAnalysisApi` for the session. | 0.040 | |
| event_driven_base_access | done | done | Extension interaction with bases is event-driven and goes through the analytical package: - Discover / pin / focus / file watch / command / UI query → open or `ensureBaseReady` - Sync, stale check, or analysis work runs - Quiet idle → [release_when_idle] No always-on connection and no continuous reconnect loop. Headless CLI / MCP already open per command via [analysis_runtime] ( [headless_sync_on_ready] ). | 0.057 | |
| release_when_idle | done | done | After an index or analysis pass finishes, and after the idle quiet period ( [idle_stale_check] / indexing_trigger_auto ), [editor_index_adapter] must release every open base SQLite handle via [base_registry] `releaseArtifacts` → [workspace_index] `deactivate`. Descriptors stay registered so the next event can reopen the same base. Idle must not leave connections open after a heal / sync, and must not schedule another open unless a real host event arrives ( file change, pin / switch, focus reopen, command, UI query ). | 0.077 | |
| sqlite_artifact_lifecycle | done | done | Base SQLite artifacts under `<base>/.reqlan/` ( [base_ownership] ) must not stay locked for the whole extension session. The host opens them only while work runs, then releases file handles so users can delete or replace the files. Aligns with [event_driven]: open and close are reactions to index or analysis events, not a permanent connection or a reconnect loop. Mediation: only [workspace_index] / [analysis_runtime] own SQLite connections. The extension adapter ( [editor_index_adapter] ) must not keep a second long-lived analysis runtime beside the index, and must not call native open without a matching dispose / release. Related: [index_sequencing], [idle_stale_check], [headless_consumers], [multi_base_environment]. | 0.052 | |
| comment_reference_import_root_alias | done | done | Qualified comment references use the same path forms as imports, including import-root aliases such as `@/`. `rq:["@/path/relative/to/import/root.rq".idea]` resolves against the alias's import-root directory ( workspace folder by default ), not the source file directory. `@reqs` without a following `/` is not an alias. Configured aliases and roots follow ["../../../language/imports.rq".configuration_import_root_alias] and ["../../host/configuration.rq".configuration_import_roots]. The language server applies this when resolving comment references in `.rq` files. The extension applies the same core resolver for `rq:[…]` links in non- `.rq` source. Related: ["../../../language/syntax.rq".comment_reference], [references_in_functional_code_comments], ["../../../language/imports.rq".import_paths]. | 0.235 | |
| comment_reference_language_blocklist | done | done | Comment rq editor features ( goto, diagnostics, ignore actions, navigate-back origin tracking ) apply to every language except a small blocklist. The blocklist is `reqlan` plus non-source editor buffers ( log, search-result, SCM input, git commit / rebase, github-issues ). The rust language client uses a file-scheme document selector and skips sync for blocklisted languages. TextMate resting underline injects into a wide `injectTo` scope list ( best-effort; goto still works without a matching grammar scope ). Related: [references_in_functional_code_comments], [comment_reference_resting_underline], ["../../../language/syntax.rq".comment_reference]. | 0.178 | |
| comment_reference_navigate_back | done | done | A go-to-definition JUMP from a comment `rq:[…]` host must push the pre-jump caret onto the same navigate-back stack as `.rq` jumps. The stack tracks every non-blocklisted language ( ["../../../../packages/extension/src/extension/comment-reference-languages.ts"] ) as well as `.rq`. Only landings in `.rq` / `reqlan` are recorded, so other language servers do not fill the stack. Alt + Left ( `reqlan.navigateBack` ) restores that caret when `reqlan.canNavigateBack` is set, in any focused editor. Related: [references_in_functional_code_comments], [comment_reference_language_blocklist], ["../syntax/features-syntax-highlighting.rq".editor_reference_activation]. | 0.114 | |
| comment_reference_resolution_error_state | done | done | [comment_reference_resolution_error] The language server must share one resolution result for clickable links and the error underline. Resolve comment references once against the loaded idea index, then reuse that result for both consumers. When you create an idea that matches a comment reference, the language server must clear the underline and the missing-idea message. Resolution and the index should be performed once and shared between consumers. | 0.053 | |
| comment_reference_resting_underline | done | done | In non- `.rq` source comments, every `rq:[…]` token must show a resting underline so it is clear the site is link-shaped. Use TextMate scope `meta.link.reqlan markup.underline` via the comment injection grammar. Do not rely on file document links for that cue — those stay stripped ( ["../syntax/features-syntax-highlighting.rq".editor_reference_activation] ). Navigation remains Ctrl + click go-to-definition for resolved sites ( [references_in_functional_code_comments] ). Related: [references_in_functional_code_comments], ["../syntax/features-syntax.rq".comment_support]. | 0.146 | |
| references_in_functional_code_comments | done | done | Comments in functional code should be able to reference reqlan ideas using the an `rq:` prefixed to an idea: `rq:[idea]` as demonstrated in python, but actually applicable to any language's comment context: Editor hosts are every language except [comment_reference_language_blocklist]. With the rust language server, a resolved `rq:[…]` site is Ctrl + click navigable via go-to-definition. File document links stay out of the editor ( ["../syntax/features-syntax-highlighting.rq".editor_reference_activation] ). Comment `rq:[…]` sites use TextMate `meta.link.reqlan markup.underline` so the resting editor shows they are link-shaped ( [comment_reference_resting_underline] ). After that goto, Alt + Left restores the comment-host caret ( [comment_reference_navigate_back] ). Missing ideas keep the error underline and have no definition target. Relates to ["../../mutation/refactor_support.rq".comment_reference_refactor_support], [comment_reference_import_root_alias], ["../support/language-server-rust.rq".language_server_rust], ["../syntax/features-syntax-highlighting.rq".editor_reference_activation]. | 0.300 | |
| import_code_completion_auto_file_import | done | done | When a code completion for an idea in another file is selected, the import should be automatically added to the current file. This prevents the user from selecting a presented idea, and then having to add the import for the same thing. Uses the same from-import insertion as [import_error] quick fixes; skips the edit when the idea is already imported. [reference_code_completion] [import_code_completion] [import_code_completion_rendering] [import_error] | 0.066 | |
| import_code_completion_rendering | done | done | Reference code completion for ideas from other files shows the relative import path in the completion detail. Same-file ideas omit the path detail. [reference_code_completion] [import_code_completion] [import_code_completion_auto_file_import] | 0.066 | |
| import_error | done | done | " Could not resolve reference " errors offer Quick Fix code actions ( lightbulb and editor right-click Quick Fix ). Fixes: - Add import from a matching idea found in loaded documents or the name catalog pushed from the index - Rewrite the reference to an anonymous qualified form `["path".name]` - Search the index for matching ideas, ideasets, or files ( same `reqlan.searchReference` path as [code_actions] / ["../../host/features-commands.rq".search_code_actions] ), then insert an import - Create the idea in a new. rq file and import it [general_errors] [code_actions] [quick_fixes] | 0.078 | |
| import_folder_targets | done | done | An import path may refer to a folder. A folder is a valid import target when that path exists as a directory. Namespace import of a folder binds the alias to that folder, the same as a file. Extensionless paths still prefer a `.rq` file: [implicit_file_extension]. Related: ["../../../language/imports.rq".import_paths], [import_does_not_exist_error], ["../syntax/features-syntax.rq".file_references]. | 0.164 | |
| index_file_search_coalesce | done | done | Index startup does not call `findFiles` or ripgrep / `searchtext.exe`. The native workspace scan ( ["../../../indexer/cache-reuse.rq".single_sync_walk] ) is the one directory walk. Overlapping rediscover and sync callers share that in-flight scan. The scan applies `.rqignore` ( ["../../../indexer/indexer.rq".nonblocking_index] ). Related: [lsp_file_watch], ["../../index-host/index.rq".rqignore], ["../../../indexer/sequencing.rq".scan_limits]. | 0.051 | |
| lsp_file_watch | done | done | The language server must register workspace file watchers for `.rq` files only. It must not register a recursive all-files watcher. Watched-file events whose path contains a `node_modules`, `.git`, `.reqlan`, `out`, or `dist` segment must not start a document rebuild. Directory URIs without a trailing slash, Windows paths, and percent-encoded URIs must use the same segment rules. A kept `.rq` change replaces that buffer's model when the text changed ( [ast_lifecycle] ). Related: [language_server], [index_file_search_coalesce], [open_file_algorithm]. | 0.085 | |
| file_reference_errors | done | done | If a [file_reference] is not found, is broken, or cannot be resolved, the language server must show an underline. Do not show the underline when [comment_reference_ignore] ( `//rq-ignore-error` ) is on the line before. Missing files: error underline, no document link. Do not underline examples inside [inline_code] or ["../../../language/syntax.rq".code_snippets]. Empty files and files with parse errors: warning underline; the document link stays. Import-root aliases such as `@/` resolve the same way as ["../../../core_analysis/check.rq".check] and ["../../../language/imports.rq".configuration_import_root_alias]. Related: [general_errors], [comment_reference_resolution_error], [inline_code], ["./features-imports.rq".import_does_not_exist_error], ["../syntax/features-syntax.rq".file_references]. | 0.092 | |
| language_server_rust | in-progress | in-progress | A Rust binary replaces the Langium language-server process. The extension launches `reqlan-lsp` over stdio. Open `.rq` buffers parse with [parser_rust]. The open sequence is ["open-file-sequencing.rq".open_file_algorithm]. Workspace names and rename targets come from the ideas index under [application_memory]. Inbound hints wait for the host snapshot of that index. The inbound inlay hover is markdown. It lists each referencer as a file link. An idea hover shows styled idea content when ["../syntax/features-syntax.rq".idea_content_tooltip] is on. A closed file has no retained syntax tree. Go-to-definition cost is ["../syntax/features-syntax-highlighting.rq".reference_goto_definition_speed]. When the engine is `rust`, the server also serves `rq:[…]` document links, go-to-definition, and missing-idea diagnostics in non- `.rq` comments. One resolution supplies the link, the definition target, and the underline. The host strips file document links; Ctrl + click uses definition ( ["../syntax/features-syntax-highlighting.rq".editor_reference_activation] ). The host serves those links when the engine is `node`. Host commands prefer LSP custom requests, napi extract, and the ideas index. Move / delete planners still use Langium until index-backed planners land. The setting `reqlan.languageServer.engine` selects the binary. The default is `rust`. `node` remains only when the setting says so, or when the binary is not on disk yet. A local extension build compiles this binary with the napi addon ( [extension_host_native_stage] ). The extension launches `bin/reqlan-lsp`, else that release build. It never launches a debug build, because a debug build is slow and the extension build does not refresh it. Code actions match ["../vsc-primitives.rq".code_actions]: ignore, create below, import fixes, search or wrap, idea move and delete, and barrel page. Attribute and name catalogs drive completion. Rename and find-references use open buffers plus the ideas index. Inlay hints honour `reqlan.referencesInlayHints.enabled` and the host inbound snapshot. An import link uses the parser token span ( [import_from] ). The path links to the file. An imported idea name and its alias each link to that declaration. A namespace reference links the alias and the idea as separate names. A missing leaf reports an unresolved idea on that name. Related: [language_server], [grammar_and_parser], [analytical_rust_port], [vsix_export]. | 0.580 | |
| ast_lifecycle | done | done | The open buffer has one model. That model is the Rust parse of this buffer. [local_symbolic_analysis] is a file-local extract. The hot path does not run it as a second parse. Related: [open_file_algorithm], [open_file_sequence_diagram], [outbound_one_hop], [token_colour_sequence], [parse_budget_timeout], ["./initialisation-and-sequencing.rq".lsp_file_watch], [language_server_rust]. Populate. Parse this buffer when the editor opens it. Semantic tokens, same-file links, and the import list come from that parse. Replace. When this buffer text changes, parse again and replace the model. Do not patch the old model. A watched `.rq` change uses the same replace path ( [document_update_handler] ). When the text is unchanged, keep the current model. Drop. When the editor closes the buffer, drop the model. A closed file keeps no syntax tree ( [language_server_rust] ). Import files. [outbound_one_hop] stores a name list for each imported file. That list is not an open-buffer model. A change to an imported file does not replace this buffer's model. Read the imported file again when its content hash changes. Incomplete parse. When the budget fails, store an empty model plus diagnostics ( [parse_budget_timeout] ). Idea names stay empty until a later successful parse. The node engine follows [node_engine_open_path]. | 0.249 | |
| comment_backlink_sequence | in-progress | in-progress | `rq:[idea]` and `rq:["path".idea]` in code comments use one resolution for the link and the underline. A qualified path reads that `.rq` file, parses idea names, then links or errors. That read belongs to the comment file. Opening a different `.rq` buffer must not re-resolve every comment buffer. An unqualified name searches open `.rq` buffers first. The index snapshot may add a match later ( [lazy_features_after_outbound] ). Do not call `list_all_ideas` for that search. Do not create a document link when the idea is not confirmed. Related: [comment_reference_resolution_error_state]. | 0.103 | |
| file_crawl_sequence | done | done | Workspace crawl is the indexer ( [index], [index_sequencing] ). It is not the open-file path. Native sync walks `.rq` files, then allowlisted code files, and writes SQLite. Open-file links must not wait for that walk ( [open_file_algorithm] ). The language server watches `.rq` files only ( ["./initialisation-and-sequencing.rq".lsp_file_watch] ). Related: [extension_startup_sequence], [background_startup]. | 0.040 | |
| lazy_features_after_outbound | done | done | After the hot path is on screen, show the rest when its data is ready. This pass must not delay [open_file_algorithm] steps 1 – 5. Related: [language_server], [unify_inbound_indexes], [background_startup], [inlay_hints], [codelens]. Work: - Confirm each qualified reference. Read that target file once, at depth 1. Then paint the idea line or the missing-idea error ( [missing_reference_colour_sequence] ). - Inbound backlinks: inlay hints and the file-top CodeLens, from the host snapshot ( [unify_inbound_indexes] ). - Attribute and name catalogs for completion and quick fixes. - Unqualified comment matches from that snapshot ( [comment_backlink_sequence] ). Until the first ready snapshot arrives, inlay hints and the file CodeLens stay empty. Those handlers do not open SQLite and do not call `list_all_ideas`. Later updates are incremental ready-state events ( ["../../../indexer/cache-reuse.rq".inbound_snapshot_events] ); last-good stays until replaced. When a ready snapshot arrives, refresh the visible editors. | 0.283 | |
| missing_reference_colour_sequence | done | done | Paint links and errors in two steps. The hot path paints same-file names and import targets ( [outbound_one_hop] ). The background pass paints qualified-reference idea lines ( [lazy_features_after_outbound] ). Related: [unresolved_reference_diagnostics], [reference_rendering], [file_reference_errors], [import_does_not_exist_error]. Hot path: - URL references: document link, no error. - Wildcard references: document link, no error underline for zero matches. - Import path missing: error underline, no document link. - Imported idea present: document link to that idea line. - Imported idea missing: error underline, no document link. - Same-file idea missing: error underline, no document link. - Qualified reference: document link to the file from the path token. Do not underline a missing idea yet. Background pass: - Qualified idea present: document link to that idea line. - Qualified idea missing: error underline, and remove the document link. - Drop a false missing-idea error only for the range that this pass confirms. Keep a different missing name on the same line. | 0.153 | |
| native_cache_retention | done | done | The language server keeps the open-buffer model for each open editor. It keeps a depth-1 name list for each imported file, and for each qualified-reference file the background pass has confirmed, while an open editor still needs that list. A closed editor drops its model. It also drops name lists that no open editor still needs. A file-watcher delete drops that URI from both caches. A check of a closed file may parse it once. That result is not stored. Related: [open_file_hot_path], [outbound_one_hop], [ast_lifecycle]. | 0.072 | |
| node_engine_open_path | done | done | This idea applies when `reqlan.languageServer.engine` is `node`. The default engine is [language_server_rust]. The node server populates a Langium AST for the open buffer at `DocumentState.Parsed`. Later builder phases attach scopes, links, and diagnostics. They must not replace that AST. Parsed-time handlers must not read Langium `.ref`. Use `$refText`, the symbolic extract, or `peekResolvedRef` until `ComputedScopes`. The linker must not resolve a cross-reference while that document is below `ComputedScopes`. Do not log the Langium warning. Do not store a failed link. Scope lookup, code lens, completion, and idea summaries use `readResolvedRef` or `peekResolvedRef` until that phase. When another `.rq` file changes, relink this document when the index says it is affected, when comment references name that file, or when this document's outbound targets include the changed URI. Do not relink every document that still has a linker error. Workspace load parses at most 8 files at a time ( [workspace_parse_concurrency] ). `fileIdeasCache` and `importPathCache` drop a URI when that document is removed ( [scope_cache_release] ). Related: [open_file_algorithm], [ast_lifecycle]. | 0.133 | |
| open_file_algorithm | done | done | When the editor opens or changes a `.rq` buffer, `reqlan-lsp` runs this sequence. Do not wait for the ideas index. Do not wait for a workspace crawl. Detail: [open_file_sequence_diagram]. Related: [open_file_reference_sequencing], [language_server_rust], [ast_lifecycle], [outbound_one_hop], [open_file_hot_path], [token_colour_sequence], [missing_reference_colour_sequence], [lazy_features_after_outbound]. The hot path finishes before the first diagnostics publish. 1. Parse this buffer once ( [ast_lifecycle] ). 2. Colour tokens from that parse ( [token_colour_sequence] ). 3. Emit same-file links and same-file missing-idea errors from that parse. 4. Parse each imported file once, at depth 1, for idea names and lines ( [outbound_one_hop] ). Link an imported name that exists. Underline an imported name that does not ( [missing_reference_colour_sequence] ). 5. Return. A qualified reference may link to its file from the path token. Do not read that file on this path. After that publish, a background pass fills the rest ( [lazy_features_after_outbound] ). That pass must not delay steps 1 – 5. The server sends the hot publish before it starts that pass. It runs that pass only while no client message waits. A hover or a definition request goes first. A change reuses the import roots loaded at open. It does not read the config again. The node engine keeps [node_engine_open_path]. | 0.330 | |
| open_file_hot_path | in-progress | in-progress | Open-file work stays on the one parse of this buffer plus one parse per imported file. Reuse that model for document links, go-to-definition, hover, and diagnostics until the text or the import-root mapping changes. Keep that model, and the import name lists, only for open editors ( [native_cache_retention] ). Look up an open import by URI. Do not match on file basename. Do not read SQLite. Do not call `list_all_ideas`. At parse time, publish lex and parse diagnostics with same-file and import errors. The editor must not wait for a workspace link phase to see a broken parse. Handlers on this path read the open model. They do not build a workspace link graph. Related: [open_file_algorithm], [outbound_one_hop], [ast_lifecycle], [comment_backlink_sequence]. | 0.186 | |
| outbound_one_hop | done | done | The hot path parses imported files only. For each import in the open buffer, parse that file once, then stop. Read idea names, idea lines, and import aliases. Do not parse files that the import references. Do not read SQLite. Prefer the open buffer text when that import is already open and dirty. Cache each parse by path plus file stamp ( modified time and size ) ( [lsp_session] ). A current stamp is one `stat`. Read and parse the file only when the stamp changed. Cache the open buffer by URI plus text fingerprint ( [open_file_hot_path] ). [local_symbolic_analysis] stays file-local. This hop is outside that extract. An extensionless import path prefers `.rq`, then the path as written ( [implicit_file_extension] ). A qualified reference that is not an import waits for [lazy_features_after_outbound]. Related: [open_file_algorithm], [non_web_reference_navigation]. | 0.246 | |
| scope_cache_release | done | done | This idea applies when the engine is `node` ( [node_engine_open_path] ). `fileIdeasCache` and `importPathCache` drop a URI when that Langium document is removed. A version change still replaces the entry for a document that stays. The rust server drops the open model in [native_cache_retention]. Related: [open_file_hot_path], [native_cache_retention]. | 0.057 | |
