search

phidatahq / phidata

Build AI Agents with memory, knowledge, tools and reasoning. Chat with them using a beautiful Agent UI.
https://docs.phidata.com
Mozilla Public License 2.0
15.17k stars 2.1k forks source link

[New Embedder Request]: Add Qdrant FastEmbed as an Embedder Option #1379

Closed sashank5644 closed 1 week ago

sashank5644 commented 2 weeks ago

Qdrant FastEmbed is a lightweight, efficient embedding library optimized for fast inference and limited dependencies. Some if its advantages include:

  1. High-Speed Embedding Generation: (50% faster) than traditional PyTorch-based transformers, making it ideal for real-time applications.
  2. Resource Efficiency: With model quantization and ONNX Runtime integration, FastEmbed minimizes memory and CPU use.
  3. Enhanced Retrieval with Prefix-Based Embeddings: FastEmbed supports specialized input prefixes like "query" and "Passage" which allows model to generate differentiated embeddings for queries and documents, optimizing retrieval accuracy and semantic search.

I though it might be a great addition to the existing embeddings, and if it seems like a fit, I would love to take this integration task.

manthanguptaa commented 2 weeks ago

Hey @sashank5644, Qdrant FastEmbed sounds like a really good addition. Go for it! I am excited to look at your PR.

manthanguptaa commented 1 week ago

Thank you @sashank5644 for the PR! It's merged now.