Architecture
The project keeps a small pipeline:
RepositoryScannerreads useful text files from a local repository.PythonAnalyzer,StructureAnalyzerandDependencyAnalyzerproduce typed facts.RepositoryRAGdelegates indexing and retrieval topython-rag-framework.DocumentationAgentdelegates prompt execution topython-agent-runtime.- Markdown generators build focused documentation files.
OutputWriterwrites generated files to disk.
The default generation path is deterministic so tests and examples run without paid LLM APIs.
Hosted OpenAI, Anthropic and Gemini adapters are optional. They receive a deterministic draft as grounding, then rewrite it through the selected provider.