reid41/QA-Pilot

What it solves

QA-Pilot helps developers quickly understand and navigate unfamiliar GitHub code repositories by allowing them to chat with the codebase using a Large Language Model (LLM).

How it works

The system clones public GitHub repositories and uses vector embeddings to index the code. It then provides an interactive chat interface where users can ask questions about the repository. It integrates with codegraph to visualize Python and Go files and supports a wide range of LLM providers, including local options like Ollama, LocalAI, and llama.cpp, as well as cloud APIs like OpenAI, Anthropic, and MistralAI.

Who it’s for

Developers who need to rapidly onboard onto new open-source projects or analyze existing GitHub repositories without manually reading through every file.

Highlights

• Broad Model Support: Compatible with numerous providers including Ollama, OpenAI, Anthropic, MistralAI, and local runners like llama.cpp.
• Code Visualization: Integrated with codegraph for viewing Python and Go files.
• Session Management: Supports multiple chat sessions with history storage and a search function to locate sessions quickly.
• Local-First Capability: Designed to validate the feasibility of fully local QA solutions using local LLMs and vector embeddings.