LanceDB support request

Environment

LocalAI/LocalRecall/LocalAGI/BMAD/KiloCode

Current backend: Chromem/Chromadb

Desired backend: LanceDB

📖 Summary

LocalRecall currently defaults to Chromem/Chromadb, with Qdrant and Milvus listed as planned integrations. I propose adding LanceDB as an additional backend option.

💡 Motivation

Lightweight & Embedded: LanceDB doesn’t require running a separate service, unlike Qdrant or Milvus. It stores vectors locally in Parquet files, making it ideal for developer‑friendly setups.

Performance: Built on Apache Arrow + Parquet, LanceDB offers fast local vector search and efficient memory usage.

Ecosystem Alignment: KiloCode has already added LanceDB support. Adding it to LocalRecall would unify the stack and reduce fragmentation between agents and allow for seamless turn-key integration of BMAD+KiloCode+LocalAI+LocalRecall+LocalAGI.

Avoids Bloat: For users who find Qdrant too heavy, LanceDB provides a leaner alternative without sacrificing retrieval quality.

✅ Expected Behavior

LocalRecall should allow users to configure LanceDB as a backend for both short‑term and long‑term memory storage. This would enable seamless memory sharing between BMAD, KiloCode, and LocalAGI agents using a unified lightweight backend.

🔧 Proposed Solution

Implement a LanceDB backend adapter that conforms to LocalRecall’s existing backend interface.

Map LocalRecall’s memory schema (embeddings + metadata) to LanceDB tables.

Provide configuration options to select LanceDB as the backend.

Add unit tests to validate insert/query/delete operations and ensure retrieval quality.

📈 Benefits

Simplifies local deployments by avoiding external DB services.

Improves performance for larger local datasets compared to Chromem.

Aligns LocalRecall with KiloCode’s new LanceDB support, strengthening ecosystem cohesion.

🙏 Request

Please consider adding LanceDB as a supported backend for LocalRecall. This would give users more flexibility and align LocalRecall with the broader ecosystem.

Start by reading LocalRecall’s existing backend interface and the current Chromem/Chromadb implementation. Trace how backend configuration is selected, then review the proposed insert, query, and delete tests. Done means LanceDB can be selected for short-term and long-term memory and those operations are covered by unit tests.