Lag between providers, docs and Python Langchain package releases

OSS Product Help LangChain
1

I am setting up an graph that has a node that is using create_agent. Because reasons, I need to have OpenRouter for local development and AWS Bedrock for production use. So I see langchain-aws and langchain-openrouter exist. Additionally the docs for init_chat_model state the provider names to use. However, the list of allowed providers in the current langchain package, 1.2.10, does not have them. The master branch does, but not the release.

So is it normal for the langgraph-* packages to get updated, which in turn update the LangChain documentation site, but not get the LangGraph release that supports that, or is this just something that happened specifically with the current pending release? Is there another release coming in the next day or so to get the updated providers list?

2

Short answer: yes — the docs site (which reflects main/master) can show new provider integrations before those changes are published in a stable release, so seeing langchain-openrouter / langchain-aws on the docs but not in langchain 1.2.10 is expected.

The docs are generated from the docs repo’s main branch and aggregate provider packages (many of which are released independently). Provider support is often added by merging provider packages or docs updates to main firstt, the published PyPI/conda/npm release for langchain/langgraph (and their pinned provider lists) may lag until a coordinated release or patch is cut. The release policy and changelog show frequent minor/patch releases, so this kind of short-term mismatch is normal while a release is pending.

# Quick checks you can run locally
python -c "import langchain_core; print('langchain-core', langchain_core.__version__)"
python -c "import langgraph; print('langgraph', langgraph.__version__)"
# Check PyPI versions
pip index versions langchain | sed -n '1,5p'

Workarounds & verification steps

  • Install the provider packages directly (they’re released separately) and use the provider-specific integration rather than waiting for a consolidated langchain release:

    pip install langchain-openrouter langchain-aws

    Then follow the provider’s docs (the provider overview links show the integrations available). Many providers also expose provider-specific constructors or accept provider-prefixed model strings.

  • If you need the absolute latest behavior from docs/main right now, install from the repo main branch (temporary, for development):

    pip install git+https://github.com/langchain-ai/langchain.git@main

    or for langgraph:

    pip install git+https://github.com/langchain-ai/langgraph.git@main

  • Track the official release/changelog and GitHub releases to know when the provider list is included in the next stable release.

Relevant docs:

1 Like
3

Thanks for confirming your experiences about the docs and the note about the ability to install from github. I did not know if LangChain had any libraries and such. I will give it a shot.

1 Like
4

@slackbits Sounds good. If it helps, can you mark the above as solution.

5

Yeah, seeing as how there is no official response from the LangChain folks, I will mark it. Although it would have been nice for them to clarify as to why there is published documentation that does not work when using the published libraries. (For which the same docs’ getting started says to install the published versions)

1 Like