feat: use haystack for the RAG backend #335

New Issue

2025-11-11T14:17:57-06:00

GiteaMirror commented

2025-11-11 14:17:57 -06:00

Originally created by @Jbollenbacher on GitHub (Feb 22, 2024).

I'd like to suggest using haystack as the backend for the RAG system. It's quickly becoming the industry standard and is well maintained. I figured now would be a good time to suggest it while the RAG system is still mostly in beta.

Haystack solves a lot of the basic problems like providing flexibility for swapping out embedding models and chunking strategies. It's also robust and comes with a lot of good default options. Basically it offloads the task of engineering a SOTA RAG system to another group of open source developers.

Originally created by @Jbollenbacher on GitHub (Feb 22, 2024). I'd like to suggest using [haystack](https://github.com/deepset-ai/haystack) as the backend for the RAG system. It's quickly becoming the industry standard and is well maintained. I figured now would be a good time to suggest it while the RAG system is still mostly in beta. Haystack solves a lot of the basic problems like providing flexibility for swapping out embedding models and chunking strategies. It's also robust and comes with a lot of good default options. Basically it offloads the task of engineering a SOTA RAG system to another group of open source developers.

GiteaMirror closed this issue

2025-11-11 14:17:58 -06:00

GiteaMirror commented

2025-11-11 14:17:58 -06:00