Crate phago_rag 

Phago RAG

Biological Retrieval-Augmented Generation.

Instead of vector similarity search, queries traverse the Hebbian knowledge graph following strongest connections. The graph learns from usage — frequently traversed paths strengthen, unused ones decay.

How it differs from standard RAG

Standard RAG	Phago RAG
Chunk → embed → vector search	Digest → wire → graph traversal
Static index	Self-reinforcing graph
No learning from queries	Traversed paths strengthen
No anomaly detection	Sentinels flag what doesn’t fit
Flat retrieval	Structured, weighted paths

Re-exports

pub use query::Query;

pub use query::QueryResult;

pub use query::QueryEngine;

pub use hybrid::hybrid_query;

pub use hybrid::HybridConfig;

pub use hybrid::HybridResult;

pub use mcp::phago_remember;

pub use mcp::phago_recall;

pub use mcp::phago_explore;

Modules

baseline

Baseline retrieval methods for comparison against bio-rag.

code_query

Code-aware query interface.

hybrid

Hybrid scoring — TF-IDF candidate selection + graph structural re-ranking.

mcp

MCP Adapter — Model Context Protocol interface for Phago.

prelude

Phago RAG Prelude — convenient imports for common usage.

query

Query engine — traverses the Hebbian knowledge graph to retrieve relevant concepts.

scoring

Information retrieval scoring metrics.