Partners LLM: First Meta's Llama API integration

[seyeong-han]

Add Llama API support

This PR adds support for the Llama API through a new ChatLlama model class.

Key features:

• Built upon BaseChatOpenAI class using Llama's OpenAI-compatible endpoint (https://api.llama.com/compat/v1/)
• Supports chat completion and streaming
• Implements structured output via the "json_schema" method

Current limitations:

• Tool calling is not yet stable in the Llama API, so it's not supported in this implementation
• Modified unit tests to only test the "json_schema" structured output format

We plan to add support for tool calling in a future update once the API stabilizes.

[vercel]

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

1 Skipped Deployment

Name	Status	Preview	Comments	Updated (UTC)
langchain	⬜️ Ignored (Inspect)	Visit Preview		May 5, 2025 7:36am

[seyeong-han]

Hi @ccurme !

I would like to know why this PR hasn't been merged yet.
Is there anything else we need to do?

[aviadr1]

if this PR works, please merge it 🙏

[ccurme]

Hi @seyeong-han, we haven't had the bandwidth to prioritize reviewing, setting up CI for releases + integration tests, etc. (this is the reason I suggested managing this in a repo you control, as we do in langchain-google, langchain-aws, and others).

I will take a look this week.

[aviadr1]

I would like to know why this PR hasn't been merged yet. Is there anything else we need to do?

hey @seyeong-han ! I am not a maintainer here, just a person who also wants to use this.
however without language_calling structured outputs I cant use it ... my pydantic classes arent supported as json_schema ...

[seyeong-han]

I would like to know why this PR hasn't been merged yet. Is there anything else we need to do?

hey @seyeong-han ! I am not a maintainer here, just a person who also wants to use this. however without language_calling structured outputs I cant use it ... my pydantic classes arent supported as json_schema ...

Hi Aviadr1,

Would it be possible to share your example code about the issue? Thank you for your interest!

[ccurme]

@seyeong-han I've made some minor adjustments and merged in your implementation here: https://github.com/langchain-ai/langchain-meta