refactor!: remove unfinished LazyBrain surfaces

.github/workflows/ci.yml

@@ -10,6 +10,10 @@ jobs:
   test:
     name: Test
     runs-on: ubuntu-latest
+    env:
+      npm_config_fetch_retries: 5
+      npm_config_fetch_retry_mintimeout: 20000
+      npm_config_fetch_retry_maxtimeout: 120000
 
     steps:
       - uses: actions/checkout@v4

.github/workflows/release.yml

@@ -22,6 +22,10 @@ jobs:
   npm-release:
     name: npm release
     runs-on: ubuntu-latest
+    env:
+      npm_config_fetch_retries: 5
+      npm_config_fetch_retry_mintimeout: 20000
+      npm_config_fetch_retry_maxtimeout: 120000
 
     steps:
       - uses: actions/checkout@v4

.gitignore

@@ -1,6 +1,7 @@
 node_modules/
 dist/
 .lazybrain/
+.claude/
 *.tsbuildinfo
 .env
 *.env
@@ -45,3 +46,6 @@ scripts/check-budget.sh
 
 # Local workflow doc (user-facing, but not for public repo — keep local)
 HOW_TO_WORK_WITH_TRAE.md
+.gstack/
+.gitnexus
+.gitnexus.*

CHANGELOG.md

@@ -1,138 +1,9 @@
 # Changelog
 
-All notable changes to this project will be documented in this file.
+## [v1.5.0]
 
