langchain[minor]: Make EmbeddingsFilters async

[pprados]

Thank you for contributing to LangChain!

• PR title: community: EmbeddingsFilters is not compatible with async

• PR message:

  • Description: EmbeddingsFilters is not compatible with async
  • Issue: Current implementation call sync methods
  • Twitter handle: pprados

• Add tests and docs: Test were updated

• Lint and test: Run make format, make lint and make test from the root of the package(s) you've modified. See contribution guidelines for more: https://python.langchain.com/docs/contributing/

@eyurtsev, another review ?

[vercel]

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

1 Ignored Deployment

Name	Status	Preview	Comments	Updated (UTC)
langchain	⬜️ Ignored (Inspect)	Visit Preview		Jun 12, 2024 10:50am

[pprados]

@eyurtsev
Once again, an async bug

[eyurtsev]

Looks good -- could you undo changes in pyproject.toml and poetry.lock -- most PRs should leave those unomdified

[pprados]

Hello @eyurtsev

In test_embeddings_filter.py, I change
from langchain_community.embeddings import OpenAIEmbeddings
to
from langchain_openai.embeddings import OpenAIEmbeddings

But, for that, I need to add a dependency in pyproject.tml

[tool.poetry.group.test_integration.dependencies]
...
langchain-openai = ">=0.1.8"

But, the poetry lock --no-update return:

Because langchain-openai (0.1.8) depends on tiktoken (>=0.7,<1)
 and no versions of langchain-openai match >0.1.8, langchain-openai (>=0.1.8) requires tiktoken (>=0.7,<1).
So, because langchain-community depends on both tiktoken (>=0.3.2,<0.6.0) and langchain-openai (>=0.1.8), version solving failed.

Do you think I should keep the original import?

[eyurtsev]

Hello @eyurtsev

In test_embeddings_filter.py, I change from langchain_community.embeddings import OpenAIEmbeddings to from langchain_openai.embeddings import OpenAIEmbeddings

But, for that, I need to add a dependency in pyproject.tml

[tool.poetry.group.test_integration.dependencies]
...
langchain-openai = ">=0.1.8"

But, the poetry lock --no-update return:

Because langchain-openai (0.1.8) depends on tiktoken (>=0.7,<1)
 and no versions of langchain-openai match >0.1.8, langchain-openai (>=0.1.8) requires tiktoken (>=0.7,<1).
So, because langchain-community depends on both tiktoken (>=0.3.2,<0.6.0) and langchain-openai (>=0.1.8), version solving failed.

Do you think I should keep the original import?

Yes, please keep the original import.

If you want to spend more effort, you could fix the test logic itself (not required for this PR) -- the test should not depend on the embedding provider, and instead can leverage fake embeddings in langchain_core. If that's done, then the test can be moved into unit tests.

[pprados]

@eyurtsev
"The Integration docs lint" checks is blocked in all the worlflow.

[eyurtsev]

Could you revert changes in pyproject toml and poetry lock please? Most PRs should not be modifying these files