v0.4.0 - Support for connecting to PostgreSQL with pgvector and improvements to retrieval quality (this will not be a breaking changes if you are already using Qdrant or Redis)

[FredyRivera-dev]

This issue affects version 0.4.0, where we will add support for vector databases for PostgreSQL with pgvector. This new version will not affect current users using Qdrant or Redis. The quality of RAG results will also be improved.

• PostgreSQL compatibility with pgvector
• Implementation of a re ranker
• Optional support for metadata (JSON) for both send and query
• Configure Reranker from the CLI
• Configure PostgreSQL from the UI-Playground
• Update documentation for all features

To improve the recovery proposed in the comments below, a re ranker will be implemented, and other things.

[FredyRivera-dev]

I was also thinking about adding the connection to supabase with the vecs library, but the asynchronous client pull request hasn't been merged into main yet, and it hasn't been released on pypi. As soon as it's available, we'll make it compatible with Aquiles-RAG :)

[FredyRivera-dev]

Implementation of BaseWrapper and standardization of wrappers to improve future integrations with other vector databases.
It is fully backwards compatible with Qdrant and Redis, you don't need to do anything

Commit: c4dbb10

[FredyRivera-dev]

Implementation of the "CreateIndexMultimodal" and "SendRAGMultimodal" models. They may be subject to change, but the basic idea for the models is there.

Commit: 3e9e2cd

[FredyRivera-dev]

Ok, it's likely this issue will expand a bit more due to some suggestions I received to improve the recovery, based on what was presented here: https://github.com/onestardao/WFGY/blob/main/ProblemMap/README.md

[FredyRivera-dev]

And I hope to be able to work on it as soon as I finish some personal matters.

[FredyRivera-dev]

Hey, quick update, we hope that Vecs will soon support the asynchronous client and therefore we will add compatibility with vecs and supabase :D

Pull requests: supabase/vecs#108

[FredyRivera-dev]

Update of the "create_index" method in the "PostgreSQLRAG" class, as much as possible to make changes, I need to validate that it works

Commit: e1e56c8

[FredyRivera-dev]

Update of 'PostgreSQLRAG' Wrapper

Commit: ff6b672

[FredyRivera-dev]

PostgreSQL has been added to the 'get_connectionAll()' function and added to lifespan (also a bug in lifespan has been fixed and is now on Pypi)

Commit: f9bb1de

[FredyRivera-dev]

The implementation of a multimodal RAG will wait, and Vecs support will be added as soon as the asynchronous client is released. We want to improve the current RAG and expand support to PostgreSQL.