Adding documentation for PGVectorStore

[dishaprakash]

Thank you for contributing to LangChain!

• Adding documentation for PGVectorStore:
  docs: Adding documentation for the new PGVectorStore as a part of langchain-postgres

• Add docs: The notebook for PGVectorStore is now added to the directory docs/docs/integrations.
  As a part of this change, we've also updated the VectorStore features table and VectorStoreTabs

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
langchain	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	May 5, 2025 2:09pm

[ccurme]

Seeing a few errors in build.

One is straightforward: we require a standard set of headers in integrations pages:

ValueError: Document /home/runner/work/langchain/langchain/docs/docs/integrations/vectorstores/pgvectorstore.ipynb is missing headers:
- ## Setup
- ## Initialization
- ## Manage vector store
- ## Query vector store
- ## Usage for retrieval-augmented generation
- ## API reference

---

Separate from this, it's possible that we're not escaping a comparison operator correctly:

Error: MDX compilation failed for file "/vercel/path0/docs/docs/integrations/vectorstores/pgvectorstore.md"
Cause: Unexpected character `)` (U+0029) before name, expected a character that can start a name, such as a letter, `$`, or `_`
Details:
{
  "column": 27,
  "message": "Unexpected character `)` (U+0029) before name, expected a character that can start a name, such as a letter, `$`, or `_`",
  "line": 327,
  "name": "327:27",
  "place": {
    "line": 327,
    "column": 27,
    "offset": 10428,
    "_index": 0,
    "_bufferIndex": 12
  },
  "reason": "Unexpected character `)` (U+0029) before name, expected a character that can start a name, such as a letter, `$`, or `_`",
  "ruleId": "unexpected-character",
  "source": "micromark-extension-mdx-jsx",
  "url": "https://github.com/micromark/micromark-extension-mdx-jsx#unexpected-character-at-expected-expect"
}
error Command failed with exit code 1.

Found this if helpful.

[dishaprakash]

@ccurme thanks for the help, escaping the comparison operator did the trick.
The PR is now ready for review.

[ccurme]

It will be confusing to users to have PGVector and PGVectorStore side by side. When should they use one vs. the other? Are we going to deprecate PGVector? Should we remove it from the docs (adding redirects to the v0.2 docs)?

[dishaprakash]

@ccurme
PGVectorStore is a 2.0 version of the langchain-postgres vectorstore, which is newly released and more information about it can be found in the readme here.
We will be deprecating PGVector, but we plan on supporting it for the time being so the users can migrate over to PGVectorStore.

We can update the docs to redirect to v2 once the support for PGVector will be stopped.
@averikitsch what would you suggest?

[averikitsch]

We are working with Eugene to deprecate PGVector; however, there are a few additional features Eugene is requesting before the deprecation.
To reduce confusion, it might be good to list both of them and add a deprecated label or warning to PGVector, so developers coming to the docs from other examples have the full context. Ultimately, we should remove or redirect docs to PGVectorStore because it uses recommend Postgres practices for better production use and enterprise scaling.

[eyurtsev]

@averikitsch where there any remaining features that were not on parity with PGVector? I think you added db agnostic filtering already?

[averikitsch]

@averikitsch where there any remaining features that were not on parity with PGVector? I think you added db agnostic filtering already?

That is correct. We don't current support nested JSON metadata, but overall we have filtering support. There are no other missing features.