What happens when your team has been using AI coding assistants for months and nobody fully understands what got built? SourceBridge is a new open-source project that tries to solve exactly that problem - generating documentation, architecture diagrams, and guided code tours from any codebase automatically.
The tool uses tree-sitter parsing (a fast, language-aware code parser) to analyze repositories in Go, Python, TypeScript, JavaScript, Java, Rust, and C++. It then feeds that structural understanding to an LLM of your choice to produce what the project calls "field guides" - cliff notes, learning paths, workflow stories, and system explanations at the repository, file, and individual function level.
What It Actually Does
- Auto-generated architecture diagrams using Mermaid, built from actual code structure rather than someone's outdated Confluence page
- AI-powered code reviews examining security, SOLID principles, performance, and maintainability
- Impact analysis that simulates a code change and traces which other parts of the system would be affected
- Requirement tracing - import requirements from Markdown or CSV, and SourceBridge links them to the code that implements them
- Conversational exploration with full codebase context, so you can ask questions about the code and get answers grounded in the actual source
The LLM flexibility is the standout feature. SourceBridge supports Anthropic, OpenAI, Google Gemini, and OpenRouter for cloud, plus Ollama, vLLM, llama.cpp, and LM Studio for local inference (running models on your own hardware without sending code to external servers). You can even assign different models to different operations to balance cost and quality.
The stack is Go for the API server, Python for AI reasoning via gRPC, React 19 with Next.js on the frontend, and SurrealDB for storage. It ships with Docker Compose for quick setup and Helm charts for production Kubernetes deployments. Air-gapped installations are supported for teams that can't send code off-network.
Licensed under AGPL-3.0, which means any modifications must also be open-sourced. This is a young project and the real test will be how well the generated documentation holds up on large, messy, real-world codebases. But the premise is right - AI-generated code is creating a documentation debt that traditional approaches can't keep up with.