VectorDB Integration

GeneralBots uses VectorDB as the vector database for storing and searching embeddings. The Rust client for the configured VectorDB is used to communicate with the service.

Configuration

The connection is configured via environment variables:

VECTORDB_URL=http://localhost:6333
VECTORDB_API_KEY=your-api-key   # optional

These values are read at startup and passed to the VectorDBClient.

Collection Mapping

Each .gbkb collection maps to a VectorDB collection with the same name. For example, a knowledge base named company-policies becomes a VectorDB collection company-policies.

Operations

Insert – Performed during indexing (see Chapter 03).
Search – Executed by the FIND keyword, which sends a query vector and retrieves the top‑k nearest neighbors.
Delete/Update – When a document is removed or re‑indexed, the corresponding vectors are deleted and replaced.

Performance Tips

Keep the number of vectors per collection reasonable (tens of thousands) for optimal latency.
Adjust VectorDB’s hnsw parameters in VectorDBClient::new if you need higher recall.
Use the FILTER option to restrict searches by metadata (e.g., source file).

Example `FIND` Usage

USE_KB "company-policies"
FIND "vacation policy" INTO RESULT
TALK RESULT

The keyword internally:

Generates an embedding for the query string.
Calls VectorDB’s search API.
Returns the most relevant chunk as RESULT.

1.5 KiB Raw Blame History Unescape Escape