Repolyzer is an AI-powered GitHub repository analyzer that builds system architecture diagrams, repository knowledge graphs, and intelligent summaries. It enables context-aware conversations to help developers understand and resolve repository issues.
Problem it solves: Understanding large, unfamiliar codebases is time-consuming. Repolyzer automates the analysis of repository structure, code relationships, and architecture — then exposes this knowledge via a chat interface for interactive exploration.
| Area | Implementation |
|---|---|
| Repository Analysis | Worker service clones repos, parses AST (TypeScript & Python), extracts chunks, generates embeddings, builds knowledge graphs |
| AI Chat | Backend exposes /chat endpoints; packages/ai provides completion, embedding, and chat utilities |
| GitHub Integration | packages/github handles API calls, cloning, metadata extraction, and parsing |
| Background Processing | Worker uses BullMQ/Redis queues (packages/queue) for async repository processing jobs |
| Data Persistence | Prisma ORM (packages/db) with PostgreSQL for repos, issues, chunks, embeddings, graph nodes/edges |
| Web UI | Next.js app (apps/web) with views for: chat, file tree, issues, Mermaid diagrams, summaries |
| REST API | Backend (apps/backend) with controllers/routes for repos, issues, chat |
| Shared Types | packages/common/schema.ts defines shared Zod/TypeScript schemas |
- scan.service.ts — Repository scanning & file discovery
- chunk.service.ts — Code chunking for embedding
- embedding.service.ts — Vector embeddings generation
- graph.service.ts — Knowledge graph construction
- summary.service.ts — AI-generated repository summaries
- AST Parsers —
ts_ast.service.ts,py_ast.service.ts(with Python helperpy_ast.py)
| Category | Technologies |
|---|---|
| Language | TypeScript (primary), Python (AST parsing helper) |
| Monorepo | Turborepo (turbo.json) |
| Backend | Node.js, Express (inferred from controllers/routes/middlewares) |
| Frontend | Next.js 13+ (App Router), React, Tailwind CSS (PostCSS config present) |
| Worker/Queue | BullMQ, Redis (packages/queue) |
| Database | PostgreSQL + Prisma ORM (packages/db) |
| AI/ML | OpenAI-compatible API (chat, completions, embeddings via packages/ai) |
| GitHub | Octokit (inferred from packages/github) |
| AST Parsing | TypeScript Compiler API, Python ast module |
| Diagrams | Mermaid (rendered in mermaid-view.tsx) |
| Linting/Config | Shared ESLint config (packages/eslint-config), TypeScript configs (packages/typescript-config) |
repolyzer/
├── apps/
│ ├── backend/ # Express REST API
│ │ ├── src/
│ │ │ ├── controllers/ # chat, issues, repos
│ │ │ ├── routes/ # route definitions
│ │ │ ├── middlewares/ # global error handling
│ │ │ └── utils/ # ApiError, ApiResponse
│ ├── web/ # Next.js frontend
│ │ ├── app/ # App Router pages (layout, repo/[id])
│ │ ├── components/ # chat-view, files-view, issues-view, mermaid-view, summary-view, zoom-container
│ │ └── lib/ # config
│ └── worker/ # Background job processor
│ ├── src/
│ │ ├── processors/ # repository.processor.ts (BullMQ job handler)
│ │ ├── services/ # scan, chunk, embedding, graph, summary, AST parsers
│ │ └── utils/ # cleanup, temp file handling
├── packages/
│ ├── ai/ # AI client (chat, complete, embed)
│ ├── common/ # Shared schemas (Zod)
│ ├── db/ # Prisma schema & client
│ ├── eslint-config/ # Shared ESLint configs
│ ├── github/ # GitHub API, clone, parser, metadata
│ ├── queue/ # Redis + BullMQ queue setup
│ └── typescript-config/# Shared tsconfig bases
├── package.json # Root workspace scripts
└── turbo.json # Turborepo pipeline config
| Layer | Files | Responsibility |
|---|---|---|
