Citer

Ask natural-language questions about any codebase and get precise, cited answers with file and line references.

Multi-turn, agentic, and evaluated.

What is Citer?

Citer is an agentic codebase Q&A tool. Point it at any GitHub repository, and it clones, parses, and embeds the source so you can ask questions like "Where is authentication handled?" — and get answers backed by exact file + line citations, not hallucinations.

The pipeline is a LangGraph agent: planner → retriever tools → synthesizer, grounded in a hybrid retrieval layer that fuses vector search, full-text search, and symbol-level matching (RRF).

Stack

Layer	Technology
Frontend	Next.js (React, TypeScript)
Backend	Python FastAPI
Agent	LangGraph (planner → tools → synthesis loop)
LLM	OpenAI GPT-4o
Embeddings	OpenAI text-embedding-3-large
Vector DB	PostgreSQL + pgvector
Search	Hybrid: vector + PostgreSQL FTS + symbol match (RRF fusion)
Cache	Redis (semantic cache)
Observability	LangSmith
Deploy	Docker Compose

Quick Start

cp .env.example .env   # fill in keys
docker compose up -d   # postgres + redis

# backend
cd backend && pip install -r requirements.txt && uvicorn app.main:app --reload

# frontend
cd frontend && npm install && npm run dev

Open the chat UI at http://localhost:3000, index a repository, and start asking questions.

Repo Layout

├── backend/                 FastAPI service
│   ├── app/
│   │   ├── api/routes/      /index /ask /sessions /eval
│   │   ├── core/            config, logging
│   │   ├── db/              models, queries, session mgmt
│   │   ├── ingestion/       clone → parse → chunk → embed → store
│   │   ├── retrieval/       hybrid: vector + keyword + symbol
│   │   ├── agent/           LangGraph state, nodes, tools, prompts
│   │   ├── cache/           Redis semantic cache
│   │   ├── eval/            golden dataset, runner, metrics, report
│   │   └── schemas/         Pydantic request/response models
│   ├── migrations/          SQL schema + pgvector index
│   ├── scripts/             CLI: index a repo, run eval
│   └── tests/               ingestion, retrieval, agent
├── frontend/                Next.js app
│   ├── app/chat/            streaming chat UI + citations
│   ├── app/eval/            eval dashboard
│   ├── components/          markdown, SSE reader
│   └── lib/                 API client
├── data/                    local repo clones / cache
└── docker-compose.yml       postgres + redis

Roadmap

• Backbone: DB + Docker, ingestion pipeline, hybrid retrieval, LangGraph agent, /ask streaming.
• Product: Next.js chat UI + citations, sessions & multi-turn, semantic cache, eval harness + dashboard, deploy.

License

MIT