Abstract
Project Brain Model Context Protocol (MCP) is a local-first, budget-aware hybrid code retrieval system for artificial intelligence (AI) developer assistants, such as Integrated Developer Environments (IDEs) and Command-Line Interfaces (CLIs), that delivers instant, explainable context on large mono repositories (monorepos). The system coordinates a two-stage fast-pass and full search (dense and learned-sparse with Reciprocal Rank Fusion (RRF)), conditional Hypothetical Document Expansion (HyDE) and fail-open reranking, deterministic de-duplicated packing, and ingest-epoch-safe caching to ensure that results stay fast, fresh, and trustworthy even during indexing. Durable memory (for prepare/resume flows) ensures cross-session continuity without needing exact titles.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
C Kuehnl, Todd and Gatzke, Alan, "BUDGET AWARE HYBRID CODE RETRIEVAL FOR AI DEVELOPER ASSISTANTS", Technical Disclosure Commons, (December 11, 2025)
https://www.tdcommons.org/dpubs_series/9013