Crate ragfs_store 

Vector storage layer for RAGFS.

This crate provides storage backends for RAGFS, implementing the VectorStore trait.

Cargo Features

• lancedb (default): Enables the LanceDB backend for production use
• Without lancedb: Only MemoryStore is available (for testing/development)

Backends

Backend	Description
LanceStore	Production backend using LanceDB (requires lancedb feature)
MemoryStore	In-memory backend for testing (always available)

LanceDB Features

When the lancedb feature is enabled:

• Vector Search: Fast approximate nearest neighbor search using HNSW
• Hybrid Search: Combined FTS and vector search for better relevance
• Full CRUD: Create, read, update, delete operations for chunks and files
• Automatic Indexing: Creates vector and FTS indices automatically

Example

use ragfs_store::LanceStore;
use ragfs_core::VectorStore;

// Create and initialize store
let store = LanceStore::new("path/to/db.lance".into(), 384);
store.init().await?;

// Store chunks
store.upsert_chunks(&chunks).await?;

// Search
let results = store.search(query).await?;

Re-exports

pub use lancedb::LanceStore;

pub use memory::MemoryStore;

Modules

lancedb

LanceDB implementation of VectorStore.

memory

In-memory store for testing without LanceDB.

schema

Arrow schema definitions for LanceDB tables.