Skip to content

CLAUDE.md

Latest commit

 

History

History
243 lines (212 loc) · 8.4 KB

CLAUDE.md

File metadata and controls

243 lines (212 loc) · 8.4 KB
Raw

Boop

Boop automates a development workflow by chaining BMAD (planning), Ralph (building), and Claude Code teams (review) into a single pipeline. CLI tool distributed via npm (npx boop).

Built as a fork of OpenClaw — an open-source multi-channel AI gateway (Node.js 22+, TypeScript, pnpm).

OpenClaw Source

Repo: https://github.com/openclaw/openclaw.git (MIT license, main branch)

What to Keep from OpenClaw

  • src/gateway/ — Core gateway server (message routing, auth, sessions, models, agents)
  • src/channels/ — Channel framework (registry, session, targets)
  • src/whatsapp/ — WhatsApp adapter (uses @whiskeysockets/baileys)
  • src/telegram/ — Telegram adapter (uses grammy)
  • src/cli/ — CLI entry point and commands
  • src/agents/ — Agent runtime
  • src/config/ — Configuration system
  • src/sessions/ — Session management
  • src/tts/ — Text-to-speech integration
  • src/shared/ — Shared utilities
  • src/types/ — TypeScript types
  • src/security/ — Security module
  • src/process/ — Process management
  • src/logging/ — Logging subsystem
  • src/browser/ — Playwright integration
  • src/hooks/ — Hook system
  • Root config: tsconfig.json, package.json, pnpm-workspace.yaml, .env.example, vitest.*.config.ts

What to Strip from OpenClaw

Remove completely:

  • extensions/ — All 36 plugin extensions (clawhub marketplace plugins)
  • skills/ — All 50+ skill directories (clawhub skills)
  • src/plugins/ — Plugin framework (loader, registry, hooks, install, manifest)
  • src/plugin-sdk/ — Plugin SDK
  • src/canvas-host/ — Canvas system
  • apps/ — Native iOS/Android/macOS apps
  • Swabble/ — Swift package
  • packages/ — Legacy bot packages (clawdbot, moltbot)
  • ui/ — Web UI
  • vendor/ — Vendored dependencies
  • docs/ — OpenClaw docs (we have our own)

Remove channel adapters (keep only WhatsApp + Telegram):

  • src/discord/
  • src/slack/
  • src/line/
  • src/signal/
  • src/imessage/
  • src/web/ (the web/HTTP channel adapter — NOT the browser module)

Remove features not needed for MVP:

  • src/canvas-host/
  • src/pairing/ — Device pairing
  • src/link-understanding/ — URL content extraction
  • src/media-understanding/ — Media content analysis
  • src/macos/ — macOS-specific integration
  • src/daemon/ — Daemon mode
  • src/terminal/ — Terminal integration
  • src/tui/ — Terminal UI
  • src/wizard/ — Setup wizard
  • src/polls.ts — Polling system
  • src/auto-reply/ — Auto-reply system
  • src/cron/ — Scheduled tasks
  • src/node-host/ — Node hosting
  • src/markdown/ — Markdown processing

After stripping, update all imports, remove dead references from src/channels/registry.ts, src/gateway/, package.json dependencies. Run typecheck to confirm clean.

