embedding-models

What problem does it solve?

Embedding models selection, tuning, and debugging to ensure reliable vector representations for downstream retrieval and similarity tasks, including handling dimension mismatches, ONNX backend setup, and query/passage prefixes.

Core Features & Use Cases

• Model cataloging and comparison for multilingual and Russian-focused contexts.
• Prefix handling for query: and passage: prompts to ensure consistent embedding generation.
• Backend configuration guidance (e.g., ONNX) and diagnostics to resolve dimension mismatches.
• Diagnostic workflow to evaluate model quality, speed, and compatibility across several embedding providers.

Quick Start

Configure EMBEDDING__MODEL to your chosen embedding model and run a quick validation pass to verify dimensions and query/passage prefixes.