| token_colour_sequence | done | done | Token colour has two layers. They must not wait for the ideas index. TextMate paints first from the buffer text. Semantic tokens paint from the one parse ( [ast_lifecycle] ). A second extract does not paint tokens. A missing reference keeps the same token colour. The error underline is a diagnostic ( [missing_reference_colour_sequence], [reference_rendering] ). Related: [syntax_highlighting]. | 0.117 | |
| workspace_parse_concurrency | done | done | This idea applies when the engine is `node` ( [node_engine_open_path] ). Workspace load parses `.rq` files with a fixed limit of 8 in flight. One file that fails to parse does not cancel the other files. The steady-state set is still every workspace `.rq` file. The rust server does not parse the workspace. A closed file keeps no syntax tree ( [language_server_rust] ). Related: [ast_lifecycle], [workspace_manager]. | 0.057 | |
| aliased_from_import_ctrl_click | done | done | A from-import alias that contains the imported idea name as a suffix or substring must offer one Ctrl + click destination for the whole written alias. Example: `from "@/_charter/plant-ontology.rq" import simulation as ontology-simulation` then `[ontology-simulation]`. The document-link underline must cover `ontology-simulation`, not only `simulation`. Ctrl + click on any character of that alias must open the imported idea in the import-root target file. Do not resolve the import path a second time against the current file directory. Do not offer a second interactive link on the original idea-name suffix. Related: [reference_document_links], [reference_underline_syntax_align], [non_web_reference_navigation], ["../../../language/imports.rq".import_from], ["../../../language/imports.rq".configuration_import_root_alias]. | 0.086 | |
| code_lens_reference_types | done | done | References should have a ["../vsc-primitives.rq".codelens] button. The button should be togleable in settings. The button should the classification of the reference. e. g. ' open reqlan file', " open { extension } file', ' open folder', ' open idea' Clicking the CodeLens should open a reference card with summary stats — not navigate, because references already act as editor links. The card should offer open / reveal / file-selector actions when useful. Folders should be handled by focusing on the ide explorer, or by opening a file selector. Some summary stats should be shown for references - e. g. last edited, count of references, count of referencers, etc. File-top inbound file references are [file_inbound_code_lens], not this outbound classification lens. Implemented by ["../../../../packages/language/src/reqlan-code-lens-provider.ts"], ["../../../../packages/language/src/reqlan-reference-code-lens.ts"], ["../../../../packages/language/src/reqlan-code-lens-settings.ts"], ["../../../../packages/extension/src/extension/register-reference-code-lens.ts"], and ["../../../../packages/extension/src/extension/reference-code-lens-card.ts"]. [file_inbound_code_lens] | 0.069 | |
| editor_reference_activation | done | done | A file reference and an idea reference have no underline in the resting editor. The editor shows an underline when the user holds Ctrl and the pointer is on that reference. Ctrl + click follows the reference in the active editor. A target in the open file moves the cursor in that editor. A target in another file opens in the active editor group. The editor does not open a second copy of the current file. The editor does not open a browser page for a file or idea reference. Alt + Left is navigate-back. It restores the caret from before go-to-definition. The extension keeps that JUMP on a reqlan stack because Cursor often skips the editor history entry for same-file definition. The stack also records jumps that start in comment-reference host languages ( ["../comment-references/functional-code-comment-references.rq".comment_reference_navigate_back] ). Definition targets prefer an open workspace tab URI so Cursor does not open a detached file copy. The definition target range must move the caret ( not an empty 0: 0 – 0: 0 range ). A click that is already on the declaration name is not a JUMP. The TextMate scope is `meta.link.reqlan`. A `markup.underline` scope forces an underline in the resting editor. `.rq` file and idea references omit `markup.underline` — no resting underline. Comment `rq:[…]` sites add `markup.underline` via the comment injection grammar ( ["../comment-references/functional-code-comment-references.rq".comment_reference_resting_underline] ). A file document link is always underlined, and its `#L` URI opens another editor. The extension does not give the editor a file document link. Ctrl + hover uses the definition request. That request is [reference_goto_definition_speed]. The interactive underline is the lexer token under the cursor: the path string or the idea id from the parser. Definition returns `LocationLink.originSelectionRange` for that token so the editor does not invent a word from brackets or quotes. Path segment names inside a quoted path are not reference tokens. The host does not register a DefinitionProvider for other languages, and does not rescan open source files for this activation. When the engine is rust, the language-server document selector includes all file-scheme documents except the comment-host blocklist, and definition navigates resolved `rq:[…]` sites ( ["../comment-references/functional-code-comment-references.rq".references_in_functional_code_comments], ["../comment-references/functional-code-comment-references.rq".comment_reference_language_blocklist] ). A web URL and a command link ( folder, wildcard ) stay document links. A click on a web URL opens that URL in the browser. Related: [reference_document_links], [non_web_reference_navigation], [reference_goto_definition_speed]. | 0.254 | |
| file_inbound_code_lens | done | done | The editor shows inbound file references in a ["../vsc-primitives.rq".codelens] at the top of the file. The CodeLens uses the same `@referenced-by: (name, …)` list form as idea inlay hints ( [view_references_as_inlay_hints] ). Idea declarations keep `@referenced-by` for idea-to-idea inbound. The file CodeLens lists ideas that point at this file with a ["../../../reference_types.rq".file_reference] ( path-only and path-with-symbol ). Click the CodeLens to open the inbound referencers summary for this file. Hide the CodeLens when the file has no inbound file references. Read referencers from the SQLite inbound snapshot ( ["../../../indexer/cache-reuse.rq".unify_inbound_indexes] ). Show the CodeLens when that snapshot arrives ( ["../support/open-file-sequencing.rq".lazy_features_after_outbound] ). Do not scan Langium IndexManager. The file CodeLens is independent of the outbound classification toggle on [code_lens_reference_types]. Applies to `.rq` files and to other indexed files that ideas reference. Implemented by ["../../../../packages/language/src/reqlan-code-lens-provider.ts"], ["../../../../packages/language/src/reqlan-file-inbound-code-lens.ts"], ["../../../../packages/language/src/reqlan-inbound-snapshot.ts"], ["../../../../packages/analytical/src/core/inbound-file-referencers.ts"], ["../../../../packages/analytical/src/index-store/sqlite-store.ts"], ["../../../../packages/extension/src/extension/inbound-file-snapshot.ts"], ["../../../../packages/extension/src/extension/register-inbound-sqlite-sync.ts"], ["../../../../packages/extension/src/extension/register-file-inbound-code-lens.ts"], ["../../../../packages/extension/src/ideas_summary_module/register-webview.ts"], and ["../../../../packages/language/src/reqlan-inlay-hint-settings.ts"]. [view_references_as_inlay_hints] [code_lens_reference_types] [inbound_inlay_index_performance] | 0.120 | |
| import_statement_highlighting | done | done | TextMate highlighting for import lines must colour the whole statement, not only the leading `from` / `import` keyword. After `import` ( and after `from` … `import` ), the quoted path, the `import` / `as` keywords, and idea / alias names must receive scopes — they must not stay default-white while still looking interactive from a path document link alone. Root cause: a short `^\\b(from|import)\\b` match ran before the path patterns and consumed the keyword so the path never scoped. Fix: line-scoped `meta.import.reqlan` begin / end rules that paint path, keywords, and names. Semantic tokens still paint import AST nodes ( and must `return` after handling them ). | 0.064 | |
| inbound_inlay_hints_whole_graph | done | done | Inbound references shown in [view_references_as_inlay_hints] should include referencers from the whole requirement graph across the workspace, not only ideas declared in the current file. Coverage comes from SQLite ["../../../indexer/sqlite-schema.rq".edges_table] via ["../../../indexer/cache-reuse.rq".unify_inbound_indexes], not from Langium IndexManager. Whole-graph coverage must not regress to per-idea full-index scans — see [inbound_inlay_index_performance]. Implemented by ["../../../../packages/language/src/reqlan-inbound-reference-inlay-label.ts"], ["../../../../packages/language/src/reqlan-inlay-hint-provider.ts"]. [view_references_as_inlay_hints] [traceable_inlay_hints] [inbound_inlay_index_performance] | 0.056 | |
| inbound_inlay_index_performance | done | done | Computing inbound reference inlay hints for a document with many ideas must look up a pre-pushed SQLite inbound snapshot ( ["../../../indexer/cache-reuse.rq".unify_inbound_indexes] ) in near-constant time per declaration. Show the hints when that snapshot arrives ( ["../support/open-file-sequencing.rq".lazy_features_after_outbound] ). Until the first ready snapshot arrives, hints stay empty. Later index sync must not clear last-good with an empty mid-sync push ( ["../../../indexer/cache-reuse.rq".inbound_snapshot_events] ). Must not call Langium `findAllReferences` / equivalent full-index scans once per idea ( that is O ( ideas × workspace refs ) and freezes the editor on dense graphs ). Implemented by ["../../../../packages/language/src/reqlan-inbound-reference-inlay-label.ts"] and ["../../../../packages/language/src/reqlan-inlay-hint-provider.ts"]. [view_references_as_inlay_hints] [inbound_inlay_hints_whole_graph] | 0.122 | |
| non_web_reference_navigation | done | done | Non-web references ( idea brackets / wikilinks, local file paths, import paths ) must be Ctrl + click navigable via document links and / or go-to-definition. Built-in http ( s ) link detection is not enough. The rust engine builds same-file links and import-target links on the open path ( ["../../../language/syntax.rq".open_file_reference_sequencing], ["../support/open-file-sequencing.rq".outbound_one_hop] ). When the engine is node, idea underlines use the path-local Rust extract ( ["../../../indexer/indexer.rq".local_symbolic_analysis] ), and go-to-definition uses Langium when the idea ref is already linked. Qualified-reference idea lines follow ["../support/open-file-sequencing.rq".lazy_features_after_outbound]. Related: [reference_document_links], [reference_goto_definition_speed], [reference_underline_syntax_align], ["../../../language/syntax.rq".open_file_reference_sequencing]. | 0.163 | |
| reference_code_completion_objects | done | done | [reference_code_completion] must offer code completion for all valid reqlan references: - reqlan files ( `.rq` ) - ideas and ideasets in a reqlan file after a quoted path ( `["path".idea]` ) - ideasets and ideaset members ( `[ideaset.member]` ) - code files - folders, including folders that do not contain `.rq` files Import path completion stays limited to `.rq` files. Anonymous file-reference paths share ranking and alias logic with import paths, but they include every file and folder target. Scanning every workspace file on each keystroke is too expensive: [reference_code_completion_performance]. | 0.051 | |
| reference_code_completion_performance | done | done | [reference_code_completion_objects] must stay interactive on large workspaces. Path completion must not re-walk the filesystem on every keystroke: cache directory listings and reuse them for import and anonymous file-reference completion. Ranking must precompute proximity keys rather than recomputing hop distance per comparator call. Idea-graph ranking must reuse per-document adjacency across completion requests while document versions are unchanged. Anonymous file-reference completion may omit distant candidates after proximity ranking so the list stays bounded. Implemented by ["../../../../packages/language/src/reqlan-path-completion.ts"] ( `clearPathCompletionCaches`, listing cache, `sortPathCompletionCandidates` ), ["../../../../packages/language/src/reqlan-completion-provider.ts"], and ["../../../../packages/language/src/reqlan-completion-context.ts"]. [reference_code_completion] [reference_code_completion_objects] | 0.046 | |
| reference_document_links | done | done | Bracket and wikilink idea references — both same-file and cross-file — must emit a document link so the editor can follow them the same way as file-path references. Editor activation is [editor_reference_activation]. All resolved `LocalReference` and `QualifiedReference` idea / ideaset tokens should emit links targeting the declaration name, regardless of whether the target lives in the same file or an imported one. Namespace-import alias tokens ( `[alias]` → file ) keep their existing file-target link. Each link must include a `target` and a `tooltip` so VS Code does not show " Loading... ". Implemented by ["../../../../packages/language/src/reqlan-file-link-resolver.ts"] ( `resolveIdeaReferenceLinks`, `collectFileLinks` ) and ["../../../../packages/language/src/reqlan-document-link-provider.ts"]. | 0.124 | |
| reference_goto_definition_speed | done | done | Hover and go-to-definition are lookups on the open parse. The parse keeps its references sorted by start byte. A request finds the reference under the cursor by binary search. A request does not walk the syntax tree and does not convert each reference to a line and column. A request resolves each authored path once per parse. It does not `stat` that path again. A same-file idea uses that parse. An open neighbor uses that neighbor's buffer. A closed neighbor is parsed once. The server keeps its model, idea names, and name-token ranges. A later request does one `stat`. It reads and parses the file again only when the modified time or the size changed. A hover on a cross-file idea renders from that cached model. It does not read the file. The target is the name token of the idea, in the open editor when the target is open. An imported alias such as `[other]` reaches the imported idea. A folder reference has no definition. Its link is the folder-picker command. A hover renders idea content only when ["./features-syntax.rq".idea_content_tooltip] is on. A request does not wait behind the confirm pass of ["../support/open-file-sequencing.rq".lazy_features_after_outbound]. The editor calls this request on Ctrl + hover. The extension does not send the document text from that handler. The extension does not open a folder picker or a wildcard panel from that handler. A folder click and a wildcard click use the command document link. When the engine is node, the provider uses cached Langium ref objects for idea tokens. It does not scan the document text when the CST already covers the cursor. Related: [editor_reference_activation], ["../../../language/imports.rq".wildcard_references_webview]. | 0.179 | |
| reference_underline_syntax_align | done | done | Idea document-link and error underlines must cover the same UTF-16 range as the syntax token ( Langium `$refNode` / semantic token ), including when non-ASCII text appears earlier in the file. Native extract spans are UTF-8 byte offsets ( ["../../../indexer/indexer.rq".local_symbolic_analysis] ); LSP `Position.character` is UTF-16. Convert before `positionAt` so the underline does not drift off the token. Related: [non_web_reference_navigation], [reference_document_links], [unresolved_reference_diagnostics], [aliased_from_import_ctrl_click]. | 0.125 | |
| same_line_named_block_highlighting | done | done | TextMate-only highlighting ( no semantic tokens ) must follow the canonical lexer for named `{` / `}`: a top-level `name {}` or `name {hello}` is a complete ["../../../language/syntax.rq".block_idea] on that line. The grammar must close on the same line. It must not leave a block open until a later own-line `}`. An open block swallows later one-liners such as a name-only idea. `name {` at end of line is a multiline block. Close it only on an own-line `}` so matching prose pairs at end of line ( `{{SITE_URL}}` ) stay in the body per ["../../../language/syntax-edge-cases.rq".nested_curly_braces]. `name text {braces}` is a ["../../../language/syntax.rq".simple_idea]. `{` after other body text must not start a block. TextMate idea-name scopes on a file must match the Langium parse of that file. | 0.052 | |
| traceable_inlay_hints | done | done | A click on one inbound referencer opens that idea via `reqlan.openInboundReferencer` → host openIndexFile ( same path as activity-bar / index opens ). The inlay label part must not use raw LSP Location for the click gesture — Cursor rejects detached file: / / opens that Location would trigger. A click on the remainder opens the inbound summary for that idea. The backlink tooltip uses the shared document-link file URI builder ( encoded path, `#L` fragment ). The Rust server emits the same markdown tooltip and open command ( ["../support/language-server-rust.rq".language_server_rust] ). A file path with a space opens that file from the inlay click and from the tooltip. A directory segment " reqlan rq " opens that file. Implemented by ["../../../../packages/language/src/reqlan-inbound-reference-inlay-label.ts"], ["../../../../packages/language/src/reqlan-inlay-hint-provider.ts"], ["../../../../packages/language/src/reqlan-inlay-hint-settings.ts"], ["../../../../packages/extension/src/extension/register-reference-inlay-hints.ts"], ["../../../../packages/language/src/reqlan-file-link-resolver.ts"], ["../../../../packages/language/src/reqlan-path-resolve.ts"], and ["../../../../crates/reqlan-lsp/src/analysis.rs"]. | 0.052 | |
| unresolved_reference_diagnostics | in-progress | in-progress | Unresolved idea references must report a linking error diagnostic. A same-file miss and an import-target miss have an error and no document link. A qualified reference may link to its file before the idea line is confirmed. After confirmation, a missing idea has an error and no document link ( ["../support/open-file-sequencing.rq".missing_reference_colour_sequence] ). A false idea link looked like a happy reference while Ctrl + click could not navigate. Missing file references stay on ["./features-syntax.rq".file_references] / ["../support/language-server-errors.rq".file_reference_errors] ( error, no document link ). Related: [reference_rendering], [non_web_reference_navigation]. | 0.117 | |
| view_references_as_inlay_hints | done | done | References should be shown as inlay hints in the editor. This should be a toggleable setting, which should be available in the command palette. N-lines ( configurable setting ) of hints should be shown below the idea. Inbound lookup must stay interactive on large graphs via [inbound_inlay_index_performance]. The Rust server honours `reqlan.referencesInlayHints.enabled` and the host inbound snapshot ( `byIdeaName` + `fileReferencers` ). Implemented by ["../../../../packages/language/src/reqlan-inlay-hint-provider.ts"], ["../../../../packages/language/src/reqlan-inbound-reference-inlay-label.ts"], ["../../../../packages/language/src/reqlan-inlay-hint-settings.ts"], ["../../../../packages/extension/src/extension/register-reference-inlay-hints.ts"], and ["../../../../crates/reqlan-lsp/src/session.rs"]. | 0.093 | |
| wildcard_reference_highlighting | done | done | Bracket and wikilink forms of ["../../../language/imports.rq".wildcard_references] ( `["path/**/*.rq".idea_*]`, `[["path/*.rq".*_pane|label]]` ) must TextMate-scope as `meta.link.reqlan` with the path as `string.other.link.reqlan` — never as `comment.block.reqlan`. The grey " comment " paint happened when `#bracket-references` only allowed ID suffixes, so `/*` inside an unmatched path opened a block comment. Fixes: TextMate `#bracket-references` / wiki path patterns accept `WILDCARD_NAME` suffixes; `#comments` block begin rejects empty slash-star-star-slash; body rules try references before comments. Semantic tokens paint `WildcardReference.pathPattern` as string and `ideaPattern` as variable ( ["./features-syntax.rq".syntax_highlighting] ). Related: ["../../../language/syntax.rq".comments], ["../../../language/imports.rq".wildcard_references_webview]. | 0.079 | |
| idea_content_tooltip | done | done | A hover on an idea name or on a reference shows that idea's content. The tooltip is markdown. It uses a heading, body lines, attributes, lists, and code fences. A reference in another file shows that file's idea content. The setting `reqlan.ideaContentTooltip.enabled` controls this hover. The default is on. When the setting is off, the hover shows the idea name and the one-line summary. Related: [syntax_hovering] ["../support/language-server-rust.rq".language_server_rust] | 0.149 | |
| syntax_highlighting | done | done | . rq files should receive full syntax highlighting coverage for all reqlan constructs. Semantic token highlighting should distinguish ideas, attributes, references, imports, and keywords. Path + wildcard idea refs ( `["…/*.rq".idea_*]` ) must highlight as links / strings — not comment-grey — see ["./features-syntax-highlighting.rq".wildcard_reference_highlighting]. Import lines must colour path and keywords after `import` / `from` — see ["./features-syntax-highlighting.rq".import_statement_highlighting]. Same-line named braces ( `name {}`, `name {hello}` ) must close in TextMate so later one-liners stay visible — see ["./features-syntax-highlighting.rq".same_line_named_block_highlighting]. | 0.093 | |
| code_actions | done | done | Language-server code actions for `.rq` files, plus comment-file Quick Fixes for ignore-error: - Quick fixes for unresolved references ( add import, rewrite qualified ref, search, create file ) — ["./support/features-imports.rq".import_error] - Refactor: search / replace or wrap as idea reference — ["../host/features-commands.rq".search_code_actions] - Refactor: move or delete an idea declaration — ["../mutation/refactor_support.rq".refactor_symbol_move] ["../mutation/refactor_support.rq".refactor_symbol_delete] - File-level source action: barrel page — ["../host/features-commands.rq".barrel_page] - Quick Fix: insert `//rq-ignore-error` before a line with diagnostics — ["../host/features-commands.rq".code_actions_ignore_error] - Quick Fix: create a missing local idea below the idea that contains the unresolved reference — ["../host/features-commands.rq".create_idea_below_idea_containing_unresolved_reference_under_cursor] The Rust server returns the same set ( ["./support/language-server-rust.rq".language_server_rust] ). Command-backed search / replace and barrel actions omit LSP command arguments so VS Code does not cache disposable delegating ids; the extension resolves the active editor ( or document ) when the command runs. The same Quick Fix appears once in one request. Langium appends validation diagnostics when a document already has diagnostics. The diagnostic list stays when the document state is below Validated. The document builder removes previous built-in diagnostics before that append. One linking error gives one Quick Fix after repeated builds. | 0.169 | |
