A FastAPI service that takes a public GitHub repository URL and returns a structured summary including what the project does, its core technologies, and its directory structure — powered by an LLM.

python -m venv venv source venv/bin/activate # macOS/Linux venv\Scripts\activate # Windows

pip install -r requirements.txt

Create a .env file in the project root:

NEBIUS_API_KEY=your_nebius_api_key_here

uvicorn main:app --reload

The API will be available at http://localhost:8000/summarize.

curl -X POST http://localhost:8000/summarize
-H "Content-Type: application/json"
-d '{"github_url": "https://github.com/psf/requests"}'

Response: { "summary": "Requests is a popular Python library for making HTTP requests...", "technologies": ["Python", "urllib3", "certifi"], "structure": "The main source code lives in src/requests/, tests in tests/, and docs in docs/." }

Error response:

{ "status": "error", "message": "Repository not found" }

This service uses Kimi K2.5 (moonshotai/Kimi-K2.5) via the Nebius API. It was chosen for its strong instruction following and reliable structured JSON output, which is critical for this use case where the response must be parsed directly into a typed Pydantic model.

Repository metadata Name, description, and primary languages from the GitHub API give a fast, reliable output about the project's purpose and tech stack

README The single best source of truth — most well-maintained projects describe their purpose, usage, and structure here

Languages API Provides an accurate list of programming languages without having to scan individual files

File tree The README reliably describes project structure for well-maintained repositories, making a separate tree fetch unnecessary

Individual source files Too large for the LLM context window and rarely needed for a high-level summary