CodeRAG

Ask questions about your codebase in plain English. CodeRAG indexes your code and uses AI to provide accurate, cited answers.

Quick Start

Prerequisites

• Docker
• Ollama running with llama3

# 1. Start Ollama (if not running)
ollama serve

# 2. Pull the model
ollama pull llama3

# 3. Start CodeRAG
docker-compose up

# 4. Open http://localhost:8000

Usage

Index a Repository

Via UI:

1. Go to the Index tab
2. Enter a GitHub URL or local path
3. Click Start Indexing

Via API:

curl -X POST http://localhost:8000/api/v1/index \
  -H "Content-Type: application/json" \
  -d '{"repo_url": "https://github.com/user/repo"}'

Query Your Code

Via UI:

1. Go to the Query tab
2. Ask a question like "Where is authentication implemented?"
3. Get AI-generated answers with source citations

Via API:

curl -X POST http://localhost:8000/api/v1/query \
  -H "Content-Type: application/json" \
  -d '{"query": "How does the login work?"}'

How It Works

1. Index → Code is parsed, chunked by function/class, and embedded
2. Search → Your question is matched against code using hybrid search
3. Answer → AI generates a response with exact file:line citations

API Reference

Endpoint	Method	Description
/api/v1/query	POST	Query the codebase
/api/v1/query/stream	POST	Query with streaming response
/api/v1/index	POST	Index a repository
/api/v1/index/status/{job_id}	GET	Check indexing progress
/api/v1/repos	GET	List indexed repositories
/api/v1/health	GET	Health check

Configuration

Set via environment variables:

# LLM
CODERAG_LLM__PROVIDER=ollama
CODERAG_LLM__MODEL_NAME=llama3
CODERAG_LLM__BASE_URL=http://host.docker.internal:11434

# Embeddings
CODERAG_EMBEDDING__MODEL_NAME=all-MiniLM-L6-v2

Troubleshooting

Issue	Solution
"LLM not available"	Ensure Ollama is running: ollama serve
Slow first query	Model loading on first use—wait ~10s
No results	Check if repo is indexed in Index tab

License

MIT