Amure Db

Create a backend knowledge graph database, named amure-db, written in Rust using the Axum framework. This system should function as a RAG backend for AlphaFactor, managing knowledge structured around Hypothesis nodes. The core functionality needs to support storing hypotheses with details like statements, abstracts, and discussion, along with associated experiments (Universe, Regime, Temporal, Combo). It must model relationships between hypotheses using edges (Reference, Superset, Subset, Orthogonal), ensuring every edge has a mandatory reason. Implement graph traversal capabilities, including an in-memory graph structure and BFS walk. Crucially, include a search mechanism that uses OpenAI embeddings for cosine similarity search, followed by MMR reranking, and also supports keyword fallback. The system should expose an HTTP endpoint for searching and loading root nodes with their immediate edges.