Crate ragfs_query 

ragfs-query

Query parsing and execution for RAGFS semantic search.

This crate handles the query pipeline: parsing natural language queries, generating query embeddings, and executing similarity searches against the vector store.

Features

• Natural language queries: Search with plain text like “authentication logic”
• Semantic matching: Uses vector similarity (cosine distance)
• Hybrid search: Combine vector similarity with full-text search
• Filtering: Narrow results by path, MIME type, or date range

Usage

use ragfs_query::QueryExecutor;
use ragfs_core::{VectorStore, Embedder};
use std::sync::Arc;

// Create the query executor
let executor = QueryExecutor::new(
    store,      // Arc<dyn VectorStore>
    embedder,   // Arc<dyn Embedder>
    10,         // Default result limit
    true,       // Enable hybrid search
);

// Execute a semantic query
let results = executor.execute("error handling implementation").await?;

for result in results {
    println!("{}: {:.3}", result.file_path.display(), result.score);
    println!("  {}", result.content);
}

Query Pipeline

1. Parse: Extract query text and optional filters
2. Embed: Convert query text to a 384-dimensional vector
3. Search: Find similar chunks using ANN (approximate nearest neighbors)
4. Rank: Order results by similarity score
5. Return: Provide results with content, scores, and locations

Hybrid Search

When enabled, combines:

• Vector similarity: Semantic matching via embeddings
• Full-text search: Keyword matching for precision

Results are fused using reciprocal rank fusion (RRF).

Components

Type	Description
QueryExecutor	Executes semantic queries against the index
QueryParser	Parses query strings with optional filters

Re-exports

pub use executor::QueryExecutor;

pub use parser::QueryParser;

Modules

executor

Query execution.

parser

Query DSL parser.