Add PGVector DocumentStore

[mathislucka]

Summary and motivation

PGVector is a popular request from our community. We should have it in 2.0.

Detailed design

TBD

Checklist

If the request is accepted, ensure the following checklist is complete before closing this issue.

• The code was merged in the main branch
• Docs are published at https://docs.haystack.deepset.ai/
• There is a Github workflow running the tests for the integration nightly and at every PR
• A label named like integration:<your integration name> has been added to this repo
• The labeler.yml file has been updated
• The package has been released on PyPI
• An integration tile has been added to https://github.com/deepset-ai/haystack-integrations (Add pgvector haystack-integrations#162)
• The integration has been listed in the Inventory section of this repo README
• There is an example available to demonstrate the feature
• The feature was announced through social media

[sahusiddharth]

@mathislucka I have started working on this, can you clarify something for me?

When it comes to metadata filtering there is way suggested in the template, but the one followed in chroma docstore is quite different, how should I proceed from here?

There is one more thing that I wanted to ask,

Context: We have different methods in different document stores, they have 4 in common [count_document, filter_document, write_document, delete_document] and there are some just to support them , but apart from them there are some extras.

Question: Do we have to implement those lacking functions in all the document store? If you could give me a list of all the function that will be required based on our current requirements.

[anakin87]

Hey @sahusiddharth, sorry for the late reply.

I recently developed the Pinecone Document Store, so I try to share some suggestions.

• the official metadata filtering is that of the template. You can find more information in the docs.
  I think filters in Chroma Document Store are the legacy filters.

• as explained in Creating Custom Document Stores, each Document Store must respect the Document Store protocol, by implementing the 4 standard methods (respecting their signature). For functions that are specific to your Document Store, you are free to implement any other methods you need.

• incremental development: I suggest you create one of the 4 standard methods at a time and test it. DocumentStoreBaseTests was recently refactored to allow this.
  An example

  1. you implement count_documents
  2. in test_document_store.py, you define the test class TestDocumentStore(CountDocumentsTest): please note that here we are only inheriting from CountDocumentsTest
  3. you make tests pass
  4. you implement write_documents
  5. in test_document_store.py, you define the testing class TestDocumentStore(CountDocumentsTest, WriteDocumentsTest): here we are also inheriting from WriteDocumentsTest
  6. ...

• Elasticsearch Document Store can be considered a good reference, also for the implementation of filters

Feel free to ask for clarification, if needed!