| context_content_styling | done | done | Pane bodies must not grow unbounded. Nested lists use [NestedSection] ( collapse + scroll ) under [CollapsiblePane]. Priority: [scope_pane]; [reference_lists_pane] kind groups; also selection / parents. [ScopePane] [ReferenceListsPane] | 0.046 | |
| dependency_pulse_table | done | done | The focus-hero [DependencyPulse] compass ( ↑ parents, → outbound, ← inbound, ↓ dependents, ● centre ) is clickable. Clicking any arm or the centre toggles a compact reference table below with Dir, Role, Idea, and Status. Rows open / focus the related idea. Duplicates of always-on inbound / outbound lists are omitted from the hero; the current-file lens still lists them. Aligns with [context_scope_v2.dependency_pulse]. Implemented by [DependencyPulse] [ScopePane]. | 0.059 | |
| fingerprint_axis_drilldown | done | done | Each [ContextFingerprint] axis row ( Files, Requirements, History, Architecture, Git, Diagnostics, Coverage ) is clickable. Clicking a row toggles a compact contributor table below it ( Source, Item, Detail ) listing what filled that axis — e. g. which lenses included a file, which ideas are in the footprint, git focus commits / authors ( dirty paths secondary ), anomalies, or relationship edges. Table rows open / focus the related idea or file when actionable. Empty axes show a short “ why care / how to fill ” message. Mirrors the interaction pattern of [dependency_pulse_table]. Aligns with [context_scope_v2.context_fingerprint] and [tooltip_detail]. Implemented by [ContextFingerprint] [ScopePane]. | 0.075 | |
| git_history_lens | done | done | Expanded git dimension in [scope_pane]: focus-centric development history, not a Source Control duplicate. Layout: - Header: branch name, HEAD short hash, muted summary ( e. g. " main · 3 commits · 2 authors " ). - Focus history ( primary ): up to ~ 8 commits touching the focus file / idea line range — relative time, short hash, subject, author. Prefer `git log -L` for `.rq` ideas; fall back to path log. Non-. rq uses path `--follow`. - Authors: compact chip row of top authors from that log window. - Working tree ( secondary, collapsed by default ): dirty staged / unstaged paths for caution / footprint only. Focus hero: history cue ( last commit age / branch ), not `git: staged|unstaged`. Fingerprint Git axis and drilldown list commits / authors ( dirty paths secondary ) per [fingerprint_axis_drilldown] and [tooltip_detail]. Aligns with ["../context-scope.rq".git_context] and [context_scope_v2.development_history_signals]. Performance: heavy focus history must not block link open or first context paint — see ["../activitybar.rq".link_open_decoupled_from_git] and ["../activitybar.rq".git_focus_cache_by_head]. Implemented by [ScopePane] [ContextFingerprint]. | 0.107 | |
| reference_resolution_parity | done | done | Reference resolution for the activity-bar references pane must use the same shared rules as the editor ( document links / go-to-definition ), not a second ad-hoc classifier. - Bare namespace-alias bracket refs such as `[ActivityBarApp]` that open a file in the editor must index as resolved `file_reference` edges ( import path ), not unresolved idea refs — per ["../../../language/imports.rq".import_namespace]. - Quoted file paths and markdown file links use the same file-target treatment as the language file-link resolver in ["../../../../packages/language/src/reqlan-file-link-resolver.ts"] / ["../../../../packages/language/src/reqlan-namespace-import-links.ts"]. - File targets are opened relative to the defining `.rq` file ( same path join as the miniature graph ), not the workspace root. - If a target resolves in the file view, it must resolve ( and be openable ) in [reference_lists_pane]. Implemented by ["../../../../crates/reqlan-index/src/extract.rs"], ["../../../../packages/analytical/src/core/file-reference-resolve.ts"], [ReferenceListsPane]. | 0.064 | |
| tooltip_detail | done | done | Scope pane tooltips are detailed enough to explain *what* a meter means and *why it matters*. - [scope_pane] widgets ( stability, dependency pulse, timeline, churn, lens chips, footprint line ) expose granular hover copy on the control and its sub-items. - [ContextFingerprint] has an ⓘ info button with semantic scoring help, per-axis hover copy, and [fingerprint_axis_drilldown]. - Scoring copy is sourced from [context_signals] ( `CONTEXT_FINGERPRINT_HELP`, `CONTEXT_FINGERPRINT_AXIS_HELP`, `buildContextFingerprint` ) so UI stays aligned with [context_scope_v2.context_fingerprint] and [context_scope_v2.ai_readiness_gauge]. Implemented by [ContextFingerprint] [ScopePane]. | 0.075 | |
| unresolved_reference_create | done | done | When an outbound reference remains unresolved after shared resolution ( missing idea, not a file / namespace-alias target ): - Do **not** treat the label as a file path or navigate to a non-existent file. - Show an unresolved badge and a ** + ** control with tooltip " Create idea ". - Activating ** + ** opens create-idea ( reuse [commands.pallete_function_create_todo] when available; until then, the existing import-error create-file flow prefilled with the reference name and source document ). Implemented by [ReferenceListsPane] and the activity-bar host message handler. | 0.064 | |
| minigraph_file_reference_links | done | done | when the user clicks on a file reference link in the miniature graph, we should open the file in the editor if the file is defined as relative, it should be opened relative to the defining file, not the workspace root Fixed in [indexer.local_graph_analysis]: external file nodes now resolve their path against the defining file's folder ( via the edge source id ) when the slice is built, so a workspace-relative path is handed to the editor open action. | 0.045 | |
| search_add_to_chat | done | done | From each [search_pane] hit, a split **Add to chat** control: - Primary button: place compact `#requirement` context into the **current** chat input without submitting. - ▾ menu ( visible secondary affordance ): **Add to current chat** or **Add to new chat** ( new session, then seed the same payload ). Tooltips name both behaviours; secondary is not right-click-only. Same payload shape as ["../../agent/features-ai.rq".add_to_chat] ( idea name, path: line, summary ) via [open_chat]. Host: [open_chat] with `{ query, isPartialQuery: true, target: 'current' | 'new' }`. Cursor note: do not use `workbench.action.chat.open` for **current** ( it always opens a new tab ); see ["../../agent/features-ai.rq".add_to_chat] / [open_chat]. Clipboard fallback when no chat command is available ( same spirit as ["../../agent/features-ai.rq".add_to_chat] ). Wire: [AddToChatButton] / [SearchPane] → [app_state] `addToChat(hit, target)` → [activity_bar_messages] → [activity_bar_webview] → [open_chat]. | 0.071 | |
| search_pane | done | done | Workspace idea search in the activity bar — find rq ideas by name, summary, or tags without leaving the sidebar. Complements the modal from [commands.search_code_actions]; this pane stays open for repeated lookups and inserts. Implemented by [SearchPane] under [CollapsiblePane]; host wire in [activity_bar_webview] / [activity_bar_messages] / [app_state]. [search_pane_query] [search_pane_layout] [search_pane_results] [search_pane_open_live_file] [search_pane_load_more] [search_pane_file_hits] [search_pane_insert_reference] [search_add_to_chat] [search_pane_advanced] [search_pane_empty_query] [search_pane_scoped_status] [search_progress] [search_pane_seed] [search_pane_match_highlighting] | 0.176 | |
| search_pane_advanced | done | done | [search_pane] header link **Advanced** opens [ideas_summary_doc.ideas_list] with the current query prefilled. | 0.063 | |
| search_pane_empty_query | done | done | Empty query in [search_pane]: prompt to type ideas or files ( do not dump the full catalog into the pane ). | 0.063 | |
| search_pane_file_hits | done | done | [search_pane] mixes [core_search.file_search] hits ( kind `file` ) into the ranked list with ideas. Click opens the file at line 0. File hits do not offer Insert ref or Pin ( those are idea actions ). Empty-state and placeholder copy mention files. | 0.115 | |
| search_pane_insert_reference | done | done | Each [search_pane] hit has an **Insert reference** button that inserts `[idea_name]` at the active editor cursor ( or replaces the selection ) via [insert_idea_reference], adding an import when the target is in another file — same rules as [commands.search_code_actions]. Helpers: [insert_idea_reference_helpers]. | 0.063 | |
| search_pane_layout | done | done | [search_pane] results fill the pane body ( 100 % width ) and scroll inside the pane; do not nest a second capped [NestedSection] scroll. | 0.069 | |
| search_pane_load_more | done | done | When ranked hits exceed the visible page, [search_pane] shows an exceeded-count label ( `40+ matches`, or `{n}+ matches` after further pages ) instead of a full total, and a **Show more** control at the bottom of the list that fetches the next page and appends. A new query resets to the first page. Uses [core_search.fuzzy_search_pages] ( `offset` + `truncated` ). | 0.095 | |
| search_pane_match_highlighting | done | done | The [search_pane_results] list highlights the parts of name, path, and summary that matched the query via [SearchHighlight]. Resilient to [core_search.fuzzy_search] / [core_search.fuzzy_search_whitespace]: contiguous and separator-insensitive spans, token prefixes / part matches, reordered words, and character-subsequence ( fuzzy / acronym ) marks on short fields such as the idea name — not scattered letters in long summaries. | 0.063 | |
| search_pane_open_live_file | done | done | A click on a [search_pane] hit must show the live workspace file at that path. If that file is already open in a workspace tab, focus that tab and move the caret. Do not open a second read-only tab with stale content ( a detached `file://` copy that Cursor does not sync ). Keep the click path cheap: no extra index read, no base discovery. | 0.074 | |
| search_pane_query | done | done | [search_pane] search input posts a debounced host query. Ranking uses [core_search.fuzzy_search] in the native search module ( `NativeWorkspaceIndex.fuzzySearch` ); separator-insensitive `_` / `-` / `...` / whitespace. File-name hits: [search_pane_file_hits] / [core_search.file_search]. Pages: [search_pane_load_more] / [core_search.fuzzy_search_pages]. Highlighting stays in the TS `./fuzzy-search` export. | 0.063 | |
| search_pane_results | done | done | [search_pane] results list is paged ( default page size 40 ) with name, path, and one-line summary; click opens the idea or file ( [search_pane_open_live_file] ). Exceeded counts and further pages: [search_pane_load_more]. File hits: [search_pane_file_hits]. | 0.078 | |
| search_pane_scoped_status | done | done | [search_pane] soft-waits when the index is not ready; scoped errors stay in the pane. Loading UX: [search_progress]. | 0.063 | |
| search_pane_seed | done | done | Host may seed [search_pane] via `focusIdeaSearch` ( query + optional pathFilter ) from the wildcard-matches webview **Open in Search** action ( ["../../../language/imports.rq".wildcard_references_webview] ). Path filter is a hard AND after ranking ( ["../../../language/imports.rq".idea_path_filter] ). | 0.090 | |
| search_progress | done | done | Activity-bar search ranks in the native [core_search.fuzzy_search] module ( rusqlite read + score in-process ). The host must not hydrate a JS idea catalog or worker before scoring — return only the capped hit list. A static " Searching … " / " Loading … " label must not be the only signal: the pane should show **live progress** so a slow wait reads as working, not hung. - Host posts progress for the in-flight search request while native scoring runs ( `Scoring ideas…` ). - [SearchPane] displays the phase label with a continuous activity affordance ( spinner / indeterminate bar ) and elapsed time while loading, for both the empty initial wait and the " Updating … " refresh path. - Stale progress from superseded queries must not overwrite a newer request; clear progress when results or a scoped error arrive. | 0.080 | |
| todo_pane | done | done | Workspace todo list in the activity bar — ideas that carry the `@todo` special attribute ( [ai_support.special_attributes] ), not ideas whose `@status` happens to be `todo`. Complements the planned Overview todos rollup in [ideas_summary_doc.overview_page]; this pane stays open for quick triage beside the editor. Layout matches [search_panel.search_pane_layout]: flat list in the pane body, not a nested boxed subsection. - List ideas in the active base where the `todo` attribute is present ( bare `@todo` / flag, or `@todo` with a note ). - Each row: idea name, source path, and the todo note when present ( omit empty / bare-true values ). - Click opens the idea in the editor ( same navigation as [search_panel.search_pane_results] ). - Results fill the pane body ( 100 % width ) and scroll inside the pane; do not nest a second capped NestedSection scroll. - List capped ( default ~ 40 ) with a “ showing N of M ” meta line when truncated; do not dump unbounded index results into the pane. - Soft-wait when the index is not ready; scoped errors stay in the pane. - Pane header **Refresh** reloads the list; also reload on index health / sync completion for the active base. - Optional affordance to create a stub via [commands.pallete_function_create_todo] when that command ships — do not block the list on it. Host query + wire: [activity_bar_webview] / [activity_bar_data] / [activity_bar_messages] / [app_state]; store query [sqlite_store]. Implemented by [TodoPane] under [CollapsiblePane]; same scroll / fill pattern as [SearchPane]. | 0.046 | |
| workspace_base_picker | done | done | Bases in the Workspace pane are chosen via a **searchable dropdown**, not a stacked list of base rows. - Closed: shows the active base label ( and compact ready / issue hint when useful ). - Open: type-to-filter over discovered bases by label and path; selecting a match **pins** the active base ( [bases.active_base_switch] ) — a pointer swap, not a sync. - Options surface label / path plus ready state and idea / edge / issue counts so the user can distinguish bases before committing. - Single-base workspaces still use the same control ( no special flat list ); empty discovery stays on the Create Base CTA. - Prefer VS Code dropdown tokens for chrome; keep the control compact so health / stats / actions remain the focus of the pane. Switch UX: - Selecting a base posts immediately; the picker may flash **Switching … ** until `indexHealth` confirms `activeBaseId` ( should be near-instant ). - Never kick soft / hard sync on switch; readiness / progress for a not-yet-ready base is whatever the index already reports. - Editor-follow must not overwrite a user-selected base on the same refresh that applies the selection. Implemented by [BasePicker] in [WorkspacePane]; filter helpers in [filter_bases]. | 0.072 | |
| workspace_content_styling | done | done | Pane bodies must not grow unbounded. Nested lists use [NestedSection] ( collapse + scroll ) under [CollapsiblePane]. File → errors grouping via [group_file_issues] in [WorkspacePane]. | 0.046 | |
| activity_bar_styling | done | done | Activity bar chrome for the [panes] stack under [ActivityBarApp] / [CollapsiblePane]. - **Open panes fill** the remaining bar height ( flex grow ); collapsed panes stay header-sized. A single open pane takes the full stack; multiple open panes share space by persisted relative weights. - Each expanded pane is **resizable** when more than one is open ( drag handle adjusts weight ); heights / weights persist with expand state via [ActivityBarApp] view state. - The bar itself **scrolls** when minimum pane sizes exceed the webview viewport; otherwise pane bodies scroll internally. - Pane shells **persist** across bootstrap / empty / invalid states — hide or soften **content** inside the pane ( e. g. [SelectionPane] empty copy ), do not unmount the section for phase or missing data. Bootstrap banner remains per [loading_state]; list overflow still uses [panel_content_styling] / [NestedSection]. Layout helpers: ["../../../packages/extension/webviews/activity-bar/lib/pane-layout.ts"]. | 0.054 | |
| error_status_bar | done | done | When the activity bar status line ( ["../../../packages/extension/webviews/activity-bar/components/HeaderBar.svelte"] ) or Ideas Summary status chrome ( ["../../../packages/extension/webviews/ideas-summary/components/StatusBar.svelte"] ) is in an error state ( `statusError` ), surface the error **prominently** — do not mute it or bury it behind healthy index counts. - Lead the status text with the error summary or issue count; healthy idea / reference counts may follow as secondary context. - Use error foreground, error validation background, a distinct error label, and `role="alert"` so the message is visually and accessibly distinct from idle status. - Bootstrap error banner ( ["../../../packages/extension/webviews/activity-bar/components/BootstrapStatus.svelte"] in [loading_state] ) must not apply muted styling to the error detail. Soft-waits for an unready index stay non-error per [loading_state] / [workspace_panel.workspace_loading]; this idea covers true error states only. Shared copy helper: ["../../../packages/extension/webviews/ideas-summary/lib/index-status-text.ts"]. | 0.075 | |
| git_focus_cache_by_head | done | done | Focus git history / stats caches are not TTL-evicted on a fixed wall-clock window. Refresh a cache entry only when its write timestamp is **before** the repository's latest commit authored time ( HEAD ); while `cache.at >= latestCommitAuthoredAt`, reuse the entry. Applies to focus history, focus stats, and range-commit count caches in ["../../../packages/extension/src/activity_bar_module/git-context.ts"]. Related UX: [current_context_scope.git_history_lens]. | 0.059 | |
| lean_resolved_open | done | done | When a link already carries a resolved `fileUri` / `targetPath`, open that URI at the given line — do not re-classify resolution or walk base discovery on the click path. Pass the active base root into ["../../../packages/extension/src/analytical_submodule/index-store/open-index-file.ts"] so ["../../../packages/extension/src/analytical_submodule/index-store/resolve-index-file-uri.ts"] skips tree discovery. `toIndexFileUri` fallbacks ( editor selection, git chrome, open-tab scan ) must use ["../../../packages/analytical/src/core/base-discovery.ts".nearestBaseRoot] ( ancestor `.reqlan` walk ), never `discoverBases` ( full workspace scan ). Clicking a search hit that already has `fileUri` + line must not wait on catalog scoring or base rediscovery. | 0.045 | |
| link_open_decoupled_from_git | done | done | Opening a resolved idea / file link must not await git focus history or idea-at-revision parsing. Host posts index-backed context ( scope, references, graph, ancestors ) first; [current_context_scope.git_history_lens] enrichment runs asynchronously afterward and may re-post context when ready. Light git chrome ( branch / dirty from the VS Code git API ) may ride along with the first paint; `git log -L`, `git show` + Langium parse-at-revision, and focus stats must not block open or the first context post. Implemented by ["../../../packages/extension/src/activity_bar_module/activity-bar-webview-provider.ts"] and ["../../../packages/extension/src/activity_bar_module/git-context.ts"]. | 0.059 | |
| link_open_interactive_quiet | done | done | Programmatic and editor file-switch navigation share the interactive quiet path with webview `openIdea`: cancel in-flight search / refresh and briefly suppress editor-driven pane rebuilds so the open settles before context rebuild. Selection moves within the same document keep the short debounce only ( no full interactive cancel ). Document-link / go-to-definition opens that change the active file are covered by active-editor-switch quieting. Implemented by ["../../../packages/extension/src/activity_bar_module/activity-bar-webview-provider.ts"]. | 0.045 | |
| loading_state | done | done | After the activity bar shell mounts, the header may appear before pane data. Cover every sequencing beat with an explicit loading or error affordance — never a blank content area: 1. HTML shell ( before Svelte boots ): inline waiting markup in [get_activity_bar_html]. 2. Connecting: header visible; a post-first-frame `ready` handshake starts the index and requests the first host state. 3. Waiting for index: [workspace_panel.workspace_loading]. 4. Ready panes: [current_context_scope.context_panes_loading], [minigraph.minigraph_loading], and [search_panel.search_progress]. Index sync progress belongs in [workspace_panel.workspace_pane]; graph slice waits stay in [miniature_graph_pane]; idea-search phases stay in [search_panel.search_pane] — not a second global overlay. Host posts index health on `ready` and on subsequent status changes; soft-waits when the index is not ready instead of erroring the whole bar. Pane shells stay mounted through these phases per [activity_bar_styling]; only content / bootstrap banner reflect invalid or loading states. True error states ( status line + bootstrap banner ) must show the error prominently per [error_status_bar]. | 0.063 | |
| minigraph_file_reference_links | done | done | File-ref click open behaviour — details in [minigraph.minigraph_file_reference_links]. | 0.040 | |
| panel_content_styling | done | done | Pane bodies must not grow unbounded. Nested lists use [NestedSection] ( collapse + scroll ) under [CollapsiblePane]. Shell resize / fill / scroll behaviour: [activity_bar_styling]. Details: [workspace_panel.workspace_content_styling], [current_context_scope.context_content_styling]. [WorkspacePane] [ScopePane] [ReferenceListsPane] | 0.047 | |
| performance | done | done | Sidebar webviews share the same constraints as [graph_lib.webview_threading]. - The HTML / Svelte shell paints before index discovery begins; the post-paint `ready` handshake starts indexing. - Require the host-matching core engine at activate; do not embed a sql. js vendor in the activation-critical host bundle. - Default neighbourhood slices; never load the full workspace graph into a pane. - Debounce editor selection and cursor events before rebuilding slices. - Lazy-mount cytoscape only for visible sub-panes; destroy when collapsed or hidden. - Reuse [indexer.local_graph_analysis] and graph query builders rather than ad-hoc index walks in the webview. - Interactive actions ( `openIdea`, insert ref, add to chat, phonebook / Ideas Summary links ) are handled on a priority path: they cancel in-flight search / refresh work and briefly quiet editor-driven rebuilds so subsequent clicks stay snappy. - Idea search scoring runs in a worker thread with a preloaded catalog; catalog load yields so it cannot monopolize the host. - Link navigation must stay snappy once a target is already resolved: [link_open_decoupled_from_git], [link_open_interactive_quiet], [lean_resolved_open], [git_focus_cache_by_head]. | 0.040 | |
