PGVector filtering improvements to support self-querying

[djo10]

Checked other resources

• I added a very descriptive title to this issue.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.
• I am sure that this is a bug in LangChain rather than my code.

Example Code

The current code does not handle top-level conditions like the example below:

filter = {"or": [{"rating": {"gte": 8.5}}, {"genre": "animated"}]}
retriever = vectorstore.as_retriever(search_kwargs={"filter": filter})

The current implementation for this example will not find results.
The current implementation does not support GTE and LTE comparators in pgvector.

Error Message and Stack Trace (if applicable)

No response

Description

• I am trying to filter documents based on metadata, including top level conditions with 'and', 'or' operators
• I am trying to create more comprehensive metadata filtering with pgvector and to create the base for pgvector self querying.
• I am trying to use GTE and LTE comparators in filter clause
• Code for pgvector can be refactored, using already defined comparators and operators from langchain.chains.query_constructor.ir

System Info

langchain==0.1.5
langchain-community==0.0.17
langchain-core==0.1.18

Python 3.10.12

[dosubot]

🤖

It's great to see your interest in improving the PGVector filtering in LangChain! Your proposed improvements sound promising and could be very valuable for the community. If you're able to implement these changes and create a pull request, it would be a fantastic contribution to the project. Keep up the good work!

---

This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

If you want to continue the conversation, start your reply with @dosu-bot.