-
Notifications
You must be signed in to change notification settings - Fork 0
Releases: papperrollinggery/lazy-brain
LazyBrain 2.0.1
LazyBrain 2.0.1 is the first clean public package release for the deterministic local capability router.
What changed
- Removed the leftover optional HuggingFace/embedding dependency from the public package.
- Kept the published npm package at zero runtime dependencies.
- Clarified first-run behavior:
npm installonly installs the CLI;lb quickstartperforms the explicit local scan and graph build. - Clarified
lb compileas local capability graph compilation, not an LLM or embedding operation.
Install
npm install -g lazybrain lb quickstart lb ready
Try it
lb "review this PR for security issues" lb combo "deploy new feature to production" lb orchestrate "deploy payment feature" lb hook status
Verification
- npm latest:
lazybrain@2.0.1 - Registry install smoke passed.
lb quickstartpassed from the published package.- Public package audit passed with no personal planning files, local paths, or tokens included.
Assets 3
LazyBrain v2.0.0
LazyBrain v2.0.0 is the first npm-published public release.\n\nInstall:\n\nbash\nnpm install -g lazybrain\nlb quickstart\n\n\nHighlights:\n- CLI entries: lb, lazybrain, lazybrain-mcp\n- Claude Code project hook lifecycle: lb hook status|plan|install|uninstall|rollback\n- Deterministic routing with 76 golden routing cases and negative checks\n- MCP stdio tools: find, orchestrate, stats, scan\n- README and install docs updated for npm, MCP, and hook usage\n\nVerification:\n- npm run build\n- npm run lint\n- npm test\n- npm run audit:public\n- npm pack --json\n- npm registry smoke install\n\nPackage:\n- npm: https://www.npmjs.com/package/lazybrain\n- tarball file count: 28\n- no local planning, handoff, prompt, or personal machine files included
Assets 3
LazyBrain v2.0.0 beta.1
e1f5f23 Beta release for LazyBrain v2.0.0.
Install current beta prerelease:
npm install -g https://github.com/papperrollinggery/lazy-brain/releases/download/v2.0.0-beta.1/lazybrain-2.0.0.tgz
lb quickstart
lb "review this PR for security issues"Status:
- Beta, local-first capability router.
- Node.js 18+.
- Deterministic hot path, no runtime LLM call for matching.
- Scans local capability metadata and writes local cache/history.
- Includes CLI commands:
lazybrain,lb,lazybrain-mcp.
Verified:
npm run buildnpm run lintnpm test, 142 tests passnpm run audit:publicnpm publish --dry-run --tag beta --access public- release tarball download and install smoke with empty HOME
lb,lazybrain, andlazybrain-mcpsmoke pass
Tarball SHA256:
c10904c6cfd9af87e7544cec1ca9959ab384eedc42f7fa2097870bb95daa3e75
Npm registry publish is still pending npm auth on the publishing machine:
npm adduser npm publish --tag beta --access public npm view lazybrain dist-tags --json npx --yes lazybrain@beta quickstart
Assets 3
LazyBrain v1.2.0
Summary
LazyBrain v1.2.0 is the public-safe beta release for trusted hook installation and non-install recommendation testing.
Highlights
- Trusted hook workflow:
hook plan, project-scoped install by default, automatic backups, rollback, and explicit--global --yesfor global install. - Non-install Lab at
/labfor visual recommendation checks, agent mapping, team gating, token strategy, and hook readiness. lazybrain readynow blocks on missing graph, unsafe hook state, stale/hung hook records, breaker state, and host load conditions that would make the hook fail closed.- Bilingual README and README_CN now document version status, skills/agents/capabilities, daily usage, install flow, rollback, and troubleshooting.
- Public-surface cleanup removes internal agent workflow protocol docs from the repository.
Recommended Flow
lazybrain scan lazybrain compile --offline lazybrain ready lazybrain server --daemon open http://127.0.0.1:18450/lab lazybrain hook plan lazybrain hook install
Safety Notes
hook planis dry-run only.- Hook install defaults to project scope.
- LazyBrain does not use
Stopas a product lifecycle. - Third-party hooks and HUD/statusline entries are preserved by default.
- Rollback restores only LazyBrain-created backups.
Known Limits
- Recommendation quality is still beta, especially for vague voice-style prompts.
- Semantic/hybrid routing requires embedding config and cache; otherwise it falls back to offline layers.
- Automatic alias promotion is still planned, not treated as mature behavior.
Verification
npm run lintnpm test— 45 files / 544 testsnpm run buildnpm pack --dry-run— 18 files, package surface limited to dist, README, README_CN, CHANGELOG, LICENSE, and package metadata
Assets 2
v1.0.2 — hook safety and diagnostics
LazyBrain v1.0.2 focuses on hook runtime safety, project-scoped activation, and operator diagnostics.
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, andlazybrain 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 --fixnow only repairs LazyBrain-owned state and refuses to silently rebind a missing project scopehook statusand startup diagnostics now surface scope, active hooks, hung hooks, breaker state, and confirm that LazyBrain does not participate inStop- Documentation updated to reflect the sidecar-agent lifecycle, project-scoped hook behavior, and CLI-first runtime guidance
Assets 2
v1.0.1 — test env hotfix
Fixes:
- budget state-machine tests had date-hardcoded entries; now use fake timers (3 tests fixed)
- package.json: remove duplicate esbuild dep that conflicted with vite's bundled version
- optionalDependencies: declare rollup native binaries for arm64/linux CI
- vitest.config.ts: exclude nested sandbox worktree from test discovery
Assets 2
lazy-brain_v1.0.0
LazyBrain v1.0.0
First tagged release. Features:
- Scan + compile + match pipeline (491 capabilities, 366 nodes, 11666 links)
- Five-layer matching engine (alias → auto-alias → tag → embedding → LLM fallback)
- Hook installer for Claude Code UserPromptSubmit
- Offline mode for tag-layer matching without API key
- Semantic embedding matching with bge-m3
- Wiki knowledge base generation
- Usage tracking and evolution (rejection learning, auto-alias, tag evolution)
- Session dashboard and stats
See README for install instructions.