| activity_constellation | incomplete | incomplete | widget, activity, graph | Spatial glow of nodes by activity ( GitHub-contribution feel, but layout- aware ). Shows hotspots without a separate legend dump. Complements [hotspot_overlay]; prefer one encoding per surface to avoid competing visual channels. | 0.054 |
| activity_river | incomplete | incomplete | widget, activity | Workspace-top ribbon of recent activity density ( yesterday / today bursts ). Shows momentum, not a file list. Consumes [project_activity_signals] and [user_signals]. | 0.057 |
| ai_readiness_gauge | incomplete | incomplete | widget, ai | Ready-to-assist gauge: requirements ✓, implementation ✓, history ✓, tests ✕, risk low — plus readiness %. Tells the user whether AI has enough synthesized context before [features_ai.ai_add_to_context] / export. Pairs with [context_fingerprint] ( what is in context ) vs readiness ( is it enough / safe ). | 0.062 |
| architectural_map | incomplete | incomplete | widget, architecture | Tiny layered minimap ( UI / Services / Core / Infra … ) with current focus highlighted and bar width for effort concentration. Consumes [architecture_signals] and [project_activity_signals]. Links out to Ideas Summary graph rather than duplicating full layout. | 0.052 |
| base_signals | in-progress | in-progress | bases, multi-base | Multi-base facts for [context_signals] / [synthesized_context]: active_base_id, active_base_root, active_base_ready, idea / edge counts for the active base, sibling_bases ( id, path, ready, counts ), discovery_empty ( prompt create-base ). Consumed by ["./activitybar-panels/workspace.rq".workspace_pane], Ideas Summary base switcher, and [project_weather]. | 0.056 |
| change_velocity | incomplete | incomplete | widget, velocity | Slow ↔ Fast bar from edits / week ( or equivalent ). Complements [churn_heat_bar] ( shape over time ) with a single magnitude readout. | 0.046 |
| churn_heat_bar | incomplete | incomplete | widget, churn | Compact sparkline of edit intensity over time for a requirement, file, folder, or subsystem ( e. g. ▁ ▁ ▂ ▁ ▁ ▇ █ ▆ ▂ ▁ ▁ ). Hover reveals synthesized story ( stable / large rewrite / quiet since ). Consumes [development_history_signals] and [project_activity_signals]. | 0.066 |
| concept_density | incomplete | incomplete | widget, semantic | Per-folder or per-subsystem bars of conceptual richness ( Authentication █ █ █ █ █ █ █ █ █ █ vs Payments █ █ █ ). Consumes [semantic_signals]; useful in Ideas Summary folders / index panel. | 0.046 |
| confidence_card | incomplete | incomplete | widget, confidence, ai | Pre-edit checklist card: High / Medium / Low confidence with check / warn rows ( stable, reviewed, tested, high fanout, … ) and a confidence %. Shown before AI modify workflows ( [features_ai.ai_integration] ) and on focus when synthesis is available. | 0.054 |
| context_consumers | incomplete | incomplete | Shared semantic representation for: UI widgets, graph traversal, AI agents, search, diagnostics, navigation-by-intent. New analyses register as signal producers; consumers bind to signals / synthesis, not to ad-hoc dimension lists. | 0.046 | |
| context_fingerprint | incomplete | incomplete | widget, context, ai | Compact bars of what the current AI / composed context actually contains: Files, Requirements, History, Architecture, Git, Diagnostics, Coverage. Lets the user see expansion when LLM reports " context expanded " ( [context_scope_v1.ai_context_export], [features_ai.ai_integration] ). Hover / info / drill-down affordances: ["./activitybar-panels/current_context_scope.rq".tooltip_detail], ["./activitybar-panels/current_context_scope.rq".fingerprint_axis_drilldown]. | 0.080 |
| context_model_layers | done | done | Canonical v2 layers. Implementation builds on [context_scope_v1.reqlan_context_model]; consumers must not reconstruct project understanding independently. | 0.045 | |
| context_radar | incomplete | incomplete | widget, context | Polar / radar chart of context richness axes ( Tests, Docs, Dependencies, History, AI, … ). Filled polygon shows what the current [context_scope_v1.context_footprint] actually covers. | 0.046 |
| context_scope_v2 | in-progress | in-progress | context, ui, knowledge, v2 | Evolves [context_scope_v1.context_scope] from a **source-oriented** model ( " where did this context come from? " ) to a **decision-oriented** model ( " what information helps a human or AI make the next decision? " ). v1 [context_scope_v1.context_dimensions] remain valid *producers* ( workspace, current file, open files, file history, edit history, manual, git ). v2 separates those producers from extracted signals, synthesized knowledge, presentation views, and consumers — so new analyses can be added without inventing new dimensions each time. Primary UI principle: [ui.show_dont_tell] — make project health, momentum, relationships, and intent legible at a glance. Host fact chips ( " 17 modified files " ) remain useful; widgets below are the preferred default for conveying meaning. Surfaces: extend [activitybar.scope_pane] / [context_scope_v1.scope_pane_design], Ideas Summary ( [ideas_summary_doc.ideas_summary] ), and graph overlays ( [graphical_graph_doc.graphical_graph] ) — same signals, different projections. | 0.089 |
| context_signals | done | done | Extracted facts. Answer: " What useful information exists? " One source may emit many signals ( e. g. git → development history, not only staged / unstaged ). Families: - [requirement_signals] - [file_signals] - [development_history_signals] - [project_activity_signals] - [semantic_signals] - [quality_signals] - [user_signals] - [relationship_signals] - [temporal_signals] - [risk_signals] - [architecture_signals] - [requirement_lifecycle_signals] - [base_signals] Signals are evidence; [synthesized_context] merges them into actionable knowledge. | 0.060 | |
| context_sources | incomplete | incomplete | Stable raw producers. Answer: " Where are we collecting information from? " Map from v1: [context_scope_v1.workspace_context], [context_scope_v1.current_file_context], [context_scope_v1.open_files_context], [context_scope_v1.file_history_context], [context_scope_v1.edit_history_context], [context_scope_v1.manual_context], [context_scope_v1.git_context] become sources ( and contributions ), not the whole model. `workspace` = VS Code workspace chrome / ambient multi-base glance ( folder roots, discovery empty state ). `base` = the active ontology [ontology_base] ( `.reqlan` - marked container ): its index, readiness, and path ( [bases.implied_context] ). Requirements graph and project index bind to the active base's store, not a merged workspace graph. | 0.055 | |
| context_views | incomplete | incomplete | Projections over the same signals — not separate data models. Timeline · Graph · File · Requirement · Subsystem · Concept · Implementation · History · Risk · AI Summary. | 0.046 | |
| dependency_pulse | incomplete | incomplete | widget, dependencies | Spatial dependency summary instead of " Parents: 4 / Children: 12 ". Compact compass ( ↑ parents, → children, ← siblings, ↓ dependents ) and / or a filled **Dependency Load** bar showing centrality growth. Consumes [requirement_signals] and [relationship_signals]. Surfaced on focus hero and graph node details. Click any arm or the centre toggles a reference table below the compass ( Dir / Role / Idea / Status ) per ["./activitybar-panels/current_context_scope.rq".dependency_pulse_table]. | 0.068 |
| historical_replay_slider | incomplete | incomplete | widget, timeline, graph | Time slider across the graph ( e. g. 2023 → Now ). Nodes and edges appear / disappear; dependencies animate. Extremely powerful for architectural storytelling; performance-capped neighbourhoods only in the activity bar; fuller replay in Ideas Summary. Consumes [temporal_signals], [requirement_lifecycle_signals], [relationship_signals]. | 0.052 |
| hotspot_overlay | incomplete | incomplete | widget, graph, hotspot | Graph-node encoding of churn / risk / discussion / instability ( size rings ○ ◎ ⬤ or green → red ). Same signal set as [activity_constellation]. Applies to [activitybar.miniature_graph_pane] and [graphical_graph_doc.graphical_graph]. | 0.059 |
| impact_radius | incomplete | incomplete | widget, impact, graph | On select / focus, fade in affected neighbours; opacity ∝ downstream impact. Consumes [relationship_signals] and [risk_signals]. Works in miniature graph and Ideas Summary graph. | 0.052 |
| implementation_priority_v2 | in-progress | in-progress | After v1 [context_scope_v1.implementation_priority] is solid: 1. [context_model_layers] types: sources → signals → synthesis contracts ( extend analytical context model; keep v1 dimensions as sources ). ✓ 2. Seed signal producers: [development_history_signals], [requirement_lifecycle_signals], [relationship_signals], [risk_signals] from index + git ( reimagined, not staged / unstaged-only ). ✓ ( thin ) 3. [synthesized_context] rollups for focus entity ( stability, confidence, AI risk, coverage ). ✓ 4. First widgets on focus hero: [stability_meter], [dependency_pulse], [timeline_ribbon], [churn_heat_bar]. ✓ 5. [requirement_card] in search / Ideas Summary lists; [context_fingerprint] + [ai_readiness_gauge] on AI context export. ✓ 6. Graph overlays: [hotspot_overlay], [impact_radius]. ✓ 7. Workspace glance: [project_weather], [activity_river], [architectural_map]. deferred 8. Deeper storytelling: [historical_replay_slider], [semantic_diff_viewer], [relationship_cloud], [traceability_matrix]. deferred | 0.040 | |
| knowledge_coverage | incomplete | incomplete | widget, coverage | Stacked coverage bars: Requirements / Implementation / Tests / Docs — prefer filled bars over bare percentages. Consumes [quality_signals] and [requirement_signals]. Pair with [traceability_matrix]. | 0.046 |
| ownership_distribution | incomplete | incomplete | widget, ownership | Horizontal author share bars ( bus-factor at a glance ). Consumes [development_history_signals] ownership and [risk_signals] ( e. g. single_contributor ). | 0.046 |
| project_weather | incomplete | incomplete | widget, dashboard, health | Synthesized workspace health widget — the engineering " dashboard " glance: Stable / Risk / Momentum / Knowledge / Recent Change, or compact meters for Project Health, Momentum, Risk, Confidence. Consumes [synthesized_context] across [risk_signals], [project_activity_signals], [quality_signals], and [base_signals] ( per-base readiness when multiple bases exist ). Belongs in activity bar workspace lens / Ideas Summary index panel — not as competing hero noise on every focus. | 0.065 |
| relationship_cloud | incomplete | incomplete | widget, semantic, relationships | Compact spatial alternative to a full graph: related concepts laid out by semantic closeness ( distance encodes similarity ). Click focuses; " Open in Graph " hands off to [graphical_graph_doc.graphical_graph]. Consumes [semantic_signals] and [relationship_signals]. | 0.052 |
| requirement_card | incomplete | incomplete | widget, card | GitHub-issue-style compact card for search results and lists: name, stability chip, meter %, last edited, ref / impl / test counts, urgency flag ( e. g. recently changing ). Composes [stability_meter], [churn_heat_bar], [dependency_pulse] as optional rows — not all at once by default ( token / UI density ). | 0.064 |
| requirement_lifecycle_wheel | incomplete | incomplete | widget, lifecycle | Four-quadrant wheel: Specify · Implement · Test · Review. Segments fill as evidence accumulates from [requirement_lifecycle_signals] and [quality_signals]. Compact lifecycle for engineering projects. | 0.046 |
| semantic_diff_viewer | incomplete | incomplete | widget, diff, semantic | Conceptual change summary instead of ( or above ) line diffs: Added / Changed / Removed ideas or policies, plus impact count ( e. g. 17 requirements ). Higher signal for AI and review than textual diffs alone. Consumes [semantic_signals], [relationship_signals], [development_history_signals]. | 0.052 |
| show_dont_tell | in-progress | in-progress | ui, widgets, show-dont-tell | Make project state **legible at a glance**. Prefer visual evidence of health, momentum, relationships, and intent over numeric fact lists. Widgets below are show-don ' t-tell components for [context_scope_v2]; each consumes [context_signals] / [synthesized_context] and may appear in activity bar, Ideas Summary, focus heroes, search results, or graph overlays. Placement guidance: - Focus / card density: [timeline_ribbon], [churn_heat_bar], [dependency_pulse], [stability_meter], [change_velocity], [requirement_lifecycle_wheel], [requirement_card], [confidence_card], [ai_readiness_gauge] - Workspace chrome / dashboard denser: [activity_river], [project_weather], [context_fingerprint], [architectural_map] - Graph / Ideas Summary: [hotspot_overlay], [impact_radius], [activity_constellation], [historical_replay_slider], [relationship_cloud] - Coverage / ownership / density: [knowledge_coverage], [traceability_matrix], [ownership_distribution], [concept_density], [context_radar], [semantic_diff_viewer] | 0.051 |
| stability_meter | incomplete | incomplete | widget, stability | Segmented meter + percent ( e. g. Stable █ █ █ █ █ █ █ █ █ ░ 91 % ) derived from edit frequency, age, dependency churn, and issue count — not raw git status. AI and humans use it as a confidence cue for change. Pair with [confidence_card] and [risk_signals]. | 0.060 |
| synthesized_context | done | done | Signals merged into decision-oriented knowledge not tied to a single source. Example projection on a requirement: - Importance, Confidence, Stability, Recent Activity - Dependencies, Implementations, Tests, Documentation - AI Risk Synthesis powers [show_dont_tell] widgets and [features_ai.ai_integration] exports — preference for compact synthesized facts over raw dimension dumps ( [context_scope_v1.ai_context_export] ). | 0.082 | |
| timeline_ribbon | incomplete | incomplete | widget, timeline | Horizontal lifecycle ribbon for a requirement ( or focused entity ), not a single " last modified " label. Show milestones along a line toward Now — e. g. created, first implementation, last reviewed, current edit, last test — so maturity and recent activity are visible in one glance. Consumes [requirement_lifecycle_signals] and [temporal_signals]. Fits [context_scope_v1.scope_pane_design]. focus_hero and [requirement_card]. | 0.052 |
| traceability_matrix | incomplete | incomplete | widget, traceability | Compact Req × Impl × Test ( and optional Docs ) check matrix for a local neighbourhood or subsystem. Exposes gaps immediately. Neighbourhood-scoped only — full-workspace matrices stay in Ideas Summary with pagination / filters per [ideas_summary_doc.ideas_summary]. | 0.058 |
| widget_composition_rules | incomplete | incomplete | Show-don ' t-tell without dashboard clutter. | 0.040 | |
| comment_reference_refactor_support | done | done | File and symbol refactors must update [comment_reference] tokens in code files and in `.rq` comments. When a file or folder moves or is renamed, rewrite `rq:["path".idea]` paths relative to the referencing file: outbound paths inside the moved file, and inbound paths in other files that pointed at it. Aliased paths such as `rq:["@/old/file.rq".idea]` keep the alias and update the remainder. When an idea is renamed, update matching comment idea tokens. When an idea moves to another `.rq` file, rewrite qualified comment paths that targeted the old file for that idea. Pathless `rq:[idea]` tokens stay as name-only links. When an idea is deleted, remove comment references that targeted it. Relates to [refactor_file_moves], [refactor_symbol_rename], [refactor_symbol_move], [refactor_symbol_delete], [refactor_changes], ["./mutation-hooks.rq".move_file], ["./mutation-hooks.rq".rename_file], ["../language/comment-references/functional-code-comment-references.rq".references_in_functional_code_comments]. | 0.215 | |
| refactor_symbol_move | done | done | An idea can move to another existing `.rq` file from a code action or the [command_palette]. Palette commands use the idea under the cursor, or a pick list when the cursor is not on an idea. 1. Move idea ( `reqlan.refactor.moveIdea` ) - Cut the declaration from the source file. - Insert it into the destination file. - Remove source imports that the move makes unused. - Add the imports that the moved idea needs to the destination. - If the source file still refers to the idea, import it from the destination. - Rewrite comment references per [comment_reference_refactor_support]. - Reindex without duplicate idea rows for the old hosting file. 2. Move idea content ( `reqlan.refactor.moveIdeaContent` ) - Do the same as ( 1 ). - Always import the destination file in the source. - Leave a one-liner stub in the source: `idea_name [alias.idea_name]`. Do not overlap [decompose]: decompose writes a new file. Move and decompose share the resolve-plan-apply sequence in ["../../../packages/extension/src/mutation_module/refactor/idea-extract-flow.ts"]. [refactor_support] [refactor_changes] | 0.106 | |
| refactor_symbol_rename | done | done | Syntactic symbols ( ideas, ideasets, import aliases, and import paths ) should be renamable via the editor rename action ( F2 / LSP rename ). Rename should update AST references across. rq files and matching `rq:["path".idea]` comment idea tokens in source files ( [comment_reference_refactor_support] ). The Rust server renames open buffers, open comment buffers, and closed files reached via the ideas index. Non- `.rq` comment rename stays on the host provider until Rust owns those buffers. [refactor_support] Implemented by ["../../../crates/reqlan-lsp/src/session.rs"], ["../../../packages/language/src/reqlan-rename-provider.ts"], ["../../../packages/language/src/reqlan-comment-rename.ts"], ["../../../packages/language/src/reqlan-name-provider.ts"], and ["../../../packages/extension/src/mutation_module/refactor/register-comment-rename-provider.ts"]. | 0.084 | |
| graph_label_auto | done | done | In Labels auto mode, visibility follows viewport zoom continuously rather than a hard cut. ( fade ) Below a low zoom threshold labels are fully hidden; above a slightly higher threshold they are fully opaque; between those thresholds opacity ramps briefly through a muted / semi-transparent phase ( GRAPH_LABEL_FADE_START → GRAPH_LABEL_FADE_END ). Labels start fully invisible during graph initialisation and layout settling so text paint does not lag those phases; zoom fade only applies once settle / reframe has finished. Hovered or dragged nodes remain fully opaque even when ambient auto opacity is low or zero. Zoom and Fit from [view_controls] / [reframe_view] must update this opacity on each paint or zoom event. Applies to Ideas Summary GraphCyController and HTML export canvas graph alike. | 0.184 | |
| graph_label_modes | done | done | Every interactive graph should provide a Labels control that cycles auto ( default ), on, and off without changing graph membership. On forces full opacity; off hides all labels; auto follows [graph_label_auto]. Reset / default restore auto. Ideas Summary: control in ["../../../packages/extension/webviews/ideas-summary/components/GraphControls.svelte"], mode persisted via ["../workspace-summary/ideas_summary/graphical_graph.rq".graph_ui_persistence]. HTML export: control in ["../../../crates/reqlan-export/src/html_template/mod.rs"] / ["../../../crates/reqlan-export/assets/app.js"]. Driven by the same viewport zoom as [view_controls]. | 0.148 | |
| graph_labels | done | done | Graph node labels must show the full idea name without character truncation. Wrapping is allowed; ellipsis truncation of display names is not. Shared contract for Ideas Summary ( cytoscape ) and HTML export ( canvas ); surface-specific wiring lives under consumers. [graph_label_modes] [graph_label_auto] | 0.098 | |
| graph_status_tag_filters | done | done | Status and tag graph filters are searchable checkbox dropdowns ( still a dropdown trigger; multi-select via checkboxes; OR within each field ). They must distinguish attribute absence from an empty declaration and from a literal unspecified value: - Not present — attribute key absent ( no `@status` / no `@tags` ) - Empty — attribute declared with no value ( `@status`, `@tags`, `@status ""`, bare flag ) - unspecified — literal authored value " unspecified " Do not invent unspecified for missing attributes in rollups, clusters, or filter options. Special options render differently from concrete values ( italic muted Not present; info-tint Empty; warning-tint unspecified ). Shared helpers live in ["../../../packages/analytical/src/core/filter-specials.ts"]. Surfaces: Ideas Summary SearchableCheckboxDropdown; HTML export SCD in ["../../../crates/reqlan-export/assets/app.js"]. | 0.134 | |
| ideas_summary_base_scope | done | done | bases, ideas-summary, multi-base | Ideas Summary is scoped to one [bases.base] at a time. - Active base defaults to the longest-matching base root for the focused editor ( [bases.implied_context] ). - Chrome includes an explicit base switcher; Overview, Bases, Ideas, Ideasets, Attributes, References, Graph, Timeline, and Index tabs bind to the selected base's index only ( Bases lists all discovered bases; Index is last ). - Switching base is a pointer swap ( [bases.active_base_switch] ): pin via `setActiveBaseId`, rebind status / tables / graph UI immediately — do not await sync on the switch call; not-ready bases get fire-and-forget open + soft-sync catch-up. - Graph UI persistence is per-base: workspaceState key `reqlan.ideasSummary.graphUi.<baseId>` via [graph_ui_state]. - Activity bar " Open in Ideas Summary " passes the current active base id into [ideas_summary_panel]. - No merged cross-base tables or graphs in this pass ( [bases.multi_base_environment] ). | 0.081 |