-The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
-and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
-
-## [Unreleased]
-
-## [v1.5.0] - 2026-04-27
-
-### Added
-- Persistent statusline with active (bold) vs dormant (dim) visual distinction, always visible in combined HUD mode.
-- Complete UI/UX redesign: unified design system, dark mode support, Chinese-first scrollable single-page layout.
-- Chinese admin panel with 6 sections: status overview, live route tester, tool browser, API config editor, system diagnostics, and setup guide.
-- Inline API config editing (compile/embedding/secretary LLM settings, API keys, routing engine) with save-to-config.
-- `POST /api/config` endpoint to write configuration from the web UI (14 whitelisted keys).
-- `GET /api/diagnostics` endpoint returning hook runtime stats, recent events, graph status, and embedding cache health.
-- Tiny gate hook now runs lightweight tag-layer matching and injects real results with scores (🟢🟡⚪), tool names, descriptions, and personality roasts.
-- Auto language detection (zh/en) for hook-injected routing suggestions.
-- Project CLAUDE.md with model-friendly install instructions.
-
-### Changed
-- Search API (`/api/search`) now returns all nodes when no query or filter is specified, fixing the empty tools display.
-- Lab page redesigned to match the unified design system.
-
-### Fixed
-- JavaScript syntax error in setup guide cmd strings due to unescaped newlines in TypeScript template literals.
-- Config fields displaying as "未配置" (not configured) due to field name mismatch in renderConfig().
-- `autoThreshold` validation now rejects NaN values.
-- `POST /api/config` now checks Origin header for defense-in-depth.
-- Tiny gate hook now writes `last-match.json` so the statusline shows real match results instead of "建议路由".
-- Platform-specific native packages moved to `optionalDependencies` for cross-platform CI compatibility.
-
-## [v1.4.5] - 2026-04-26
-
-### Added
-- `RouteSpec` schema version, `no_route_needed` mode, `tokenStrategy`, and route rationale fields for stable cross-surface routing.
-- `lazybrain prompt "<query>" --target claude|codex|cursor|generic` with explicit `--copy` clipboard support.
-- Read-only MCP stdio server via `lazybrain mcp --stdio`, exposing `lazybrain.route`, `lazybrain.search`, `lazybrain.skill_card`, and `lazybrain.combos`.
-- Privacy-preserving route counters through `lazybrain route stats`.
-- `lazybrain hook status --json` for runtime diagnostics including skip reason, breaker state, active/hung runs, and p95 duration.
-
-### Changed
-- The default Claude hook is now a tiny gate: it performs a fast complexity/vagueness/risk check and injects only a short reminder to call `lazybrain.route`.
-- `/api/route` now uses the same history/profile inputs as CLI routing and enforces query/body size limits.
-- Public docs now position MCP and prompt output as the main low-intrusion route surfaces, with hook as a reminder gate.
-
-### Security
-- MCP tools are read-only, do not execute skills, do not install hooks, do not return agent bodies, and do not read transcripts.
-- Hook route telemetry stores only hashes and compact metadata, not raw prompts.
-
-## [v1.4.0] - 2026-04-25
-
-### Added
-- Advisory Route Plan orchestrator via `lazybrain route "<query>"`, `--json`, and `--target generic|claude|codex|cursor`.
-- Stable `RouteSpec` output with intent, scenario, skills, workflow, context needed, guardrails, verification, done conditions, adapter prompts, warnings, and clarification questions.
-- Optional SKILL.md frontmatter schema fields: `useWhen`, `avoidWhen`, `inputs`, `workflow`, `verification`, `doneWhen`, `contextNeeded`, and `guardrails`.
-- Built-in combo templates for frontend pages, redesigns, CEO dashboards, public install docs, regression review, stuck-runtime debugging, and public release audit.
-- `lazybrain combos [category]` for read-only combo discovery.
-- Verification catalog for UI screenshots, dashboard operating questions, docs readability, code checks, hook dry-run, rollback, privacy scan, and package dry-run.
-- `POST /api/route` and GUI Try Router Route Plan display.
-
-### Changed
-- Route planning stays outside the matcher; `match()` remains retrieval-only while the orchestrator builds execution guidance.
-- README and README_CN now document RouteSpec, combo templates, advisory-only behavior, and schema metadata.
-
-### Security
-- Route planning does not execute skills, install hooks, read transcripts, return agent bodies, or write Claude/Codex/Cursor configuration.
-
-## [v1.3.0] - 2026-04-25
-
-### Added
-- Local Web GUI via `lazybrain ui`, with Overview, Try Router, Skill DB, Hook Safety, Lab, Health, Troubleshooting, and Settings pages.
-- Read-only GUI/status APIs: `/`, `/ui`, `/api/status`, `/api/health`, `/api/stats`, `/api/search`, `/api/embeddings/status`, and Lab API aliases.
-- Explicit action APIs for `POST /api/test` and `POST /api/embeddings/rebuild` with confirmation gates.
-- CLI status homepage as the default `lazybrain` output.
-- `lazybrain api test` for compile LLM, secretary LLM, and embedding checks without printing keys.
-- `lazybrain embeddings status` and `lazybrain embeddings rebuild --yes` with temp-file atomic cache writes.
-- Public audit gate through `npm run audit:public`, PR template, optional Codex review guide, and GitHub release workflow.
-
-### Changed
-- `lazybrain --version`, `/health`, `/api/health`, package metadata, and changelog now share one package-version source.
-- CI keeps a stable required `Test` check while covering Node 18, 20, and 22, package dry-run, public privacy scan, hook tests, and Lab/server smoke.
-- README and README_CN now document GUI usage, API testing, embedding cache rebuild, release gates, and bug recovery.
-
-### Security
-- Public audit blocks private paths, local planning docs, personal email markers, token-like secrets, private runtime directory markers, and internal workspace-name leaks.
-- Root `AGENTS.md` is no longer tracked in the public repository.
-- GUI v1 does not install hooks, read Claude transcripts, return agent body text, or write `.claude/settings.json`.
-
-## [v1.2.0] - 2026-04-25
-
-### Added
-- Non-install LazyBrain Lab at `/lab` for visual recommendation testing, agent mapping, team gating, token strategy, and hook readiness.
-- Agent inventory scanner for project, user, and plugin agents using metadata only.
-- Trusted hook install workflow with dry-run plan, automatic backups, rollback, readiness checks, and global-install confirmation.
-- Advisory team model guidance, runtime adapters, and subagent prompt suggestions.
-
-### Changed
-- Documentation now recommends scan, offline compile, ready check, Lab preview, hook plan, then project-scoped install.
-- Hook docs now separate implemented behavior from planned capabilities and clarify semantic fallback behavior.
-- `lazybrain ready` now blocks when hook breaker state, hung records, or host load would make the hook fail closed.
-- README and README_CN now include v1.2.0 release positioning, skill/agent metadata coverage, daily usage, and troubleshooting guidance.
-
-### Security
-- Redact sensitive config values in CLI output.
-- Lab and hook plan responses avoid agent body text, Claude private transcripts, local home paths, and statusline secret parameters.
-- Project-scope runtime guard now canonicalizes symlinked workspace paths before comparing cwd.
-- Remove internal agent workflow protocol documents from the public repository.
-
-## [v1.1.0] - 2026-04-23
-
-### Added
-- Add baseline token cost calculation for accurate token savings in session statistics and dashboard.
-
-## [v1.0.2] - 2026-04-20
-
-### Added
-- Project-scoped hook install metadata and workspace `cwd` guard so LazyBrain only runs inside the intended repo by default
-- Hook runtime registry, active run inspection, and breaker diagnostics via `lazybrain doctor`, `lazybrain hook ps`, and `lazybrain hook clean`
-
-### Changed
-- Hardened hook runtime safety with concurrency limits, hung/stale run handling, overload breaker checks, and fail-closed scope behavior when install metadata is missing
-- `doctor --fix` now only repairs LazyBrain-owned state and refuses to silently rebind a missing project scope
-- `hook status` and startup diagnostics now surface scope, active hooks, hung hooks, breaker state, and confirm that LazyBrain does not participate in `Stop`
-- Documentation updated to reflect the sidecar-agent lifecycle, project-scoped hook behavior, and CLI-first runtime guidance
-
-## [v1.0.0] - 2026-04-19
-
-### Added
-- Step 1: Cleanup embedding dead code and fixup decision type identification quality
-- Step 2: Decision type identifier for classifying user intents
-- Step 3: Team recommender for intelligent agent team formation
-- Step 4: Thinking trigger for proactive tool suggestions
-- Step 5: Duplicate detector for identifying redundant tools/skills
-- Step 6: HTTP API server for desktop UI integration
-- Step 7: Real usage data tracking for analytics and improvement
+- Kept the verified route, MCP, compile, embeddings, ready, statusline, status, and diagnostics path.
+- Replaced the unfinished web console with a compact Workbench for status, diagnostics, route, compile, embeddings, and API tests.
+- Removed unfinished choice preference, route adoption, route regression, public jobs, repairs, doctor-fix, and config-test/schema surfaces.
+- Removed unpublished planning documents and the bundled Cytoscape asset from the public package.
+- Kept RouteSpec `1.5.0` output stable for CLI, HTTP, and MCP consumers.

