How hsk-2004/codemind works
CodeMind is a local-first AI codebase assistant. A user connects a GitHub repository, previews and indexes its files, then explores summaries, semantic search, and code-grounded questions; the dashboard also surfaces commit summaries and breaking-change analysis. The app exposes these workflows through Next.js pages and tRPC, with PostgreSQL/pgvector for project data and retrieval, local Ollama models by default, and optional cloud chat models. The README documents several flows; unsampled implementations and some UI-to-router wiring are represented only where documented, not inferred.
Main entry points: project.ts Project API · github-loader.ts Repository indexing · retriever.ts Vector retrieval · llm LLM providers
Components 24
App experience
- Workspace pages
src/app/(workspace) - Connect repository
src/app/(workspace)/create/page.tsx - Ask AI
src/app/(workspace)/dashboard/ask-question-card.tsx - Semantic search
src/app/(workspace)/search/page.tsx - Repository files
src/app/(workspace)/files/page.tsx - Dashboard
src/app/(workspace)/dashboard/page.tsx - AI metrics
src/app/(workspace)/metrics/page.tsx - System status
src/app/(workspace)/system/page.tsx
Repository workflows
- Repository indexing
src/lib/github-loader.ts - Commit analysis
src/lib/github.ts - Repository insights
src/lib/repo-insights.ts
AI services
- RAG and direct QA
src/services/rag/chain.ts - Vector retrieval
src/services/rag/retriever.ts - Context excerpts
src/services/rag/excerpt.ts - Embedding provider
src/services/embedding/index.ts - LLM providers
src/services/llm
Server platform
- Project API
src/server/api/routers/project.ts - tRPC API
src/server/api/trpc.ts
Stores and providers
- PostgreSQL pgvector
src/server/db.ts - GitHub API
- Ollama
- Gemini / Groq
- Qdrant
src/services/vectorstore/qdrant.ts
Other components
- Developer
Connections 29
- Developer → Workspace pages: uses
README.md - Developer → Connect repository: connects repository
README.md - Developer → Ask AI: asks questions
README.md - Workspace pages → Project API: uses tRPC
- Connect repository → Project API: starts indexing
- Project API → Repository indexing: loads and indexes
src/server/api/routers/project.ts - Project API → Commit analysis: requests commit work
src/server/api/routers/project.ts - Project API → Repository insights: requests insights
src/server/api/routers/project.ts - Repository indexing → GitHub API: loads repository
README.md - Repository indexing → LLM providers: summarizes files
README.md - Repository indexing → Embedding provider: embeds summaries
README.md - Repository indexing → PostgreSQL pgvector: stores vectors
README.md - Commit analysis → GitHub API: polls commits
README.md - Commit analysis → LLM providers: summarizes and classifies
README.md - Ask AI → RAG and direct QA: submits question
README.md - Semantic search → Vector retrieval: searches embeddings
- RAG and direct QA → Vector retrieval: retrieves sources
README.md - Vector retrieval → Embedding provider: embeds query
README.md - Vector retrieval → PostgreSQL pgvector: queries vectors
README.md - Vector retrieval → Context excerpts: selects excerpts
README.md - RAG and direct QA → LLM providers: generates answer
README.md - LLM providers → Ollama: runs locally
README.md - LLM providers → Gemini / Groq: routes optionally
README.md - Repository files → Project API: loads indexed files
- Dashboard → Project API: loads project data
- AI metrics → tRPC API: requests metrics
- System status → tRPC API: requests health
src/app/(workspace)/system/page.tsx - tRPC API → PostgreSQL pgvector: uses database
src/server/api/trpc.ts - System status → Qdrant: reports provisioned status