| file_treatment_toggle | done | done | Toolbar dropdown selects how hosting `.rq` files ( implicit ideasets ) appear in the graph — distinct from explicit AST `ideaset` declarations and from external `file:` reference nodes. Options ( with info tooltips / help text ): - invisible / Hidden — no file containers or file leaf nodes; ideas float free - compound — one compound parent per hosting file ( basename label ); nests under folder / tag compounds when Controls → Compound is also on. Containers are draggable ( members move with the box; live physics pins descendants while dragging ). Clicking the title opens the file; double-click also opens. - linked / Linked nodes ( default ) — synthetic ideaset-styled leaf per hosting file ( `rq-file:…` ) with `ideaset_member` edges to ideas in that file; click / dbl-click opens the hosting file like other nodes. Client reshape via `applyFileTreatment` in ["../../../../packages/extension/webviews/shared/graph/file-treatment.ts"]; no re-query on mode flip. Ideas Summary: `FileTreatmentSelect` on GraphControls; persisted as `fileTreatment` in [graph_ui_persistence]. HTML export: same three options via a select; compound boxes are hittable / draggable with clickable titles; linked nodes carry `pageUrl` / `hostFilePageUrl` for navigation. Related: [graphical_graph], [ui_controls], [webview_export_graph_parity], ["../../../ontology.rq".file], ["../../../ontology.rq".ideaset], [html_export_graph_page]. | 0.170 | |
| webview_export_graph_parity | done | done | Interactive graph behaviour that authors rely on in Ideas Summary must have a matching affordance on the HTML export workspace graph ( [html_export_graph_page] ) unless a requirement explicitly scopes it to one host. Shared today: Labels auto / on / off, live physics, Fit, status / tag filters, path / search filters, hide external files, hide ideasets ( explicit AST ), [file_treatment_toggle] ( dropdown with the same three modes, compound drag / title-open, linked-node open ), and [wildcard_refs_toggle] ( default on; dotted wildcard edges ). Ideas Summary may still differ where the host model differs: Indirect refs rebuilds hop depth via `buildGraphViewSlice` ( export freezes neighbourhood at snapshot time — no Indirect chip ); Key / Controls panels and node-budget controls are webview-only; export uses client-side filters on the embedded graph JSON rather than re-querying the index; Ideas Summary uses cytoscape compound parents while export draws AABB compound boxes on canvas. When adding a new Ideas Summary graph chip or edge style, update [html_export_graph_page] ( or document an intentional exception here ) and keep labels / defaults aligned across surfaces. Related: [graphical_graph], [wildcard_refs_toggle], [file_treatment_toggle], [ui_controls], [graph_labels], [graph_label_modes], [graph_status_tag_filters]. | 0.165 | |
| wildcard_refs_toggle | done | done | Toolbar chip " Wildcard refs " ( next to Indirect refs ) on GraphControls toggles `GraphViewQuery.includeWildcardRefs`. Default true — edges of kind `wildcard_reference` from the indexer fan-out stay visible ( ["../../../indexer/indexer.rq".wildcard_reference_edges] ). When false, `buildGraphViewSlice` omits those edges and does not expand neighbourhood through them ( `includeWildcardRefs!== false` normalisation in Ideas Summary host + analytical query ). Distinct from plain `references` and from the Indirect refs depth toggle; Key legend shows a dotted " Wildcard reference " row; cytoscape marks `edge[?isWildcard]` with `line-style: dotted`. Query field flows Ideas Summary loadGraph → `buildGraphViewSlice` in ["../../../../packages/analytical/src/index-store/webview-graph-queries.ts"]. HTML export workspace graph has the same chip ( client filter on embedded edges; dotted stroke ) per [webview_export_graph_parity] and [html_export_graph_page]. Related: ["../../../language/imports.rq".wildcard_references], ["../../../indexer/indexer.rq".wildcard_reference_edges], ["../../../indexer/indexer.rq".index], [graphical_graph], [ui_controls], [webview_export_graph_parity]. | 0.308 | |
| bases_tab | done | done | Table of discovered [ontology.base] entries in the workspace ( label, path, ready, idea / edge / issue counts, state ). Selecting a row pins that base ( pointer swap ) per [ideas_summary_base.ideas_summary_base_scope]. Filterable / sortable / column options like other tables; data comes from index status, not a merged cross-base SQLite dump. Client-side search and column filters via [BasesTable] / bases-filter matcher. | 0.045 | |
| ideas_list | done | done | ideas, git, table | - [IdeasTable] - filterable and sortable via [TableToolbar], [ColumnFilterRow], and [SortableTh] - column show / hide via [TableOptionsMenu] / [table_column_options] - wide tables scroll horizontally in a `.table-scroll` viewport ( columns keep min widths; no crush-to-fit ) - groupable by idea kind ( block / oneliner ) per [group_by_type] - other attributes shown as expandable chips per [ChipList] - clicking an attribute chip adds a column, filters to ideas with a non-null value for that attribute, and sorts by it ( combinable with other filters ) - references split into outbound and inbound chip columns with counts - clicking a reference chip filters ideas that share that edge ( combinable with other filters ) - block idea bodies shown in a scrollable cell per [IdeaBodyCell] - columns for [git_idea_timeline_analysis]: first committed, last modified, and change count - read from indexed `git_created_at` / `git_modified_at` / `git_change_count` only - empty cells until ["../../index-host/git-codelens.rq".git_dates_background_indexing] fills them; main table load never runs git [ideas_list_git_columns_performance] [ideas_list_git_date_format] | 0.079 |
| ideas_list_git_columns_performance | done | done | ideas, git, performance | Ideas table pages that show [git_idea_timeline_analysis] columns must read indexed `git_created_at` / `git_modified_at` / `git_change_count` only — never invoke git or the `git_dates` analyser on the UI query path. At ~ 2 k ideas, a 50 - row page ( including sorts by git modified / change count ) must stay interactive: absolute budgets plus relative guards vs path sort and vs a simulated per-row git walk. Missing-id backlog queries for background fill must also stay fast once the index is filled. Implemented by ["../../../../packages/analytical/src/index-store/sqlite-store.ts"] ( `listIdeasPage`, `listIdeaIdsMissingGitDates` ). [ideas_list] [git_idea_timeline_analysis] | 0.050 |
| ideas_list_git_date_format | done | done | ideas, git, formatting | First committed / last modified cells in [ideas_list] must display a stable `YYYY-MM-DD` authored calendar date from indexed git `%aI` timestamps ( not locale-dependent `toLocaleDateString`, and not raw ISO / patch junk ). Invalid, empty, or non-date values render as an em dash ( `—` ). The `git_dates` analyser must only persist ISO author-date lines so patch hunks never become stored " dates ". Implemented by ["../../../../packages/extension/webviews/ideas-summary/lib/format-git-date.ts"] and ["../../../../crates/reqlan-index/src/git_dates.rs"] ( `parseGitAuthorDates` ). [ideas_list] [git_idea_timeline_analysis] | 0.050 |
| ideas_table_filters | done | done | the ideas table should support text search, column sorting, and attribute-column filters driven by chip clicks. attribute filters require a non-null attribute value in sqlite json per [webview_table_queries] reference chip filters match outbound / inbound edges by target or source id per [webview_table_queries] chips are right-aligned so truncated labels keep the informative suffix visible. active attribute and reference filters appear as removable chips in the table toolbar. Also [table_column_filters], [table_column_options], and [group_by_type]. | 0.040 | |
| index_panel_base_support | done | done | The [index_panel] / Ideas Summary chrome supports switching between bases in the workspace. Switch is a pointer swap per ["../../../bases/base.rq".active_base_switch] and ["./base-scope.rq".ideas_summary_base_scope] — rebind views, do not sync. | 0.040 | |
| overview_coverage_scores | done | done | overview, coverage, metrics | Collapsible " Coverage " section on [overview_page] / [OverviewPanel]. Product-facing metrics for how thoroughly ideas map into the active base's project files ( respecting ["../../host/configuration.rq".configuration_rqignore] ): - Linked files: percent of project files ( non-requirement, non-ignored ) that ideas reference — exact path match after resolve, or under a referenced folder. Show " N of M project files ". - Requirement density: ideas per thousand lines of eligible project code; also show idea count and line count. - Linked paths: distinct files or folders referenced from ideas. - Requirement files: count of requirement documents in the base. Collapsed by default. Calculate only when the user expands the section ( not on Overview open / tab switch ). Non-blocking: host work must not stall Overview chrome or other tabs; show a loading state until the result arrives. When the index is not ready or calculation fails, show a clear message — never fake zero scores. Recalculate on expand when the active base changes or when the user refreshes; otherwise reuse the last result for the session. | 0.047 |
| table_column_filters | done | done | Every data table supports a global search and a per-column filter row ( [ColumnFilterRow] ), matching the usability of [html_export.html_export_list_views]. Text / path / body columns use searchable inputs; closed sets ( idea kind, reference type, ideaset kind, base ready ) use select / multi-select from a fixed option list. Typed filter / search text must keep spaces through the host round-trip ( do not trim the echoed query — SQL matchers may trim only when building LIKE patterns ). Select filters must support deselect / clear back to " All " ( controlled select value + clear control for multi-select ). Column filters are applied in the ui-sql layer ( [webview_table_queries] ) together with existing chip filters. Table panels must bind AppState query / rows via Svelte 5 `$derived` ( not legacy `$:` ) so filtered results from the extension message listener re-render — same pattern as [graphical_graph_doc.graphical_graph] / GraphView. Bases table filters client-side in the webview; other tables round-trip search through [ideas_summary_panel]. | 0.072 | |
| timeline_page | done | done | timeline, ideas, git | Timeline tab shows the evolution of ideas for the active base — not raw file-index noise. Sources: - Git idea lifecycle from indexed `git_created_at` / `git_modified_at` ( created + last edited as separate events when they differ ) - Session idea reindex activity ( one event per idea persisted, with name / status / summary when available ) Missing git dates are filled by the silent background indexer ( ["../../../indexer/indexer.rq".git_dates] / ["../../index-host/git-codelens.rq".git_dates_background_indexing] ) — opening Timeline must not run git log on the UI path. Reindex must preserve existing git dates across upsert. Entries are reverse-chronological, filterable by source ( git / reindexed ), and clickable to open the idea. Each event surfaces idea-centric info: name, short summary, status, kind / tags, and path. Full live multi-commit git log browsing remains out of scope; this surfaces indexed idea history first. | 0.085 |
| avoid_full_list_all_ideas | draft | draft | Prefer `fuzzySearch`, paged SQL, or catalog deltas over materialising every idea. Avoid: - reference-search / wildcard panels calling `listAllIdeas` when search APIs suffice - Ideas Summary `sendFullGraph` dumping all ideas and edges ( [graphical_graph] guidance ) - repeated full-catalog pushes when only a few documents changed [index_cache_reuse] [consumers.extension_ui_consumers] | 0.051 | |
| inbound_snapshot_events | done | done | Host → LSP inbound snapshots are event-driven and incremental. Push only when the ideas index is ready ( catalog update, ready-edge, visible-editor change, or window focus ). While Syncing or not ready, do not push an empty batch that clears the language-server last-good map; set a deferred flag and flush on the next ready event. The language client often starts before the index — registration must ensure-open / defer, not skip forever. Each notification is a per-document merge for visible editors; URIs not in the batch stay as last-good until replaced or the document closes. After a single-file enqueue reaches ready, notify catalog so consumers get a ready-state event ( mid-sync catalog notifies alone are not enough ). Coalesce concurrent push work so catalog spam drains to one ready pass. Related: [unify_inbound_indexes], ["../extension/language/support/open-file-sequencing.rq".lazy_features_after_outbound], ["../extension/language/syntax/features-syntax-highlighting.rq".inbound_inlay_index_performance]. | 0.133 | |
| index_cache_reuse | draft | draft | Identify where the ideas index, diagnostics store, or in-process caches already hold data that other modules recompute. Goal: one walk, one sync, one open store where possible; reuse SQLite edges and mtime maps instead of rebuilding graphs from scratch. Related: [nonblocking_index], [indexing_incrementality], [consumers.index_consumers], [sequencing.index_sequencing]. | 0.164 | |
| overview_coverage_reuse | draft | draft | Overview coverage must not re-walk the base with `.rqignore` when a soft sync just produced an equivalent file set. Reuse the last sync document list or a cheap SQLite documents query. [index_cache_reuse] [sequencing.soft_sync] | 0.040 | |
| path_completion_listing_cache | draft | draft | Import and anonymous file-path completion ( [import_code_completion_ranking], [anonymous_reference_code_completion], [reference_code_completion_objects] ) walk directories via [path_completion]. Reuse directory listings across keystrokes ( TTL cache ) so each keystroke does not re-walk the filesystem; clear the cache in tests and on demand. [index_cache_reuse] | 0.040 | |
| reuse_sqlite_edges_for_hops | draft | draft | Activity bar and Ideas Summary already query hop neighbourhoods from SQLite edges. Language-server reference ranking ( [reference_code_completion_sequencing] ) rebuilds adjacency from open Langium documents in [completion_provider]. When the index is ready, completion hop ranking should prefer indexed edges ( or a shared hop API ) over a second AST-derived graph. [index_cache_reuse] [consumers.extension_ui_consumers] | 0.040 | |
| single_open_store | draft | draft | The extension adapter owns a live [workspace_index] / SQLite connection per base. `getAnalysisApi()` / [analysis_runtime] must not open a second runtime that re-runs `ensure_ready` → `sync_workspace` when the adapter already synced. Reuse the open store for file-context, deprecation, and semantic commands. [index_cache_reuse] [consumers.extension_ui_consumers] | 0.040 | |
| single_sync_walk | done | done | One native directory walk records base roots, `.rq` paths, and allowlisted code paths. The walk does not read file bodies. Startup, Refresh, and create-base each run that walk once. They do not call `findFiles` and they do not run a second native `syncWorkspace` walk. Soft sync indexes the path list and skips unchanged mtimes. Idle stale check stats paths already stored in SQLite. New files arrive from the file watcher. The language server does not walk. The extension sends the base roots after the scan. Overlapping callers share one in-flight scan. CLI and MCP `sync_workspace` uses the same walk ( one pass, stop at nested bases ). [index_cache_reuse] [sequencing.soft_sync] [sequencing.scan_limits] | 0.080 | |
| unify_inbound_indexes | done | done | Inlay hints and code lens must read inbound edges from [ideas_schema.edges_table], not Langium IndexManager. Query: edges whose `target_id` is an idea in the open file, plus `comment_link` / `file_reference` rows whose `target_file` matches that file. File-top CodeLens uses the `file_reference` subset ( ["../extension/language/syntax/features-syntax-highlighting.rq".file_inbound_code_lens] ). The extension host runs the query and pushes a snapshot to the language server. The language server does not open SQLite ( [consumers.extension_ui_consumers] ). Updates follow index sync independently of open-file parse ( ["../language/syntax.rq".open_file_reference_sequencing], [inbound_snapshot_events] ). Mutation hooks already use the same store. Share one graph. Related: [index_cache_reuse], ["../indexer/indexer.rq".local_symbolic_analysis], ["../extension/language/syntax/features-syntax-highlighting.rq".inbound_inlay_index_performance]. | 0.213 | |
| index_consumers | done | done | Modules that read or write [ideas_schema.ideas_index_sqlite] / [index]: - Engine write / read: [workspace_index], Rust sync / store under [ownership] - Extension adapter: [editor_index_adapter] ( watchers, Uri conversion, multi-base ) - Headless: [analysis_runtime] for [cli_package] and [mcp_server] - Activity bar: [activity_bar_host] ( ideas-in-file, hop refs, search via fuzzy ) - Ideas Summary: [ideas_summary_panel] ( table SQL, graph slices, overview ) - Search: [fuzzy_search], [file_search] - Check / analyse / export: [check], analytical export APIs - Local graph: [local_graph_analysis] Product aliases: [index_ideas], [index_technology], [consumption]. Related: [cache_reuse.index_cache_reuse], ["./sequencing.rq".index_sequencing]. | 0.169 | |
| diagnostics_sqlite | done | done | Per-base timing and index-issue store at `<base>/.reqlan/index-diagnostics.sqlite` ( [index_diagnostics_store] / [application_memory] ). Sibling of [ideas_schema.ideas_index_sqlite]; Clear & rebuild of the ideas index must not delete it. Authoritative schema lives in Rust ( [diagnostics_rust], `DIAGNOSTICS_SCHEMA_VERSION = 2` ). The TypeScript facade ( [diagnostics_store_ts] ) is thinner and focuses on timing rows. Related: [index_diagnostics_timing], [index_diagnostics], [index_diagnostics_metrics], [index_diagnostics_webview], [sidecars.sqlite_wal_sidecars], [consumers.diagnostics_consumers]. | 0.210 | |
| git_dates | done | done | Creation and last-modified dates, and change count, for ideas via git history where available. Values are written into the ideas index by the `git_dates` analyser as a silent background job after soft-sync / catalog updates ( batched, yielded ) — not triggered by Activity Bar focus, opening Timeline, or Ideas table load. Host queue and rate caps: ["../extension/index-host/git-codelens.rq".git_dates_background_indexing]. | 0.143 | |
| index_code_files | done | done | Index non- `.rq` source files that contain [comment_reference] ( `rq:[idea]` / `rq:["path".idea]` ). The same directory scan records allowlisted text paths ( respecting `.rqignore`, [app_memory_doc.binary_ignore], [app_memory_doc.commentless_file_ignore], and nested child bases ). Content reads use that list. They are not a second directory walk. [app_memory_doc.text_prefix_skip] skips non-text before a full read. Binary files stay out unless a negation such as `!*.bin` opts them in. The walk cheap-filters on `rq:`, and persists matching files as documents plus `comment_link` edges. Edge shape: source = referenced idea, `target_file` = code file URI, kind `comment_link` ( no synthetic file ideas ). Reindex of a `.rq` file updates kept ideas in place, so their comment links stay. Deletion of a removed idea cascades to its links. Code documents are re-linked only when the set of idea names in that `.rq` file changes, because links resolve on names only. A body edit must not read the code files again. Watcher + idle stale check cover code documents as well as `.rq` files ( [triggers], [nonblocking_index] ). Consumers: [graph_analysers.file_related_requirements], ["../core_analysis/search.rq".file_search], ["../core_analysis/html_export.rq".html_export_code_reference_pages], Ideas Summary comment-reference rows, local graph ( idea → external file node ), ["../cli/click.rq".click_code_file]. See [index_comment_reference_inclusion], ["../ontology.rq".referenced_files]. | 0.394 | |
| index_comment_reference_inclusion | done | done | Comment `rq:[…]` sites are indexed as `comment_link` edges plus first-class code-file documents, not a separate idea kind. Graph surfaces them as idea → external file nodes ( same file-node id as outbound ["../reference_types.rq".file_reference] ). Live editor links still use the comment grammar / resolver; the index path is native [index_code_files]. | 0.098 | |
| indexer_rust | in-progress | in-progress | The [index] engine ( parse-for-index → extract → SQLite persist → incremental sync ) should be written in Rust. Crate: `reqlan-index` over `reqlan-parse` per ["../core_analysis/rust_port.rq".rust_crate_layout]. Extractor ports ["../../crates/reqlan-index/src/extract.rs"] including [wildcard_reference_edges]. Store ports schema v4 ( ["../../packages/analytical/src/index-store/schema.ts"] ) via rusqlite; diagnostics live in `index-diagnostics.sqlite` on the same engine. Sync keeps [nonblocking_index]: per-file upsert, mtime skip, hard rebuild, cancellable progress. Not the Langium LSP. Comment-file indexing: [index_code_files]. | 0.172 | |
| local_symbolic_analysis | done | done | Open-file symbolic analysis must run in Rust on one file only. Input: file URI plus source text ( and optional import-root mappings ). Output: ideas in that file, import binding names in that file, outbound edges with UTF-8 source byte spans, and same-file inbound backlinks ( edges whose target is an idea declared in this file ). Import binding names let a qualified reference name an import alias ( ["../core_analysis/check.rq".check_qualified_missing_idea] ). LSP document-link and error underlines must convert those spans to UTF-16 ( ["../extension/language/syntax/features-syntax-highlighting.rq".reference_underline_syntax_align] ). Cross-file inbound still comes from SQLite via ["./cache-reuse.rq".unify_inbound_indexes]. Do not read SQLite. Do not walk other workspace documents. Do not load an idea catalog. The open-buffer model is ["../extension/language/support/open-file-sequencing.rq".ast_lifecycle]. The hot path parses this buffer once. It does not run this extract as a second parse. Import-target confirmation is a separate depth-1 parse ( ["../extension/language/support/open-file-sequencing.rq".outbound_one_hop] ). That parse is not this extract. The language server caches the open buffer by URI plus text fingerprint ( ["../extension/language/support/open-file-sequencing.rq".open_file_hot_path] ). Wildcard forms stay one `wildcard_reference` pattern edge ( unresolved ); sync still expands them later via [wildcard_reference_edges]. Flag: `ExtractOptions.local_symbolic` forces an empty candidate list. CLI: ["../cli/cli_package.rq".commands] `analyse --local-symbolic <file>`. Sidebar: ["../extension/local-context/activitybar-panels/current_context_scope.rq".reference_lists_pane] merges live same-file inbound / outbound with indexed refs ( [local_symbolic_sidebar] ). Related: ["../language/syntax.rq".open_file_reference_sequencing], [index], ["./cache-reuse.rq".unify_inbound_indexes]. | 0.387 | |
| local_symbolic_sidebar | done | done | The activity bar reference lists and current-file scope must consume [local_symbolic_analysis] for the open buffer text so same-file inbound backlinks and outbound refs appear even when the SQLite index is stale or not ready. Merge live local rows with indexed `listReferencesForIdea` / hop-depth rows; dedupe by direction plus edge id ( and by idea id for scope idea chips ). Prefer the referencer name as the inbound label ( not the edge target label ). | 0.108 | |
