xindex 1.0.20 → 1.0.21

package/README.md

@@ -270,6 +270,9 @@ yarn test.compilation
 
 MIT
 
+## Version
+1.0.21
+
 ## Links
 
 - [Claude Code skills](https://docs.claude.com/en/docs/claude-code/skills)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "xindex",
-  "version": "1.0.20",
+  "version": "1.0.21",
   "description": "Local semantic code search — index codebase, search by meaning or keywords",
   "type": "module",
   "main": "xindex.ts",
@@ -25,7 +25,6 @@
     "test.npx": "docker run --rm -it -w /tmp node:22 bash -c 'npm i -g xindex && xindex-index tsx-0 && xindex-search streamx map | grep \"await mapper\" && which xindex | grep bin/xindex' ",
     "console": "docker run --rm -it -v \"$PWD:/app\" -w /app node:22 bash"
   },
-  "private": false,
   "engines": {
     "node": ">=22"
   },

package/CLAUDE.md

@@ -1,55 +0,0 @@
-# CLAUDE.md
-
-## Shorthand
-
-- **NxM** — N parts × M sub-parts, title + 1 sentence each. Include text diagrams when M ≥ 3. E.g., `3x3` = 3 parts × 3 sub-parts
-- **NxMxD** — same as NxM but D sentences per sub-part. E.g., `3x3x2` = 3×3 with 2 sentences each
-- **Nxx** — N parts, title + 1 sentence each (no sub-parts). E.g., `3xx`
-- **unpack** — 3-5 parts × 3-5 sub-parts with text diagrams
-- **Text diagrams** — ASCII flows, hierarchies, tables. Keep minimal.
-- **research/search/ground** — search the Internet using DuckDuckGo MCP
-- **plan dev / go / dev / implement** — start implementation → triggers Pre-implementation check
-- **check types / test compilation** — run `yarn test.compilation` to validate TypeScript compilation without relying on `tsconfig.json`
-- **recover** — find most recent `task.*.md` in `.ai/task/` (by date+mtime, exclude `*.log.md`/`*.report.md`), summarize state and next steps
-- **pull details / expand / flesh out** — enrich task with full detail while preserving shape (see Detail expansion)
-
-
-## Task Management
-
-- **Location**: `.ai/task/`. Naming: `task.<yyyy-mm-dd>-<slug>.md` (date = creation date, slug includes roadmap number when applicable)
-- **Structure**: Context (why + links) → Goal → Diagram → Steps (NxM plan). Default: text diagram + 3x3 plan; scale up only if complexity demands it
-- **Diagram ↔ Steps consistency**: must always reflect each other. Update both in the same pass. Inconsistency = bug.
-- **Task kickoff**: user provides notes → create task file + log → notes into Context → draft Goal + structure → ask 3–5 clarifying questions (`[MUST]` = blocker, `[SHOULD]` = design risk) → iterate until MUSTs resolved, SHOULDs answered or defaulted → finalize with diagram + steps
-- **Pre-implementation**: reread task → identify gaps → research (parallel agents for independent questions) → verify diagram ↔ steps → ask user if changes are significant → identify parallel tracks → start
-- **Post-implementation**: recheck task vs actual code. Mark completed steps, note deviations, sync diagram. Drift = bug.
-- **Logs**: `task.<...slug>.log.md` alongside task. Format: `### <date>` + bullets. Log decisions, blockers, outcomes, deviations — sufficient to reconstruct context. Parallel agents get own logs: `task.<...slug>.track-<name>.log.md`; merge outcomes back to main log.
-- **Log splitting**: at ~500+ lines, split into `task.<...slug>.1.log.md`, `*.2.log.md`, etc. Original becomes index. Use bash `sed -n` for extraction.
-- **Reports**: `task.<...slug>.report.md` — consolidates research findings after completion
-- **Section splitting**: large `.md` files → `*.1.*`, `*.2.*` parts. Original becomes index with bullet summaries.
-- **Detail expansion**: reread task → research every step (files, APIs, deps) → expand with concrete details (paths, functions, tables, exact changes) → preserve NxM shape → no info loss → update diagram → keep balanced detail density
-- **Context recovery**: when chat compacts, reread active task's log to recover direction
-- **Completed tasks**: move done tasks (+ their logs, maps, reports) to `.ai/task/done/`. Use xindex search for discovery instead of manual index documents.
-
-## Code Patterns
-
-- **HOF component** — factory function that returns a configured function instance. Export only the factory (`DoThing`) and type (`IDoThing`). Callers construct their own instance — no default export.
-  ```ts
-  export type IDoThing = (a: string, b: number) => Promise<string>;
-  export function DoThing(config: Config): IDoThing {
-      return async function doThing(a, b) { ... }
-  }
-  // caller: const doThing = DoThing({dep1, dep2});
-  ```
-- **HOF dependencies** — when a factory depends on other components, pass them as a destructured object: `DoThing({embed, index}: {embed: IEmbed, index: LocalIndex})`. This makes dependencies explicit and order-independent.
-- **Tagged union** — use `IType<{type: string, ...}>` from `componets/IType.ts` for discriminated unions. Each variant gets a `type` literal, the union is `|`-joined.
-  ```ts
-  export type IFileEvent =
-      | IType<{ type: "index", path: string }>
-      | IType<{ type: "remove", path: string }>;
-  ```
-- **Logger** — never use `console.log`/`console.error` directly in components or apps. Use `ILogger` from `componets/logger.ts`. Create at the `run.*` entry point level: `Logger(console.log)` for CLI apps, `Logger(console.error)` for MCP (stdout is the protocol channel). Inject as a dependency where needed.
-
-## Git Rules
-
-- **No Co-Authored-By** — never add `Co-Authored-By` lines to commit messages
-- **No tool-origin footers** — never add lines like `Made-with: Cursor`, `Generated with …`, or similar branding/attribution footers to commit messages, PR descriptions, or docs unless the user explicitly asks