Target Project Structure

boop/
├── package.json
├── pnpm-lock.yaml
├── tsconfig.json
├── CLAUDE.md
├── README.md
│
├── src/
│   ├── cli/
│   │   └── program.ts              # CLI entry point (npx boop)
│   │
│   ├── gateway/                     # OpenClaw gateway (stripped)
│   │   └── server.ts               # Core message routing
│   │
│   ├── channels/                    # Notification adapters
│   │   ├── whatsapp/
│   │   └── telegram/
│   │
│   ├── voice/                       # ElevenLabs TTS (dormant until Growth)
│   │   └── boop-voice.ts
│   │
│   ├── profile/                     # Developer profile system
│   │   ├── onboarding.ts
│   │   ├── schema.ts
│   │   └── defaults.ts
│   │
│   ├── pipeline/                    # Core pipeline orchestrator
│   │   ├── orchestrator.ts
│   │   ├── epic-loop.ts
│   │   └── state.ts
│   │
│   ├── planning/                    # Planning phase (BMAD-derived)
│   │   ├── viability.ts
│   │   ├── prd.ts
│   │   ├── architecture.ts
│   │   └── stories.ts
│   │
│   ├── bridge/                      # BMAD → Ralph format converter
│   │   ├── parser.ts
│   │   └── converter.ts
│   │
│   ├── build/                       # Build phase (Ralph-derived)
│   │   ├── ralph-loop.ts
│   │   ├── story-runner.ts
│   │   ├── reality-check.ts
│   │   └── progress.ts
│   │
│   ├── review/                      # Review phase (Claude Code team)
│   │   ├── team-orchestrator.ts
│   │   ├── code-reviewer.ts
│   │   ├── tech-debt-auditor.ts
│   │   ├── gap-analyst.ts
│   │   ├── refactoring-agent.ts
│   │   ├── test-hardener.ts
│   │   ├── security-scanner.ts
│   │   ├── qa-smoke-test.ts
│   │   └── fix-runner.ts
│   │
│   ├── scaffolding/                 # Project scaffolding from profile
│   │   ├── generator.ts
│   │   └── defaults/
│   │
│   ├── benchmark/                   # Benchmark harness
│   │   ├── types.ts                 # Suite, case, result, metrics interfaces
│   │   ├── metrics-collector.ts     # Per-phase timing/token/retry accumulator
│   │   ├── mock-provider.ts         # Canned Claude responses for dry-run
│   │   ├── suite-loader.ts          # YAML suite parser and validator
│   │   ├── runner.ts                # Execute cases against pipeline
│   │   ├── scorecard.ts             # JSON + markdown report generation
│   │   ├── history.ts               # Persist/load runs to ~/.boop/benchmarks/
│   │   ├── compare.ts               # Diff runs, detect regressions
│   │   └── commands.ts              # CLI subcommand registration
│   │
│   └── shared/                      # Shared utilities
│       ├── logger.ts
│       ├── retry.ts
│       └── types.ts
│
├── prompts/                         # BMAD planning prompt library
│   ├── viability/
│   ├── prd/
│   ├── architecture/
│   ├── stories/
│   ├── personas/
│   ├── templates/
│   └── checklists/
│
├── benchmarks/                      # Benchmark suite definitions + fixtures
│   ├── suites/                      # YAML suite files (smoke, planning-only)
│   └── fixtures/
│       └── mock-responses/          # Canned responses for dry-run mode
│
└── test/
    ├── unit/
    ├── integration/
    └── fixtures/

User-side directories (created at runtime):

~/.boop/
├── profile.yaml
├── memory/
├── benchmarks/                      # Benchmark run history
│   ├── index.json
│   └── runs/
└── logs/

<project>/.boop/
├── state.yaml
├── prd.json
├── progress.txt
├── planning/
└── reviews/

Tech Stack

  • Runtime: Node.js 22+
  • Language: TypeScript 5.x
  • Package manager: pnpm
  • AI model: Claude Opus 4.6 (all phases)
  • Agent runtime: Pi Agent Core (from OpenClaw's @mariozechner/pi-agent-core)
  • Test runner: vitest (OpenClaw uses it)
  • Linter/formatter: oxlint + oxfmt (OpenClaw's tooling)
  • Build: tsdown (OpenClaw's bundler)

Conventions

  • Files: kebab-case.ts
  • Classes/Types: PascalCase
  • Functions/Variables: camelCase
  • Constants: UPPER_SNAKE_CASE
  • Directories: kebab-case

Pipeline State Machine

IDLE → PLANNING → BRIDGING → SCAFFOLDING → BUILDING → REVIEWING → SIGN_OFF → DEPLOYING → RETROSPECTIVE → COMPLETE

SCAFFOLDING runs once per project (first epic only). DEPLOYING requires cloudProvider config. State persisted to .boop/state.yaml.

Architecture Reference

Full architecture doc: docs/architecture.md Full PRD: docs/PRD.md Epic breakdown: docs/epics.md

Key Dependencies from OpenClaw

These are the critical packages to keep in package.json after stripping:

  • @mariozechner/pi-agent-core — Agent execution engine
  • @whiskeysockets/baileys — WhatsApp
  • grammy — Telegram
  • express — Gateway HTTP server
  • ws — WebSocket
  • playwright-core — Browser automation
  • commander — CLI
  • sharp — Image processing (if needed)
  • vitest — Testing
  • tsx — TypeScript execution
  • tsdown — Bundling
  • oxlint / oxfmt — Linting/formatting