| nonblocking_index | done | done | Index sync must stay non-blocking for the host UI ( notably [activity_bar_doc.activity_bar] and [workspace_pane_doc.workspace_loading] ). Editor startup is gated by ["../extension/host/activation.rq".first_paint_startup]; discovering a base must not construct Langium services, which are initialized lazily on the first parse. While syncing, surface progress — state, processed / total, and the current file — in [workspace_pane_doc.workspace_pane] and [ideas_summary_doc.index_panel]. In-flight sync must be cancellable from those surfaces; cancel keeps already-indexed rows and returns the index to ready. Incremental behaviour ( see also [indexing_incrementality] ): - Single-file create / change / delete via watchers indexes only that file — never a full base walk. - Soft full sync ( `syncWorkspace` / Refresh ) skips parse + persist when the file's mtime matches the value stored on the documents row. - Hard refresh ( `clearAndRebuildIndex` / Clear & rebuild ) wipes the store and reindexes every file. - Every sync and stale check applies `.rqignore` to the paths it receives, the same as the base walk; the native single-file index also drops an ignored path. Thus base discovery cannot make the first sync index ignored files. Implemented by [workspace_index]; editor adapter [editor_index_adapter]. | 0.372 | |
| wildcard_reference_edges | done | done | Native extract ( `["../../crates/reqlan-index/src/extract.rs"]` ) expands each AST wildcard against the idea catalog ( `match_wildcard` / workspace candidate list ). Persist one edge of kind `wildcard_reference` per matching idea ( label = target idea name ); never emit plain `references` for wildcard forms. Zero matches → one unresolved edge of the same kind ( `isResolved: false`, label = pattern text ). Edge ids use `edgeId(source, 'wildcard_reference', targetOrUnresolvedKey)`. Sync walks sorted `.rq` files and grows the catalog as it goes, so a wildcard file is expanded against ideas already stored plus files earlier in that walk ( ["../../crates/reqlan-index/src/sync.rs"] ). Ancestor / reference-chain walks treat `wildcard_reference` like `references` ( ["../../packages/analytical/src/index-store/sqlite-store.ts"] ). Ideas Summary references table groups them under sub-idea with other idea-to-idea links; graph visibility is [ideas_summary_doc] / ["../extension/workspace-summary/ideas_summary/graphical_graph.rq".wildcard_refs_toggle]. Soft mtime sync skips unchanged files — after introducing this edge kind, reindex ( or touch ) `.rq` files that contain wildcards so stored rows leave any legacy `references` fan-out. Langium `idea-extractor` keeps the same kind for remaining TS extract paths. Open-file local extract ( [local_symbolic_analysis] ) does not expand wildcards against a catalog. Related: ["../language/imports.rq".wildcard_references], [index], [indexer_rust], ["../extension/workspace-summary/ideas_summary/graphical_graph.rq".wildcard_refs_toggle], [local_symbolic_analysis]. | 0.353 | |
| index_sequencing | done | done | Describe when the ideas index starts, updates, and idles, and which trigger owns each pass. Capability aliases live in [triggers] and [idea_index_triggers]. Detail ideas below; keep [nonblocking_index] and [indexing_incrementality] as the UX rules. Related: [consumers.index_consumers], [cache_reuse.index_cache_reuse], [ideas_schema.documents_table], [diagnostics_schema.sync_runs_table]. | 0.179 | |
| scan_limits | done | done | The workspace scan has two caps. There is no unlimited setting. Defaults and clamp bounds are the `scan` object in ["../../packages/analytical/src/core/defaults.json"]. `reqlan.scan.maxDepth` uses `defaultMaxDepth`, `minDepth`, and `maxDepthLimit` from that file. The workspace folder is depth 0. A directory deeper than the cap is not entered. `reqlan.scan.maxEntries` uses `defaultMaxEntries`, `minEntries`, and `maxEntriesLimit` from that file. The count is directories plus files visited. When a cap hits, the walk stops, keeps the partial result, and the editor shows a warning. It does not start another scan to finish. Refresh uses the same caps. CLI and MCP use the same defaults. Built-in ignore also skips `target/` so a Cargo tree does not consume the entry budget. [extension_startup_sequence] [cache_reuse.single_sync_walk] | 0.064 | |
| reqlan_directory_layout | done | done | Each base owns `<base>/.reqlan/` ( [application_memory] ). Contents: - `ideas-index.sqlite` — [ideas_schema.ideas_index_sqlite] - `index-diagnostics.sqlite` — [diagnostics_schema.diagnostics_sqlite] / [index_diagnostics_store] - `click-sessions.sqlite` — ["../cli/click.rq".click_session] - SQLite WAL / SHM / journal sidecars — [sqlite_wal_sidecars] - `config.json` — [json_siblings] / [base_configuration] - `.rqignore` — [rqignore] - `.gitignore` — [base_initialisation_ignore] / [vcs_ignore_for_stores] - `export_settings.json` — [json_siblings] Path helpers: [application_memory_impl]. Override: `REQLAN_INDEX_PATH` for a single store. | 0.177 | |
| sqlite_wal_sidecars | done | done | When a SQLite file opens in WAL mode, the engine may create sibling files next to the DB: - `ideas-index.sqlite-wal` / `ideas-index.sqlite-shm` / `ideas-index.sqlite-journal` - `index-diagnostics.sqlite-wal` / `-shm` / `-journal` - `click-sessions.sqlite-wal` / `-shm` / `-journal` These are true sidecars of the open database, not separate product stores. Index walks skip them through [rqignore] ( `*.sqlite-wal`, `*.sqlite-shm`, `*.sqlite-journal`, and the same suffixes on `.sqlite3`, `.db`, and `.db3` ). Seeded `.reqlan/.gitignore` must keep `*.sqlite*` artifacts out of VCS ( [base_initialisation_ignore], [ignore_rust] `default_gitignore_file_contents` ). [reqlan_directory_layout] [ideas_schema.schema_migrations] | 0.138 | |
| ideas_index_sqlite | done | done | The ideas graph for one [base_ownership] lives at `<base>/.reqlan/ideas-index.sqlite` under [application_memory]. Schema version is in ( `SCHEMA_VERSION` in [schema_ts] and [schema_rust] ). Runtime migrate runs in Rust ( [store_rust] ); TypeScript opens the store and calls native migrate ( [sqlite_store] ). Keep [schema_ts] aligned with [schema_rust]. Related: [index], [indexer_rust], [ownership], [sidecars.sqlite_wal_sidecars], [diagnostics_schema.diagnostics_sqlite], [consumers.index_consumers]. | 0.226 | |
| idea-tooltip-file-link-accuracy | done | done | Superseded by ["../extension/language/syntax/features-syntax-highlighting.rq".traceable_inlay_hints]: inbound inlay label `location` and tooltip targets both use the shared goto / document-link URI builder. | 0.040 | |
| anonymous_imports_allowed | done | done | A quoted path plus idea — for example `["../ontology.rq".attribute]` — opens that file as an ideaset. Every idea in the target file is in scope for that reference. A from-import or qualified import of the same path only binds the listed names for bare `[name]` references. It must not hide other ideas from a quoted-path reference. | 0.073 | |
| idea_path_filter | done | done | Shared path-filter helper for [wildcard_references_webview] search seeding. Glob or substring match against workspace-relative path and file URI; leading `./` / `../` segments are stripped for relative filters; `**/` is implied when the filter has glob meta but is not already rooted with `**/`. Used only as a hard AND after fuzzy ranking — it never replaces the idea-name query. | 0.078 | |
| wildcard_references | done | done | Qualified path + idea-pattern references match multiple ideas across files. Surface forms: bracket `[…]` and wikilink `[[…]]` ( same `WildcardReference` target ). Path is a quoted string that may include shell-style globs ( `*`, `**`, `?` ) and `@/` import-root aliases ( [configuration_import_root_alias] ). Relative globs resolve against the defining document directory; `.` / `..` segments are collapsed before matching. Idea pattern is an unquoted `WILDCARD_NAME` that must contain at least one `*` or `?` Exact quoted-path plus idea stays ["./syntax.rq".reference_qualified]. Matching uses anchored globs ( not full JS regex ): path `*` is one segment, `**` crosses segments; idea `*` / `?` match within the idea name. Zero matches produce a warning diagnostic. One match also produces a warning — a wildcard is for many ideas; use a concrete ["./syntax.rq".reference_qualified] when only one idea matches ( ["../core_analysis/check.rq".check_wildcard_one] ). Path segment `/**/` must not be lexed as an empty block comment — see ["./syntax.rq".comments]. Lexer note: `ID` is a prefix of many wildcard names ( `import_` vs `import_*` ); the token builder sets `ID.LONGER_ALT = WILDCARD_NAME` so `import_*` is one token. Body prose `*` tokens are accepted via `WILDCARD_NAME` in PlainText so ordinary `*` characters still parse. Without that longer-alt, a live wildcard inside a block body can fail Idea recovery and leave later top-level ideas unindexed ( soft mtime sync will not rebuild until the `.rq` file changes ). Navigation does not emit N LocationLinks. A document-link click opens `reqlan.openWildcardReference` with `{ pathPattern, ideaPattern, fromUri }` ( see [wildcard_references_webview] ). Go-to-definition does not open that panel. Indexing fans out per ["../indexer/indexer.rq".wildcard_reference_edges] ( owned under ["../indexer/indexer.rq".index] ): native extract writes one `wildcard_reference` edge per catalog match; empty catalog / zero matches → one unresolved edge of that kind labeled with the pattern text. ["../core_analysis/check.rq".check] handles 0 - match and 1 - match fan-out through ["../core_analysis/check.rq".check_wildcard_zero] and ["../core_analysis/check.rq".check_wildcard_one] ( ["../core_analysis/check.rq".check_wildcard_sparse] ). Ideas Summary graph toggles those edges via ["../extension/workspace-summary/ideas_summary/graphical_graph.rq".wildcard_refs_toggle]. The server can answer `reqlan/wildcardReferenceAt`. Go-to-definition does not call it. Related: ["./syntax.rq".reference_embedding], ["./syntax.rq".reference_brackets], ["./syntax.rq".reference_wikilink], ["../indexer/indexer.rq".wildcard_reference_edges], ["../core_analysis/check.rq".check_wildcard_sparse], ["../core_analysis/check.rq".check_wildcard_zero], ["../core_analysis/check.rq".check_wildcard_one]. Illustrative forms ( fenced so this file does not embed live wildcards ): | 0.486 | |
| wildcard_references_webview | done | done | A click on the command document link for a [wildcard_references] target runs `reqlan.openWildcardReference` and opens a dedicated **wildcard matches** webview ( beside the editor ) that lists: - Stats: matching idea count, distinct file count, and @ status breakdown - Requirements: each matching idea with name, status, path, and summary ( click opens the idea ) - Files: each matching file with idea count ( click opens the file ) Secondary action **Open in Search** seeds the activity-bar ["../extension/local-context/activitybar-panels/search.rq".search_pane_seed] via `focusIdeaSearch` with: - `query`: idea pattern with glob meta softened for fuzzy search ( `import_*` → `import` ) - `pathFilter`: the path glob as a hard AND filter ( [idea_path_filter] ) Payload builder: `buildWildcardMatchesPayload` filters the index catalog by idea glob + path filter. Protocol for the matches panel is inline host ↔ webview ( `init` / `openIdea` / `openFile` / `openSearch` ). The document-link target is the command URI. Go-to-definition does not open the panel. Highlighting of the source reference must stay link-styled — see ["../extension/language/syntax/features-syntax.rq".syntax_highlighting] and ["../extension/language/syntax/features-syntax-highlighting.rq".wildcard_reference_highlighting]. | 0.199 | |
| parse_budget_timeout | done | done | Lexing, parsing, or a parse subprocess for a single `.rq` file must finish within `parseBudgetMs` from ["../../packages/analytical/src/core/defaults.json"]. If it cannot, do not hang the language server or block the rest of the workspace. Surface both a warning ( budget exceeded; semantic features may be incomplete ) and an error ( file left unloaded ) on that document, then continue with other files. Happy path stays in-process ( sync ) for typical file sizes so edits and workspace load stay fast. Killable worker enforcement is used when the input is large, a file previously hit the budget ( sticky escalate ), or a caller forces the worker path; sync throws still become the same incomplete result shape. The open buffer stores that result as an empty model plus diagnostics ( ["../extension/language/support/open-file-sequencing.rq".ast_lifecycle] ). | 0.197 | |
| context_sensitive_lexer_scaling | done | done | Context-sensitive token classification for structural / prose braces and top-level import keywords must remain linear in document size. The lexer records brace-depth changes sparsely ( not per-character tables ) and looks them up in log time while tokenising, so workspace language-server initialization and editor re-lexes stay bounded. Code fences are skipped during the brace scan so example braces cannot inflate work or desync depth. Bounded work still assumes the lexer / parser *returns*; unbounded hangs are cut by ["./parser_lexer.rq".parse_budget_timeout]. | 0.097 | |
| fencing_comments | done | done | A complete same-line fence keeps `//` and `/*` as prose, not comments. Closed double quotes: " / / this line " contains no comment. Closed backticks: `//this line` contains no comment. Closed backticks with a glob: `@reqlan/*` does not open a block comment. An unclosed backtick is not a fence, so this line ` An unclosed quote is not a fence, so as does " The Langium lexer, the Rust lexer, comment scanners, and TextMate must use this rule. Related: ["./syntax.rq".comments], ["./syntax.rq".inline_code], ["./syntax.rq".naked_strings_in_body], ["../core_analysis/rust_port.rq".comment_span_align]. | 0.078 | |
| attribute_namespacing | done | done | A bracket link `[idea.attribute]` targets one attribute of that idea. The first name is the idea name. The second name is the attribute name. The grammar types that name as an attribute or an idea declaration. `[attribute_namespacing.example]` targets the `example` attribute of this idea. The form `[ideaset.member]` still targets an ideaset member. See [references_to_subidea]. The editor opens the attribute when the first name is an idea. The editor reports an error when that idea has no attribute with the second name. The link target is only that attribute on that idea. HTML export uses the same link form. The export link opens that attribute on the idea page. See [attribute_namespacing.example]. Related: [reference_brackets], [attribute_marker], ["../core_analysis/html_export.rq".html_export_internal_links]. | 0.129 | |
| comment_reference | done | done | Reqlan references may appear in line comments of arbitrary source files. Comment reference syntax is `rq:[idea]` or `rq:["path".idea]` ( quoted path, optional extra qualification ). Qualified paths use the same path forms as imports, including import-root aliases such as `@/` — ["imports.rq".configuration_import_root_alias], ["../extension/language/comment-references/functional-code-comment-references.rq".comment_reference_import_root_alias]. Discovery skips quoted strings, including Rust raw strings ( `r#"…"#`, `r##"…"##`, `br#"…"#` ), so `rq:[…]` text in a string is not a live comment. Editor language coverage is a blocklist, not an allowlist: ["../extension/language/comment-references/functional-code-comment-references.rq".comment_reference_language_blocklist]. Indexed as `comment_link` by ["../indexer/indexer.rq".index_code_files]. Unresolved targets: [comment_reference_resolution_error]. Indexer listing: ["../core_analysis/core.rq".test_references]. Refactors: ["../extension/mutation/refactor_support.rq".comment_reference_refactor_support]. | 0.365 | |
| comment_reference_ignore | done | done | Comment reference ignore syntax uses `//rq-ignore-error`. Comment reference ignore suppresses diagnostics on the immediately following line. The language server and the CLI use the same core scanner [rq_ignore]. | 0.511 | |
| comment_reference_resolution_error | done | done | If a comment reference cannot be resolved, show it as an error diagnostic. Do not create a document link to a missing file ( that underline then raises a generic cannot-find-file editor error ). Related: [comment_reference], [comment_reference_ignore], ["../reference_types.rq".comment_reference], ["../extension/language/comment-references/functional-code-comment-references.rq".references_in_functional_code_comments]. | 0.151 | |
| comments | done | done | Line comments use `//` through end of line ( hidden `SL_COMMENT` ). Block comments use `/* … */` and may span multiple lines ( hidden `ML_COMMENT` ). Meta comments on the same line after content are allowed ( `keep this // meta` ). Comments must not open inside a string / reference-string context: - Naked body quotes ( ["naked_strings_in_body"] ): a quoted slash-slash that is not a comment, quoted urls, `"../mod/**/*.rq"`. - Reference / import `STRING` paths ( `["…"]`, `from "…" import …` ) — the lexer consumes the whole quoted path as one `STRING`, so neither `//` nor `/*` inside the quotes starts a comment. Example: a file reference whose path text contains slash-slash, `["../extension/**/*.rq".*webview*]`. - URL / path-segment guard: `//` immediately after `:` or `/` is not a line comment. Same-line fences: ["./syntax-edge-cases.rq".fencing_comments] — complete `"…"`, `'…'`, and ` ` ` … ` ` ` spans keep ` Unclosed quote or backtick openers are not fences, so `//` after them is a comment. Comment-range identity across lexer, scanner, and TextMate: ["../core_analysis/rust_port.rq".comment_span_align]. Multiline / glob clash: the empty block form `/**/` is **not** a comment — it is the recursive glob segment used in paths such as `../mod/**/*.rq`. A real block comment needs a non-empty body ( `/* note */`, `/*\n note\n*/` ). The token builder rejects length-4 `/**/` matches; comment-span scanners skip the same sequence so embedded `rq:[…]` discovery stays aligned. Related: [comment_reference_ignore], [string_literals], ["./imports.rq".wildcard_references], ["../core_analysis/rust_port.rq".comment_span_align]. | 0.145 | |
| file_layout | incomplete | incomplete | A file begins with zero or more import statements, followed by top-level ideas and ideasets. The file forms an implicit ideaset; its name may default to the filename without the suffix. Front matter syntax is not yet specified. | 0.070 | |
| inline_code | done | done | Single-backtick spans ( `…` ) are opaque `INLINE_CODE`. They may show example syntax such as a quoted path plus idea. The grammar does not parse bracket or file references inside the span. Check and the language server must not report those examples as unresolved ["../reference_types.rq".file_reference] or idea refs. Related: [code_snippets], ["../core_analysis/rust_port.rq".comment_span_align]. | 0.165 | |
| markdown_links | done | done | Markdown `[label](target)` links are supported; the label is not treated as an idea reference. Example: testing requirements and the golden corpus. Related: ["../development/core.rq".testing], ["../core_analysis/rust_port.rq".golden_corpus]. | 0.055 | |
| no_name_idea_safe_warning | done | done | A top-level nameless curly-brace block — `{... }` with no idea name — must not take down the parser or the language server workspace. Parse the block as a recoverable anonymous top-level element, emit a descriptive warning that a name is required before `{`, and continue so later ideas, imports, and links still work. Nameless blocks inside parenthesized lists remain valid list items per [lists]; only top-level nameless blocks warn. Related recovery posture: ["./imports.rq".import_error_recovery]. Block shape: [block_idea]. Fenced examples must stay opaque per [code_snippets] so braces inside snippets cannot desync structural depth. Non-termination ( lex / parse never finishes ) is a different failure mode: ["./parser_lexer.rq".parse_budget_timeout]. | 0.072 | |
| open_file_reference_sequencing | done | done | When the editor opens a `.rq` buffer, reference marks follow ["../extension/language/support/open-file-sequencing.rq".open_file_algorithm]. The open buffer stays file-local ( ["../indexer/indexer.rq".local_symbolic_analysis] ). Cross-file inbound waits for ["../indexer/cache-reuse.rq".unify_inbound_indexes] and ["../extension/language/support/open-file-sequencing.rq".lazy_features_after_outbound]. | 0.193 | |
| reference_resolution_order | done | done | For each `.rq` file, identify the names in local scope, then resolve references. Sequence: 1. Collect all idea, one-liner, and ideaset declarations in the file. Declaration order does not change which names are in scope. 2. Collect reference candidates in the file. 3. Resolve each candidate against those local declarations. 4. Resolve imports only for candidates that have no local match. Stop import lookup for a candidate when a local match exists. The language server and the indexer must use this sequence. Open-file editor underlines follow [open_file_reference_sequencing]. Related: [same_file_reference], [ideaset], ["./imports.rq".import_tokenisation], ["../indexer/indexer.rq".local_symbolic_analysis]. | 0.053 | |
| reference_url | done | done | A bracket reference may target an unquoted absolute URL ( `scheme://…` ). Example: `[https://reqlan.com/]`. The URL is one token; `://` is not a line comment. Unbracketed URLs in body prose stay text. Quoted remote paths stay [reference_file] ( `["https://example.com/file.ts"]` ). Index and check: ["../reference_types.rq".url_reference]. | 0.137 | |
| references_to_subidea | done | done | It is valid to reference a member of an [ideaset] as `[ideaset.member]`, for example [example_ideaset.subidea1]. The indexer must record that reference on the namespace leaf node. In this example the leaf is [subidea1], not [example_ideaset]. Related: [reference_brackets], [same_file_reference], ["../reference_types.rq".idea_reference], ["../indexer/indexer.rq".index]. | 0.167 | |
| idea_graph_distance | done | done | allow searches to match based on partial match and exponential decay of distance to to ideas matching the search term. Used by ["../core_analysis/search.rq".contextual_search]: `finalScore = textScore * 2^(-minHop)` with unreachable floor 8. | 0.088 | |
| katz_centrality | done | done | Score each idea by inbound Katz centrality on the citation graph. An idea scores higher when important ideas reference it. The score uses resolved reference edges and wildcard reference edges between ideas. The analysis package exposes the scores as `katzCentrality()`. The top score is 1. A graph with no citation edges gives every idea the score 1. Search can opt in through ["../core_analysis/search.rq".katz_search]. HTML export shows the score through ["../core_analysis/html_export.rq".html_export_katz_column]. | 0.061 | |
