WDBX memory & retrieval

What it can help with

Build document recall and decision-history prototypes with explicit source ownership.

These are potential pilot applications, not claims of customer outcomes.

What the source implements

The durable store combines snapshots, write-ahead recovery and a layered HNSW search index. The hybrid retrieval library exposes score components. Canonical episode persistence is a separate v3 surface; consumers must use its defined semantics rather than assuming every legacy vector record is an admitted episode. The episode store can sign each episode digest with the canonical writer's Ed25519 key, records quarantine, contradiction and resolution memory-edge episodes, and caps the ledger it will append to.

Available today

Rust persistence, HNSW search and episode-store components are implemented.

How to evaluate it

Start with a synthetic or authorized corpus. Check recovery, source identifiers, update behavior and retrieval relevance. Treat application permissions, data retention and deployment evidence as additional requirements.

Supporting research and guides

Read the accompanying archive entries: WDBX: A Weighted-Backtrace Memory Store for Traceable Retrieval; WDBX Graph Weights for Traceable Neural Retrieval; Vector Index Maintenance Under Continuous Ingestion. Each distinguishes implemented behavior from proposed applications and includes pinned implementation evidence.