CLAUDE.md

@@ -81,3 +81,47 @@ Key routing rules:
 - Ship/deploy/PR → invoke /ship or /land-and-deploy
 - Save progress → invoke /context-save
 - Resume context → invoke /context-restore
+
+<!-- gitnexus:start -->
+# GitNexus — Code Intelligence
+
+This project is indexed by GitNexus as **lazy-brain** (3580 symbols, 10213 relationships, 267 execution flows). Use the GitNexus MCP tools to understand code, assess impact, and navigate safely.
+
+> If any GitNexus tool warns the index is stale, run `npx gitnexus analyze` in terminal first.
+
+## Always Do
+
+- **MUST run impact analysis before editing any symbol.** Before modifying a function, class, or method, run `gitnexus_impact({target: "symbolName", direction: "upstream"})` and report the blast radius (direct callers, affected processes, risk level) to the user.
+- **MUST run `gitnexus_detect_changes()` before committing** to verify your changes only affect expected symbols and execution flows.
+- **MUST warn the user** if impact analysis returns HIGH or CRITICAL risk before proceeding with edits.
+- When exploring unfamiliar code, use `gitnexus_query({query: "concept"})` to find execution flows instead of grepping. It returns process-grouped results ranked by relevance.
+- When you need full context on a specific symbol — callers, callees, which execution flows it participates in — use `gitnexus_context({name: "symbolName"})`.
+
+## Never Do
+
+- NEVER edit a function, class, or method without first running `gitnexus_impact` on it.
+- NEVER ignore HIGH or CRITICAL risk warnings from impact analysis.
+- NEVER rename symbols with find-and-replace — use `gitnexus_rename` which understands the call graph.
+- NEVER commit changes without running `gitnexus_detect_changes()` to check affected scope.
+
+## Resources
+
+| Resource | Use for |
+|----------|---------|
+| `gitnexus://repo/lazy-brain/context` | Codebase overview, check index freshness |
+| `gitnexus://repo/lazy-brain/clusters` | All functional areas |
+| `gitnexus://repo/lazy-brain/processes` | All execution flows |
+| `gitnexus://repo/lazy-brain/process/{name}` | Step-by-step execution trace |
+
+## CLI
+
+| Task | Read this skill file |
+|------|---------------------|
+| Understand architecture / "How does X work?" | `.claude/skills/gitnexus/gitnexus-exploring/SKILL.md` |
+| Blast radius / "What breaks if I change X?" | `.claude/skills/gitnexus/gitnexus-impact-analysis/SKILL.md` |
+| Trace bugs / "Why is X failing?" | `.claude/skills/gitnexus/gitnexus-debugging/SKILL.md` |
+| Rename / extract / split / refactor | `.claude/skills/gitnexus/gitnexus-refactoring/SKILL.md` |
+| Tools, resources, schema reference | `.claude/skills/gitnexus/gitnexus-guide/SKILL.md` |
+| Index, status, clean, wiki CLI commands | `.claude/skills/gitnexus/gitnexus-cli/SKILL.md` |
+
+<!-- gitnexus:end -->