Enterprise Framework RAG
A local hybrid lexical-plus-vector RAG system over a proprietary enterprise Java framework's public API surface, served to agentic coding tools over MCP.
This project brings code intelligence to a proprietary enterprise Java framework that has effectively zero representation in public LLM training data.
Hybrid retrieval
Lexical search preserves exact API names and framework terminology, while vector retrieval captures semantic intent. Combining the two gives coding agents relevant context even when a request does not use the framework's exact vocabulary.
MCP delivery
The retriever is exposed through an MCP server, so agentic coding tools can request grounded framework context inside their normal workflow rather than relying on pasted documentation.
Local and framework-aware
The system runs locally over the framework's public API surface. It helps models produce idiomatic code while keeping proprietary implementation details out of external services.