Google Gemini

[ktibbs9417]

Feature request

With Gemini Pro going GA today (Dec. 13th). When can users of LangChain expect an update to use the new LLM?

Motivation

This will allow users of LangChain to use the latest LLM that Google is providing along with their safety settings.

Your contribution

I can try and help. Happy to contribute where needed

[dosubot[bot]]

🤖

That's great to hear that you're interested in contributing to LangChain! Your willingness to help is much appreciated. If you're able to work on integrating the new LLM from Google Gemini into LangChain, that would be a valuable contribution to the project. Thank you for your support!

---

This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

If you want to continue the conversation, start your reply with @dosu-bot.

[jamesbraza]

@ktibbs9417 check this: https://github.com/langchain-ai/langchain/pull/14621

You can run pip install langchain-google-genai to get started with this, note it was just added about 15 mins after you made this issue haha

[baskaryan]

Like @jamesbraza said, should be able to do this with langchain-google-genai. Let us know if you see any issues! Closing GH issue for now

[dheerajiiitv]

You can also use ChatVertexAI with model_name gemini-pro

[gpy1234]

@dheerajiiitv Did you try it with ChatVertexAI? I tried to use it with ConversationalRetrievalChain.from_llm, but it's not working.

[dheerajiiitv]

No, I am using System message. I can't use ChatVertexAi, you have to use VertexAI

On Sat, Dec 16, 2023, 4:48 PM gpy1234 @.***> wrote:

@dheerajiiitv https://github.com/dheerajiiitv Did you try it with ChatVertexAI? I tried to use it with ConversationalRetrievalChain.from_llm, but it's not working.

— Reply to this email directly, view it on GitHub https://github.com/langchain-ai/langchain/issues/14671#issuecomment-1858793110, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFY7PMHXXFGG2CE7VW537WDYJV7PZAVCNFSM6AAAAABATU7VH6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNJYG44TGMJRGA . You are receiving this because you were mentioned.Message ID: @.***>

[JoAmps]

You can also use ChatVertexAI with model_name gemini-pro

I tried it, but it doesnt stream when incorporated into retrieval chain, it has streaming in its parameters but it doesnt stream, does it stream for you?

[pseudotensor]

@JoAmps Seems still that way ever since the langchain google package was updated. It got better for exception handling, but lost streaming once went to:

langchain-google-genai==0.0.6
google-generativeai==0.3.2

[pseudotensor]

https://github.com/langchain-ai/langchain/pull/15459

https://github.com/h2oai/h2ogpt/commit/972d0b69813343290e219302e25f955bf482105c

[dharless-eli-lilly]

I'm confused because I see most things related to this are closed, but not merged. What's the recommended route here? Should we abandon LLMChain since it's marked deprecated and use the new notation? Langchain is all pretty new to me so would love a little bit of clear direction.