Dubsbot

Model-agnostic local coding agent CLI inspired by Claude Code, built with TypeScript and powered by Vercel AI SDK.

Highlights

• Local-first agent loop (gather -> reason -> act -> verify)
• Structured outputs with strict Zod validation
• Provider adapters for OpenAI, Anthropic, and Google
• Approval-gated execution policy for safer command/tool actions
• PGLite-backed local persistence for sessions, tools, retrieval metadata, and context indexing
• Hybrid context retrieval (lexical + vector + graph-scored ranking)
• Optional daemon for automations and file/git watch re-indexing
• Extensible command/hook model via AGENTS.md

Requirements

• Node.js 24+
• pnpm 10+

Quick Start

pnpm install
pnpm db:migrate
pnpm build

CLI Commands

pnpm dev -- chat
pnpm dev -- chat "summarize this repo"
pnpm dev -- plan "create a rollout plan for indexing"
pnpm dev -- index .
pnpm dev retrieval-proof --profile smoke
pnpm dev -- automations list
pnpm dev -- automations add --name "Hourly Check" --cron "0 * * * *" --prompt "summarize local status"
pnpm dev -- automations run

Run daemon:

pnpm dev:daemon

Interactive Loop UX

• The interactive chat loop now exposes lifecycle phases in the TUI: initializing, planning, awaiting_approval, executing, interrupted, resuming, completed, failed.
• Press Ctrl+C during active execution to request a controlled interrupt. The loop stops at the next safe checkpoint and saves resumable state.
• Use /resume to continue an interrupted turn from the last durable checkpoint.
• Sensitive tool actions (write and destructive) require an explicit in-TUI decision:
  • /approve to execute
  • /deny to skip
  • /dismiss to skip

Quality Commands

pnpm lint
pnpm lint:fix
pnpm typecheck
pnpm test
pnpm build

Project Layout

• src/cli - Ink TUI and command entrypoints
• src/agent - orchestrator and turn schemas
• src/providers - AI SDK provider adapters
• src/policy - approval and safety policy logic
• src/context - indexing, retrieval, file/git watchers
• src/db - PGLite client and migrations
• src/automation - scheduler, hooks, automation runner
• src/observability - traces/transcripts and optional OTel
• src/mcp - MCP process client
• tests - unit/integration-facing tests for core behavior

Configuration

Environment variables (BYOK):

• DUBSBOT_PROVIDER (openai | anthropic | google) - defaults to google
• OPENAI_API_KEY
• ANTHROPIC_API_KEY
• GOOGLE_GENERATIVE_AI_API_KEY
• DUBSBOT_OPENAI_MODEL
• DUBSBOT_ANTHROPIC_MODEL
• DUBSBOT_GOOGLE_MODEL (defaults to gemini-3.1-pro-preview)
• DUBSBOT_OTEL_ENABLED=1 to enable telemetry export hooks
• DUBSBOT_EMBEDDING_STRATEGY_V2=1 to enable explicit embedding strategy resolution/fallback
• DUBSBOT_EMBEDDING_STRATEGY_CONFIG_JSON to provide explicit strategy config
• DUBSBOT_EMBEDDING_PROVENANCE_LOG=1 to emit embedding provenance log lines

Notes

• Anthropic embeddings currently fall back to deterministic local vectors.
• This project intentionally uses Biome only (no ESLint/Prettier).
• Retrieval proofing benchmark schema/workflow docs: docs/retrieval-proofing-benchmark-schema.md and docs/retrieval-proofing.md.
• Embedding strategy rollout guide: docs/embedding-strategy-rollout.md.