| search_with_context | done | done | function ( exposed in cli ) that allows user to specify current contex ( set of ideas or ideasets or files ), and a search term, yielding the closest matches, in a context aware way. Implemented as ["../core_analysis/search.rq".contextual_search] ( `AnalysisApi.searchRequirements` + `reqlan search --context` ). | 0.067 | |
| atlas | draft | draft | A challenge with navigating complex graphs is that they can be overwhelming. One extension to the product is to have atlases — named, reusable cartographic cuts beyond the live [cartographic_map]. | 0.046 | |
| attribute | done | done | Attributes are classification, functional, or nonfunctional metadata. Syntax: ["./language/syntax.rq".attribute_marker], ["./language/syntax.rq".attribute_forms], ["./language/syntax.rq".attribute_values], ["./language/syntax.rq".attribute_location]. Policy: ["./language/attributes.rq".attribute_configuration]. Indexed on [analytical_types] `IdeaAttributeMap`; extract: [idea_extractor], [rust_extract]. Product surface: [ideas_summary_tabs] Attributes tab. | 0.070 | |
| attribute_body | done | done | The first attribute in an idea, if not marked otherwise, is the main body attribute. It is unstructured text. Syntax: ["./language/syntax.rq".block_idea]; extract summary from that body: [idea_extractor], [rust_extract]. | 0.054 | |
| base | done | done | A base is a container for ideas, ideasets, and files — the boundary of one ideas graph. Discovery: a directory is a base only when it owns a `.reqlan` application-memory folder. There is no implicit base from a bare workspace folder; the product prompts the user to create `.reqlan` at the workspace-folder root when none exist. Bases may be nested: a child directory with its own `.reqlan` is a separate base; the parent base's index stops at that child boundary. Each base holds its own ideas index under its `.reqlan`. Details: ["./bases/base.rq".base]. | 0.251 | |
| cartographic_map | done | done | A cartographic map is a visual representation of a graph, optimised / curated by the user for context. Product: ["./extension/workspace-summary/ideas_summary/graphical_graph.rq".graphical_graph] ( [graph_view] ). Distinct from a full-graph dump; [atlas] is a further curated overlay, not yet implemented. | 0.046 | |
| configuration | done | done | There are multiple scopes for configuration: at the [base] level and at the [extension] level. Base config: ["./extension/host/configuration.rq".configuration_location] ( `<base>/.reqlan/config.json` ). Base vs extension split: ["./bases/base.rq".base_configuration]. | 0.040 | |
| extension | done | done | The extension contains the editor product features. Host: [extension_main]; scope: ["./extension/host/scope.rq".vscode_extension], ["./extension/host/scope.rq".lsp_support]. Architecture: ["./extension/host/architecture.rq".event_driven]; packages: ["./core-architecture.rq".extension_package]. Domain hierarchy: ["./extension/extension.rq".extension_product], ["./extension/extension.rq".extension_domains]. | 0.102 | |
| file | done | done | Files store ideas and form ideasets. Documents use the `.rq` suffix ( ["./language/syntax.rq".file_suffix] ) and ["./language/syntax.rq".file_layout]. Indexed as ideaset kind `file` in [analytical_types]. | 0.094 | |
| grammar_rule | done | done | Grammar rules describe the syntax features. Requirements: ["./language/syntax.rq"]; lexer bridge: ["./language/parser_lexer.rq".lexer_bridge_to_syntax]. Langium: [reqlan_grammar]; comment grammar: [comment_grammar]; index-path AST: [rust_ast]. | 0.040 | |
| idea | done | done | A reqlan core unit is an idea. Syntax: ["./language/syntax.rq".simple_idea], ["./language/syntax.rq".block_idea]. Grammar: [reqlan_grammar] `Idea` / `OneLinerIdea`; index records: [analytical_types] `IdeaRecord`. Extract: [idea_extractor], [rust_extract]; AST: [rust_ast]. Product surface: [ideas_summary_tabs] Ideas tab per ["./extension/workspace-summary/ideas_summary/webview.rq".ontology_aligned_tabs]. | 0.110 | |
| idea_name | done | done | The idea name is a string / handle / variable that uniquely identifies the idea ( within the ideaset / file ). It should not be a reserved [keyword]. Syntax: ["./language/syntax.rq".idea_name]; grammar: [reqlan_grammar] `IdeaName`. | 0.069 | |
| ideaset | done | done | Ideasets are namespace containers for ideas. Syntax: ["./language/syntax.rq".ideaset]. Grammar: [reqlan_grammar] `IdeaSet`. Index kind `ideaset` plus `ideaset_member` edges: [analytical_types], [idea_extractor], [rust_index_types]. A [file] is an implicit ideaset ( ["./language/syntax.rq".file_layout] ). Product surface: [ideas_summary_tabs] Ideasets tab. | 0.134 | |
| import_statement | done | done | Imports describe references. Some [keyword] s are reserved for imports. Forms: ["./language/imports.rq".import_from], ["./language/imports.rq".import_namespace], ["./language/imports.rq".import_qualified]. Grammar: [reqlan_grammar] `Import`; resolve: [import_resolver]. | 0.063 | |
| keyword | done | done | As a language, reqlan has keywords that are reserved in some contexts. Import statement keywords: ["./language/imports.rq".import_keywords] ( `import`, `from`, `as` ). Grammar: [reqlan_grammar] — those tokens are keywords only at statement start. Idea names should not use reserved keywords ( [idea_name] ). | 0.057 | |
| reference | done | done | References link ideas to other ideas, files, and symbols. Syntax: ["./language/syntax.rq".reference_brackets], ["./language/syntax.rq".reference_wikilink], ["./language/syntax.rq".reference_qualified], ["./language/syntax.rq".reference_file], ["./language/syntax.rq".reference_url], ["./language/syntax.rq".reference_embedding]. Kinds: [reference_types]. Grammar: [reqlan_grammar] `ReferenceTarget`; index edges: [idea_extractor], [rust_extract]. | 0.655 | |
| referenced_files | done | done | Non- `.rq` files that ideas reference, or that reference ideas. These may be shared. They should be captured in the ["./indexer/indexer.rq".index]. Outbound file refs: ["./language/syntax.rq".reference_file], [file_reference_resolve], edge kind `file_reference` in [analytical_types]. Inbound comment refs: ["./extension/language/comment-references/functional-code-comment-references.rq".references_in_functional_code_comments] via [comment_resolver] / [comment_grammar]. Full non- `.rq` file indexing of comment-bearing sources: ["./indexer/indexer.rq".index_code_files]. Click treats those files as centres: ["./cli/click.rq".click_code_file]. | 0.169 | |
| github_funding | done | done | The GitHub sponsor button is configured in ["../.github/FUNDING.yml"]. The GitHub Sponsors username is the account in phonebook link id `github-sponsors`. Discord is community, not a funding platform: do not list it in FUNDING. yml. | 0.046 | |
| show_in_footer | done | done | Each entry under `links` in ["./phonebook.json"] may set `show-in-footer`. The property is optional. Omitted means true. Set `show-in-footer` to false to keep a link out of the site footer and the landing contact list. Packages do not have this property. Schema: ["./phonebook.schema.json"]. Typed helpers: `isShownInFooter`, `phonebookFooterLinks` in ["../scripts/phonebook.ts"] and ["../site/src/lib/phonebook.ts"]. per ["./site/site.rq".footer] per ["./site/site.rq".links] | 0.072 | |
| comment_reference | done | done | Inbound ["./ontology.rq".reference] from comments in non- `.rq` source: `rq:[idea]` / `rq:["path".idea]`. Qualified paths may use import-root aliases ( `rq:["@/path".idea]` ) per ["./language/imports.rq".configuration_import_root_alias] and ["./extension/language/comment-references/functional-code-comment-references.rq".comment_reference_import_root_alias]. Spec: ["./extension/language/comment-references/functional-code-comment-references.rq".references_in_functional_code_comments] Syntax note: ["./language/syntax.rq".comment_reference]. Resolve: [comment_resolver]. Unresolved: ["./language/syntax.rq".comment_reference_resolution_error]. Index edge kind `comment_link`. Related capture of those files: ["./ontology.rq".referenced_files]. Indexer: ["./indexer/indexer.rq".index_code_files]. | 0.119 | |
| file_reference | done | done | Outbound ["./ontology.rq".reference] to a file, folder, or symbol — not only `.rq`. Syntax: ["./language/syntax.rq".reference_file] ( bracketed quoted path, optional `L#` / symbol path ). Index edge kind `file_reference`. Resolve: [file_reference_resolve]; extract: [idea_extractor]. Editor links: ["./extension/language/syntax/features-syntax.rq".file_references]. Inbound file-reference CodeLens at the top of the target file: ["./extension/language/syntax/features-syntax-highlighting.rq".file_inbound_code_lens]. Non- `.rq` targets and inbound comment traffic: ["./ontology.rq".referenced_files]. Markdown `[label](target)` also stores this edge kind — see [link_reference]. Quoted body text is not this kind: [reference_edgecase]. | 0.211 | |
| idea_reference | done | done | Idea-to-idea ["./ontology.rq".reference]: bracket `[name]` / `[ideaset.name]` targets a local, imported, or same-file idea. Syntax: ["./language/syntax.rq".reference_brackets], ["./language/syntax.rq".same_file_reference], ["./language/syntax.rq".reference_qualified], ["./language/syntax.rq".reference_embedding], ["./language/syntax.rq".references_to_subidea]. Index edge kind `references` on [analytical_types] / [rust_index_types]. Extract: [idea_extractor], [rust_extract]. A `[ideaset.member]` form records the edge on the member leaf, not on the ideaset container. An attribute uses the same brackets. See ["./language/syntax.rq".attribute_namespacing]. Wikilink surface for the same edge kind: [link_reference]. Fan-out across files: [wildcard_reference]. | 0.125 | |
| ideaset_reference | done | done | References to an ["./ontology.rq".ideaset] or a namespace that behaves as one. Syntax: ["./language/syntax.rq".ideaset]; namespace import alias: ["./language/imports.rq".import_namespace] ( `[alias]` opens the imported file and links only the alias name ). Membership edges are `ideaset_member`; a bare namespace-alias bracket is stored as `file_reference` ( [idea_extractor] ), not `references`. | 0.057 | |
| link_reference | done | done | Link-shaped ["./ontology.rq".reference] surfaces that are not a bare `[idea]` bracket. Wikilink `[[target]]` ( optional alias pipe ): ["./language/syntax.rq".reference_wikilink] — indexes as idea-to-idea `references`, same as [idea_reference]. Markdown `[label](target)`: ["./language/syntax.rq".markdown_links] — label is not an idea ref; the target indexes as `file_reference` ( [file_reference] ). | 0.112 | |
| reference_edgecase | done | done | A quoted string is a [file_reference] only when it is inside brackets. Example of a file reference: ["./reference_types.rq"] Example of body text: " this is not a reference. rq " The file suffix does not change this rule. The parser must not make a file reference from quoted body text. The language server must not make a document link or a missing-file error from quoted body text. Related: [file_reference], ["./language/syntax.rq".naked_strings_in_body], ["./language/syntax.rq".reference_file]. | 0.112 | |
| url_reference | done | done | A bracket reference whose target is an absolute URL with a scheme and `://`. Example: [https://reqlan.com/]. Syntax: ["./language/syntax.rq".reference_url]. Index edge kind `url_reference` on [analytical_types] / [rust_index_types]. Extract: [idea_extractor], [rust_extract]. [check] does not treat a URL reference as a missing idea or a missing file. Unbracketed URLs in body prose stay text, not this kind. Related: [file_reference], [link_reference]. | 0.075 | |
| wildcard_reference | done | done | One ["./ontology.rq".reference] that matches many ideas: quoted path glob + idea pattern ( `*` / `?` ). Syntax and navigation: ["./language/imports.rq".wildcard_references], ["./language/imports.rq".wildcard_references_webview]. Index: one `wildcard_reference` edge per catalog match ( ["./indexer/indexer.rq".wildcard_reference_edges] ); never stored as plain `references`. Native extract: ["../crates/reqlan-index/src/extract.rs"]. Sparse match handling: ["./core_analysis/check.rq".check_wildcard_sparse] ( ["./core_analysis/check.rq".check_wildcard_zero], ["./core_analysis/check.rq".check_wildcard_one] ). | 0.279 | |
| assessment | done | done | A named quiz with a stable id, a title, a blurb, and sections of questions. Questions live in ["../../site/src/content/assessment.ts"] per ["./site.rq".copy]. Current sole quiz id `core`. Sections: ontology, installation, syntax, graph, cli. Ontology covers [idea], [ideaset], [file], [base], [attribute], [attribute_body], [reference]. Advanced syntax covers imports, file refs, wildcards, and comment refs. CLI covers click and `reqlan check` for reference integrity in CI / CD. Pass at 80 %. Then collect a display name and mint a certificate token. The token claim `a` identifies this assessment. After mint, go to [certificate_page] with the token and the just-completed query param. | 0.065 | |
| assessment_page | done | done | `/certs/assessment/` shows the sole [assessment] quiz when there is one. When there is more than one [assessment], it lists them. `/certs/assessment/[id]/` is that assessment's quiz. Unknown ids are not found. Legacy `/assessment/` and `/tutorials/assessment/` redirect here and keep the query string. The home ["./site.rq".quiz_sticker_tab] links here. It does not run the quiz in the tab. | 0.075 | |
| certificate_page | done | done | `/certs/certificate/` shows a minted certificate. Token query param `c` holds the packed claims, including [assessment] id `a`. Query param `just=1` marks the just-completed view: - show the certificate - show a shareable link that omits `just` - embed the sticker form per [sticker_form_embed] Without `just`, the page is the shareable certificate only. Do not embed the form. Invalid or missing tokens send the user back to [assessment_page]. The completion line on a valid certificate names the [assessment] and links to that quiz. Legacy `/tutorials/certificate/` redirects here and keeps the query string. | 0.062 | |
| certs | done | done | Certification lives under `/certs/`. Pages: - [assessment_page] at `/certs/assessment/` ( sole quiz when one [assessment]; catalog when more ) - one quiz per [assessment] at `/certs/assessment/[id]/` - [certificate_page] at `/certs/certificate/` The tutorials catalog and ["./support-page.rq".support_page] advocacy link send the user to [assessment_page]. Tokens are honor-system claims in ["../../site/src/lib/certificate-token.ts"]. Claim `a` is the [assessment] id ( not a section id ). | 0.054 | |
| sticker_form_embed | done | done | The just-completed [certificate_page] embeds phonebook link id `sticker-form` as a Google Form iframe. Href comes from [phonebook], not a hardcoded URL in the view. Convert the href to an embed `src` ( `viewform` plus `embedded=true` ). Keep a direct link to the same phonebook href if the iframe does not load. Hide the form when the user prints the certificate. | 0.048 | |
| accurate_rq_snippets_must_parse | done | done | Every site snippet that demonstrates accurate reqlan must parse against the Langium grammar. In scope: language- " rq " blocks ( landing syntax pairs, the example slice, the hero snippet, showcase `.rq` blocks ). Out of scope: ts / md / py snippets that only show `rq:` comment links. Invalid-on-purpose samples must not be labelled language- " rq ". Enforced by tests plus the content parse gate in ["../../site/scripts/validate-showcase-rq.mts"]. Showcases are a subset per [every_rq_block_must_parse] / [phase_1_parse_gate]. | 0.090 | |
| agent_context_showcase | done | done | ai, mcp, planning | Flagship. Context for the agent, not the repo. An agent answers a question about session expiry. Show the repo-dump cost, then MCP click returning compact file context, then the plan the agent writes. Absorbs former agentic-planning and feature-tracking pages. | 0.050 |
| antipatterns_showcase | done | done | antipattern, craft | Depth. Three ways to write. rq badly. Data instead of pointers, prose restating code, one monolithic file. Showing negative space is a taste signal. Absorbs former data-not-ideas page; expands it. | 0.045 |
| attribute_dialect_showcase | done | done | attributes, extensibility | Depth. Attributes as a domain dialect. The full attribute grammar: flags, negated flags, block values, nested and named lists — forms nobody currently knows about. Absorbs former arbitrary-attributes page. | 0.045 |
| audit_trail_showcase | done | done | compliance, audit, export | Depth. An audit trail you can export. Custom clause / evidence / verified_by attributes, then Completion Status showing the coverage gap, then HTML export as the auditor's deliverable. Domain: regulated software — IEC 62304 or SOC 2. Absorbs former compliance page. | 0.045 |
| broken_links_showcase | done | done | diagnostics, refactor, cli, ci | Flagship. Links that break loudly. Someone renames a file. Show the diagnostic, the quick-fix menu, and the rewritten refs. Then the same issue from ["../core_analysis/check.rq".check] ( `reqlan check`, exit 1 ) and that command as a CI step so a merge cannot land rot. This is the argument against a wiki. Domain: refactor safety, any stack. | 0.190 |
| claim_cycle | done | done | Show one claim at a time and cycle in order. Pause on hover or focus. Tracks jump to a claim. The active claim sits in the vertical centre of a slot sized for the longest claim, so shorter lines do not jump to the top. | 0.056 | |
| claim_expand | done | done | A toggle expands every [elevator_pitch_section] claim into one list. The rotating cycle stays the default. Expanding pauses the cycle. Collapsing returns to the cycle. Reduced motion still starts on the list per [claim_reduced_motion]. | 0.056 | |
| claim_progress | done | done | A track per claim shows hold-time state: past tracks are full, the current track fills left to right, later tracks stay empty. Pause holds the fill. The fill and the cycle share one clock so they do not drift. | 0.056 | |
| claim_reduced_motion | done | done | When the visitor prefers reduced motion, show the full claim list with no cycle and no progress animation. | 0.064 | |
| copy | done | done | Marketing copy lives in typed TS modules under ["../../site/src/content"], composed by ["../../site/src/content/site.ts"]. Shared types: ["../../site/src/content/types.ts"]. Copy stays decoupled from layout / styling. Copy should be really super minimal and concise if at all. ( show don't tell ) Copy should be user facing; not simply a description of the feature requirements. Production code doesn't have latent dev notes in visible components. Landing modules: [hero_section], [elevator_pitch_section], [featured_in_section], [motivation_section], [syntax_section], [example_section], [roadmap_section], [nav_graph], [faq_page], [sidebar_nav], [links], [footer], [sticky_cta_node], [quiz_sticker_tab]. | 0.336 | |
| cta_icon | done | done | The hero has a separate " Install extension " menu button. The menu lists Cursor, VS Code, and Download VSIX. It must not include Open VSX. Open VSX stays on [quickstart_page]. It must not assume the user's editor or pre-select one silently. The CLI control is a separate button per [install_cli_menu]. Get-started CTA copy lives in ["../../site/src/content/hero.ts"] per [hero_section]. Each menu row shows the target's icon beside its full label; icons resolve through [phonebook_icon_registry] via the quickstart icon set / name refs in ["../../site/src/content/install-actions.ts"]. per [components] per [development_core.iconify_stack] | 0.078 | |
| elevator_pitch_section | done | done | Landing `#elevator-pitch` after [hero_section] and before [featured_in_section]. This is the home " Why reqlan " destination: [sidebar_nav] child and [nav_graph] Why node use this id. One-line thesis plus rotating claims. Copy lives in ["../../site/src/content/meta.ts"] per [copy]. Thesis follows ["../constitution.rq".elevator_pitch]. Behaviours: [claim_cycle], [claim_progress], [claim_reduced_motion], [claim_expand]. Styling per [styling_architecture]. Must stay compatible with [simple_lightweight_and_static]. | 0.123 | |
| example_section | done | done | Landing `#example`: one worked graph slice ( imports, file anchors, wildcards, status, tags, tests ). Copy lives in ["../../site/src/content/example.ts"] per [copy]. The slice must parse as accurate reqlan per [accurate_rq_snippets_must_parse]. | 0.091 | |
| faq_drift | done | done | The FAQ includes this question: How do you keep `.rq` files from drifting once the code moves on? `.rq` files are code. They are the source of truth. They are not a document. Specify material work as an idea first. Implementation and tests point back with rq-comments per [development_core.code_comment_references] and [development_core.testing]. A check requires each test to serve a requirement. Agents see only the ideas that attach to the files they touch. Mark a replaced idea deprecated and point to the new idea. The indexer flags missing references like a compile check. Copy lives in ["../../site/src/content/faq.ts"] ( item id `drift` ) per [copy] and [faq_page]. The sidebar nests this question as a same-page child per [sidebar_nav]. | 0.057 | |
| faq_page | done | done | The site exposes `/faq/` as a statically generated page answering a short set of common questions. Copy lives in ["../../site/src/content/faq.ts"] and stays minimal per [copy]. Support answers point at [support_page_reqs.support_page] and at phonebook surfaces ( GitHub stars, marketplace stars, issues / contribution, GitHub Sponsors, Discord ) rather than inventing URLs. Required topics: - how reqlan supports token efficiency - whether reqlan is the de facto standard for efficient developer-centric context management and semantic mapping - when to use reqlan - how `.rq` files stay in sync with code per [faq_drift] - how to support reqlan ( GitHub stars, GitHub Sponsors, VS Marketplace stars, Discord, contribution, bug reporting ) Sidebar nests FAQ question anchors as same-page children per [sidebar_nav]. Landing nav graph includes FAQ as a page node. | 0.132 | |
| featured_in_section | done | done | Landing `#featured-in` after [elevator_pitch_section] and before [motivation_section]. Rolling social-proof banner: label " As featured in: " plus catalogue cards that scroll slowly in a continuous loop. Each card shows an icon and an outbound link label. Required listings ( labels match the catalogue names ): - agentlanguages. dev → https://agentlanguages.dev/languages/reqlan/ - awesome-spec-driven-development → https://github.com/Engineering4AI/awesome-spec-driven-development - awesome-docs → https://github.com/testthedocs/awesome-docs Copy and icons live in ["../../site/src/content/featured.ts"] per [copy]. Icons resolve through [phonebook_icon_component] / [icon_registry]. Links open in a new tab with `rel="noopener noreferrer"`. Pause the scroll on hover or focus. When the visitor prefers reduced motion, show a static row of the same cards with no scroll. Do not put these URLs in the phonebook footer / contact list. Styling per [styling_architecture]. Must stay compatible with [simple_lightweight_and_static]. | 0.122 | |