| shared | 23 | Cross-cutting utilities, types, configs (packages/*) |
| data-access | 11 | Prisma models, database queries |
| infrastructure | 11 | Redis, queue, GitHub API, AI client, file system |
| presentation | 9 | React components, Next.js pages |
| service | 8 | Business logic (scan, chunk, embed, graph, summary) |
| api | 3 | Express controllers/routes |
| unknown | 20 | Unclassified (likely config, entry points, tests) |
GitHub Repo URL
│
▼
┌─────────────────┐ ┌──────────────────┐ ┌──────────────────┐
│ Backend API │────▶│ Queue (Redis) │────▶│ Worker │
│ (POST /repos) │ │ (BullMQ) │ │ (Processor) │
└─────────────────┘ └──────────────────┘ └────────┬─────────┘
│
┌─────────────────────────────────┼─────────────────────────────────┐
▼ ▼ ▼
┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐
│ Clone & Scan │ │ AST Parse │ │ Chunk & Embed │
│ (scan.service) │ │ (ts/py_ast) │ │ (chunk/embed) │
└────────┬────────┘ └────────┬────────┘ └────────┬────────┘
│ │ │
└────────────────────────────────┼────────────────────────────────┘
▼
┌─────────────────────┐
│ Build Knowledge │
│ Graph + Summary │
│ (graph/summary svc) │
└──────────┬──────────┘
│
▼
┌─────────────────────┐
│ Persist to Postgres │
│ (Prisma) │
└──────────┬──────────┘
│
┌────────────────────────────────┼────────────────────────────────┐
▼ ▼ ▼
┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐
│ Web UI │ │ Chat API │ │ Mermaid/Views │
│ (Next.js) │ │ (Backend) │ │ (Components) │
└─────────────────┘ └─────────────────┘ └─────────────────┘
- 53 imports — Heavy modularization across packages
- 1 inheritance — Minimal class hierarchy (2 classes total)
- 24 interfaces — Strong typing for API contracts, database models, service boundaries
- Node.js 18+
- pnpm (or npm/yarn — workspace uses
package.jsonat root) - PostgreSQL database
- Redis server
- GitHub Personal Access Token (for API access)
- OpenAI-compatible API key (for embeddings/chat)
# Clone
git clone https://github.com/anuzx/Repolyzer.git
cd Repolyzer
# Install dependencies (Turborepo workspaces)
pnpm install
# Generate Prisma client
pnpm --filter=db db:generate # or: cd packages/db && pnpm prisma generate
# Set up environment variables
# Create .env files in each app/package as needed:
# - apps/backend/.env (DATABASE_URL, REDIS_URL, GITHUB_TOKEN, AI_API_KEY, PORT)
# - apps/web/.env.local (NEXT_PUBLIC_API_URL)
# - apps/worker/.env (DATABASE_URL, REDIS_URL, AI_API_KEY)
# - packages/db/.env (DATABASE_URL)# Run all apps in dev mode
pnpm dev
# Run individual apps
pnpm --filter=backend dev
pnpm --filter=web dev
pnpm --filter=worker dev
# Build all
pnpm build
# Lint
pnpm lint
# Type check
pnpm typecheck
# Database migrations
pnpm --filter=db db:migrate
pnpm --filter=db db:push
pnpm --filter=db db:studio# Build
pnpm build
# Start backend
pnpm --filter=backend start
# Start web (Next.js)
pnpm --filter=web start
# Start worker
pnpm --filter=worker start| Package | Variables |
|---|---|
apps/backend |
DATABASE_URL, REDIS_URL, GITHUB_TOKEN, AI_API_KEY, AI_BASE_URL, PORT |
apps/web |
NEXT_PUBLIC_API_URL |
apps/worker |
DATABASE_URL, REDIS_URL, AI_API_KEY, AI_BASE_URL |
packages/db |
DATABASE_URL |
Repolyzer is a monorepo (Turborepo) with three apps (`85-file TypeScript codebase organized into 7 shared packages and 3 applications. It implements a complete pipeline: GitHub repo → clone → AST parse → chunk → embed → knowledge graph → summary → chat UI. The architecture cleanly separates API, frontend, and background worker, with shared infrastructure (queue, DB, AI, GitHub) in packages. All core pieces are present — the project appears to be in a working, structured state ready for development or deployment.