| firmware_cloud_contract_showcase | done | done | integration, cross-stack, embedded | Flagship. The contract between firmware and cloud. A 10 Hz sample rate and a 50 ms ingest deadline: one invariant, two codebases, neither of which can express it alone. Absorbs half of former glue-semantics and static-connection. | 0.050 |
| footer | done | done | The site footer is on every page via ["../../site/src/components/SiteShell.tsx"]. Copyright copy lives in ["../../site/src/content/meta.ts"] per [copy]. The footer shows phonebook `links` where `show-in-footer` is not false per [phonebook_reqs.show_in_footer]. Icons render through [phonebook_icon_component]. | 0.109 | |
| get_started_cta_motion | done | done | The get-started control on [hero_section] shows a moving colour wash on the border and on the button fill. The wash uses the aqua and rust brand colours. The border wash and the fill wash share one spin period and one spin direction. When the visitor prefers reduced motion, show a static fill and a static border. per [styling_architecture] | 0.058 | |
| graph_view_showcase | done | done | extension, graph | Depth. See the whole graph. Needs real captured screenshots or a short loop from the actual Cytoscape view — not fake ASCII arrows dressed as. rq. Aligns with [diagrams_are_never_dressed_as_code]. | 0.050 |
| hero_github_star | done | done | The [hero_section] action row shows a Star on GitHub control next to get-started and the install controls. The control opens the GitHub repository from [phonebook_reqs.phonebook] in a new tab. The icon comes from [phonebook_icon_component]. Copy lives in ["../../site/src/content/hero.ts"] per [copy]. The hero control is in addition to [sticky_cta_node] and [support_page_reqs.support_page]. | 0.064 | |
| hero_section | done | done | Landing `#hero`: mascot plus a short `.rq` snippet. Copy lives in ["../../site/src/content/hero.ts"] per [copy]. The snippet must parse as accurate reqlan per [accurate_rq_snippets_must_parse]. Install controls are [cta_icon] and [install_cli_menu]; get-started CTA href is `/quickstart` per [quickstart_page]. Get-started motion follows [get_started_cta_motion]. A GitHub star control sits in the action row per [hero_github_star]. | 0.135 | |
| icon_registry | done | done | [phonebook_icon_registry] maps phonebook icon keys to bundled Iconify data for the site. A link or package can omit `icon`. If `icon` is present, it must have a matching registry entry. The module throws at startup if a present icon does not resolve. When you add or change an icon in ["../phonebook.json"], update the registry import for that slug. per [development_core.iconify_stack] per [development_core.phonebook_icon_keys] | 0.064 | |
| install_cli_menu | done | done | The hero has a separate " Install CLI " control. It must not nest inside [cta_icon]. The primary action copies the selected package manager command for [cli_package.cli_package] per [package_manager_preference]. A chevron menu lists npm, pnpm, yarn, and bun. Each command uses the CLI package name from [phonebook_reqs.npm_packages]. A click copies the selected command to the clipboard and stores the preference. The [quickstart_page] CLI and MCP install rows use the same package manager options and the same preference. A copy must not change the last-used editor. | 0.099 | |
| interlock_showcase | done | done | industrial, safety, sequencing | Flagship. The interlock that must not fail. `["./plc/interlock.stL#41-58"]` A safety requirement bound to that PLC symbol and line range, line range, to a vitest case by test name, with an rq-comment in the ST source pointing back. Domain: industrial control / IEC 61131. | 0.050 |
| legacy_archaeology_showcase | done | done | legacy, discovery, deprecation | Depth. Archaeology on inherited code. An undocumented function, an rq-comment seeded into it, @ deprecated on the old path, then Deprecation Impact listing everything that breaks. Absorbs former scattered-docs page. | 0.045 |
| meta_tags | done | done | The site HTML head must include a document title, a description, a canonical URL, Open Graph tags, and Twitter card tags. Canonical origin is the phonebook `site` href per [phonebook_reqs.phonebook]. Default title and description live in ["../../site/src/content/meta.ts"] per [copy]. Root layout ["../../site/src/app/layout.tsx"] sets the shared metadata: title template `%s · reqlan`, description, icons, manifest, theme colour, Open Graph image, and Twitter `summary_large_image`. Each marketing page must set a title and a description. Showcase and tutorial detail pages use the item title and summary. Open Graph image is 1200 × 630 PNG at `/og.png`, generated with favicons per [images]. Compatible with [simple_lightweight_and_static]. Do not index the presentation player route per [presentation_player_route]. | 0.051 | |
| module_surface_showcase | done | done | modularity, imports, ideasets | Depth. A module's published surface. billing / interface. rq exports an ideaset; checkout. rq imports only that; reaching past the boundary is an unresolved-import error. Ideasets-as-public-API is undemonstrated today. Absorbs half of former glue-semantics and module-interface. | 0.045 |
| motivation_section | done | done | Landing `#motivation` after [featured_in_section]: what / why slides for language, extension, CLI, and MCP. Copy lives in ["../../site/src/content/motivation.ts"] per [copy]. | 0.108 | |
| nav_graph | done | done | Sticky landing page graph. Node / edge model and labels live in ["../../site/src/content/nav.ts"] per [copy] / [sidebar_nav]. Section nodes match home anchors; page nodes match top-level routes in [routes]. | 0.123 | |
| package_manager_preference | done | done | The site keeps one global JavaScript package manager value: npm, pnpm, yarn, or bun. The default is npm. All CLI and MCP install commands read and write this one value. Do not keep a separate choice on each control. The value persists in the browser across pages and reloads. Surfaces: [install_cli_menu], [quickstart_page]. Compatible with [simple_lightweight_and_static]. | 0.068 | |
| phase_0_honesty_audit | pending | pending | tooling, honesty | In parallel with content rewrite. Fix what the honesty audit surfaced. | 0.040 |
| phase_1_parse_gate | done | done | Make dishonesty impossible. Add a build step that runs the Langium parser over every language- " rq " block in showcase content files and fails on parse errors or unresolved local refs. Do this first so the rewrite cannot reintroduce the current problem. Also expand Shiki language set per [showcase_languages]. | 0.071 | |
| phase_2_content_schema | done | done | Extend ShowcaseBlock union and Showcase type per [showcase_module]. Presentation polish can follow in [phase_5_detail_layout]; schema first. | 0.040 | |
| phase_3_reqs_as_truth | done | done | This file is the source of truth: one idea per showcase with mechanism, domain, and status; [showcase_set] listing them; refs into showcase content files. Reqlan's own showcase plan is tracked in reqlan. | 0.040 | |
| phase_4_write_the_eleven | done | done | Write the eleven, one at a time, each validated by [phase_1_parse_gate] before moving on. Flagships first — [agent_context_showcase], [interlock_showcase], [broken_links_showcase], [firmware_cloud_contract_showcase] — so the format can be judged before committing to all eleven. | 0.040 | |
| phase_5_detail_layout | done | done | Detail page layout: beats with captions, mechanism badges distinct from tags, and cross-block highlighting of the shared identifier. | 0.045 | |
| phase_6_capture_media | done | done | Capture real media from the extension for [broken_links_showcase] ( diagnostics / quick-fix ) and [graph_view_showcase] ( Cytoscape ). | 0.040 | |
| presentation_player_route | done | done | `/presentations/player/` is a statically exported Next page that loads runtime deck JSON from `/presentations/decks/` and Reveal from `/presentations/vendor/reveal.js/`. Runtime JSON is generated from authored JSONC by [sync_presentations]. It must not use SiteShell; it is iframe-hosted by tutorial detail pages. Do not place `public/presentations/player/index.html` — that path is owned by the App Router export. | 0.068 | |
| quickstart_page | done | done | The site exposes `/quickstart/` as the install / get-started page. Primary path: editor install chooser. Also explains headless [cli_package.cli_package] ( `@reqlan/cli` ) and ["../core-architecture.rq".mcp_package] ( `@reqlan/mcp` ) on the same `.reqlan` index. Package npm hrefs come from [phonebook_reqs.npm_packages], not hardcoded URLs. CLI and MCP install commands offer npm, pnpm, yarn, and bun per [install_cli_menu]. They share one stored preference per [package_manager_preference]. After install steps, it links onward to Tutorials, Showcases, and FAQ so visitors can continue without returning to the home nav. Related link copy lives in ["../../site/src/content/quickstart.ts"] per [copy]. Sidebar nests Extension / CLI / MCP as same-page children per [sidebar_nav]. | 0.113 | |
| quiz_sticker_tab | done | done | Home only: a right-edge tab on the home page. It does not mount in root layout or on other routes. Collapsed: a vertical label and the [phonebook_reqs.phonebook] `sticker-form` icon. Open: a short panel. The CTA is a link to [certs_reqs.assessment_page] through ["../../site/src/lib/certs-paths.ts"]. It is not an in-place quiz. Copy lives in ["../../site/src/content/quiz-sticker.ts"] per [copy]. Style with CSS modules per [styling_architecture] and [brand]. Use theme colour tokens. Hide on narrow viewports so it does not cover content. Keep it compatible with [simple_lightweight_and_static] and [sticky_cta_node]. No looping decorative motion. Open and close may change width once. When the visitor prefers reduced motion, do not animate. | 0.101 | |
| roadmap_section | done | done | Landing page section `#roadmap` after example and before links. Compact vertical scaled-center carousel ( snap + distance-based scale / opacity peeks ). Flat Now / Next / Later items — one focused at a time; neighbours peek smaller. Copy lives in ["../../site/src/content/roadmap.ts"] ( per [copy] ); nav and sticky page graph include the section. Keep the list honest and thin — no fake dates, no wall of tickets. | 0.091 | |
| showcase | done | done | site, showcase, content | Every showcase demonstrates a mechanism, not a metaphor. Domain problem, the. rq that binds it, the real artifact in another language it binds to, and the thing the tooling hands back. Content is authored in TS objects per [showcase_module], not generated from this file at runtime. This file is the source of truth for intent, ordering, and editorial rules. | 0.040 |
| showcase_module | done | done | Showcases are standardised and typed in TS content objects. Types live in ["../../site/src/content/showcases/types.ts"]. Each showcase is its own file under ["../../site/src/content/showcases"]. ["../../site/src/content/showcases/index.ts"] composes them in [showcase_set] order. Each showcase file carries rq: comment references to its showcase idea and to the product ideas it demonstrates. Short version ( list card ): id, title, summary, tags, mechanism badge. Long version ( detail page ): full blocks array including payoff. Exist as a list page and as a full page per slug. Routes: `/showcase/` and `/showcase/[slug]/` — static export. Block kinds needed beyond today's code | features union: callout — the one-sentence point of the page exchange — query + response, terminal-styled diagnostic — editor-style error with quick-fixes diagram — visually distinct from code; never language- " rq " caption / mechanism fields on the showcase object Copy stays minimal and show-don ' t-tell, aligned with site copy rules. | 0.057 | |
| sidebar_nav | done | done | The site sidebar is a hierarchical page nav, not a flat list of mixed anchors and routes. Top-level entries are pages ( or the home hub ). Same-page section anchors nest as children under their parent page. Children render only while the visitor is on that parent page — off-page section lists stay collapsed so the sidebar reflects the current route. Until pathname is known after mount, children stay closed ( do not default to Home ) so static HTML / hydration does not flash a mismatched hierarchy. The active page uses `aria-current="page"`; the in-view same-page section uses `aria-current="location"` when known. Nav model and labels live in ["../../site/src/content/nav.ts"] per [copy]; the Sidebar client reads pathname ( and hash / section when on home ) to decide expansion and current markers. FAQ is a top-level page entry per [faq_page]. Support is a top-level page entry per [support_page_reqs.support_page]. Sidebar nests Support section anchors as same-page children. | 0.187 | |
| spec_html_export | done | done | The site static build exports the workspace requirement graph ( excluding `*.secret.rq` ) under `out/spec/` via `reqlan export`. It must pass [html_export.html_export_url_base] for `/spec` ( honouring `SITE_BASE_PATH`, e. g. `/reqlan/spec` in production ) so `/spec` loads with or without a trailing slash. It must pass [html_export.html_export_header_link] pointing at the marketing site home so readers can leave the embedded spec and return to the parent site. Embedded `/spec` graph pages inherit [html_export.html_export_graph_background_load] so large workspace graphs do not freeze the page while settling. Pipeline: ["../../site/scripts/build-static.mts"] → ["../../site/scripts/export-spec.mts"] → ["../../site/scripts/verify-static-export.mts"]. Host options come from [html_export.html_export] / CLI [cli_package.commands], including native CLI export via ["../core_analysis/rust_port.rq".native_bridge]. CI must build / stage host `reqlan-napi` before this pipeline ( [distribution_reqs.site_distribution] ); the loader probes repo-root `crates/target`, not `packages/crates`. | 0.142 | |
| sticky_cta_node | done | done | Root layout ["../../site/src/app/layout.tsx"] hosts a sticky, draggable button tray on marketing pages. The tray stays in the viewport. Drag starts from a dedicated handle only — action links must not capture pointer or begin a drag. The drag / hit box is the button tray only ( `width` / `height`: fit-content ). Labels and tooltips overlay outside that box; they must not enlarge it. Compact state is icon-only. Hover / focus expands the matching label as a tooltip beside ( web ) or above ( mobile ) the icon, flipped if the tray is near that viewport edge. Narrow ( mobile ) viewports show one action: star the GitHub repository. Wider ( web ) viewports also show " Try the extension ", which opens [quickstart_page]. GitHub href and icons come from [phonebook_reqs.phonebook] / [phonebook_icon_component], not hardcoded URLs or ad-hoc glyphs. Copy lives in ["../../site/src/content/sticky-cta.ts"] per [copy]. Do not show the tray on the presentation player route per [presentation_player_route]. | 0.114 | |
| story_codebase_showcase | done | done | story, traceability, delivery | Flagship. A story that ships with the code. Write the feature as one idea. Bind source with @ implementation. Bind proof with @ tests. Point back with rq: comments. Click on that name returns the bound files — not a ticket outside the repo. Domain: product delivery. One worked name across every block per [one_worked_example_per_page]. Payoff is click output per [every_showcase_ends_in_a_payoff] and ["../cli/click.rq".click]. Uses [development_core.code_comment_references] and [development_core.testing]. Demonstrates [sub_file_binding] and [bidirectional_comment_links]. | 0.046 |
| sync_config_schema | done | done | `scripts/sync-config-schema.mts` copies the canonical base config JSON Schema from ["../../packages/extension/schemas/config.schema.json"] into `public/schemas/v1/config.schema.json`. Public URL: `https://reqlan.com/schemas/v1/config.schema.json` ( path `v1` is schema compatibility; not site package semver ). Canonical authoring and `$id`: ["../extension/host/configuration.rq".configuration_schema_file]. Init seed `$schema`: ["../extension/host/installation.rq".base_config_seed]. Wired into site `predev` and `build-static` ( same pattern as [sync_presentations] ). Deployed with existing `site/v*` pipeline. Synced artifact is not a second authoring source; do not edit under `public/schemas/`. | 0.054 | |
| sync_presentations | done | done | `scripts/sync-presentations.mts` reads authored `*.jsonc` decks and `manifest.jsonc`, emits comment-free `*.json` for the static player, and syncs assets / vendor into `public/presentations/`. | 0.067 | |
| syntax_section | done | done | Landing `#syntax`: self-explanatory rule + practical snippet pairs ( show, don't tell ). Copy lives in ["../../site/src/content/syntax.ts"] per [copy]. Highlighting languages follow [code_block_styling]. language- " rq " pairs must parse per [accurate_rq_snippets_must_parse]. | 0.091 | |
| test_proves_showcase | done | done | testing, traceability | Depth. What this test actually proves. Keep the intent of the former test-explanation page; upgrade it with a real test-name anchor and the reverse direction from test file back to intent. | 0.045 |
| tutorial_location_crumb | done | done | Tutorial detail chrome shows a compact location trail: catalog → course → lesson → slide. Hierarchy must be obvious without a wide single-line title dump. Layout: each tier is a stacked chip — role ( and optional completion control ) on the top row; title / index value muted underneath. Titles colour up on hover, focus-within, or when the peer menu is open. **Folder navigation: ** each tier acts like a path segment / folder. Hover or focus reveals a peer menu for that level of the tree so the user can jump without using the transport row: - catalog → courses ( enter first lesson of the chosen course, or the catalog index ) - course → peer courses - lesson → peer lessons in the active course - slide → peer slides in the active lesson ( updates the embedded player / `?slide=` ) Current peer is marked; menus stay keyboard-reachable via focus-within. **Sizing: ** the crumb container grows to fit text; wrap before truncating. Ellipsis only as a last resort on very narrow viewports — do not cap chip width so aggressively that titles ellipsize while space remains. Course deep-links also exist at `/tutorials/#course-<series>`; catalog at `/tutorials/`. Slide shows `n/total` and updates live with the embedded player. Completion toggles stay on the role row. The crumb lesson tier always shows the **current** lesson only — it must not share hover / preview state with the transport lesson step dots. Step-dot hover may preview a title near the dots, never in the crumb. | 0.046 | |
| tutorial_player_transport | done | done | The `<` control must show the previous slide when the player can go back. When the player cannot go back, `<` must open the previous lesson if that lesson exists. Open that lesson on the last slide. The `>` control must show the next slide when the player can go forward. When the player cannot go forward, `>` must open the next lesson if that lesson exists. Open that lesson on the first slide. The Left Arrow key and the Right Arrow key must do the same as `<` and `>`. Shift + Left Arrow and Shift + Right Arrow must open the adjacent lesson from any slide. The `<<` and `>>` controls must continue to open the adjacent lesson. After the current lesson player is ready, the shell must start to load the next lesson player in the background. The shell must prefetch the next lesson page. The shell must ignore state messages from the background player. | 0.046 | |
| tutorials_section | done | done | The site exposes `/tutorials` and `/tutorials/[slug]/` as statically generated pages. Catalog and deck ids come from authored ["../../presentations/manifest.jsonc"]. Each detail page embeds the Reveal player iframe at `/presentations/player/?deck=<id>` using [assets] / SITE_BASE_PATH-aware `sitePath`. Each detail page hosts a unified transport chrome ( `<<` `<` steps `>` `>>` ) that drives slide nav in the embedded player via postMessage, plus lesson jumps ( [tutorial_player_transport] ). Location hierarchy chrome: [tutorial_location_crumb]. Blurb and try-this live on the catalog and inside the slides, not as page chrome. The player is an App Router page ( [presentation_player_route] ) so it static-exports under `trailingSlash` without colliding with `public/` HTML. Sync ( [sync_presentations] ) converts authored deck / catalog JSONC under ["../../presentations/"] into runtime JSON under `public/presentations/`, and copies assets plus vendored Reveal. js ( player shell excluded ). The catalog links to [certs_reqs.assessment_page] for certification. | 0.046 | |
| support_page | done | done | Rank actions by ease and likelihood to nudge. Star on GitHub is first. Default view is a graph. Toggle Graph / Tiles / Table as icon buttons on the right of the header, with hover tooltips. Graph: cola layout, settled on first paint to the visible window; copy has no graph coordinates; each node moves with its label; nodes are draggable; a drop keeps the new position; a drag does not start the action; all nodes offset slightly toward the cursor; nodes repel so some canvas edges stay open; action radius is impact × ease and stays latent to users; action nodes are solid and flat; each action links to a smaller passive original-group node in secondary colour ( advocacy, community, cite, contribute, sponsor ); group labels grow on group hover; labels do not overlap. Hover on an action node shows a tooltip with the action blurb. Star and sponsor tags are featured in the orange brand colour: the Star action, sponsor actions, and the Sponsor group. Tiles: responsive flex layout. Table: action and group only. Ease, impact, and score stay latent. Action names use the aqua link colour in Graph, Tiles, and Table, including copy and share, except featured graph tags. Each table row runs the same action as the graph and tiles: open a link, copy, or share. Support via, in tile order: - one click - star on github - share - discord - a minute - cite - issues - post your certificate - say something nice on the Visual Studio Marketplace - say something nice on Open VSX - FREE stickers: show the world you are a reqling, via [phonebook] sticker-form - write - reach out and send an email - write a testimonial - sponsor - github sponsorship - direct sponsorship integrated with [phonebook]. Copy lives in ["../../site/src/content/support.ts"] per ["./site.rq".copy]. The site exposes `/support/` as a statically generated page. Sidebar nests Support section anchors as same-page children per ["./site.rq".sidebar_nav]. Landing nav graph includes Support as a page node per ["./site.rq".nav_graph]. | 0.112 |