Support OpenAI Assistant reuse (assistant_id)

[pletinsky]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

agent = OpenAIAssistantV2Runnable.create_assistant(
                assistant_id="e2b_tool"
                name="langchain assistant e2b tool",
                instructions="You are a personal math tutor. Write and run code to answer math questions.",
                tools=tools,
                model="gpt-4-1106-preview",
                as_agent=True
            )

Description

I see the latest version of OpenAIAssistantV2Runnable in the dev build does not support custom assistant_id.
But it's absolutely necessary to support it to reuse existing assistants.
Does anyone know if this feature is expected or not?

System Info

python 3.12, macos

[dosubot[bot]]

Hey @pletinsky! I'm here to help you with any bugs, questions, or contributions you have. Let's tackle this together!

The latest version of OpenAIAssistantV2Runnable does support custom assistant_id for reusing existing assistants. You can specify the assistant_id when creating an assistant. However, there are no specific constraints or validation checks on the assistant_id field, so ensure that the assistant_id you provide is valid.

Here is an example of how to create an assistant using an existing assistant_id:

from langchain_experimental.openai_assistant import OpenAIAssistantRunnable

# Create an assistant using an existing assistant_id
assistant = OpenAIAssistantRunnable(
    assistant_id="your_existing_assistant_id",
    model="gpt-4-1106-preview"
)

# Invoke the assistant with some input
output = assistant.invoke({"content": "What's 10 - 4 raised to the 2.7"})
print(output)

In your example code, you should pass the assistant_id directly to the OpenAIAssistantV2Runnable constructor instead of the create_assistant method, as the latter does not accept assistant_id as a parameter:

agent = OpenAIAssistantV2Runnable(
    assistant_id="e2b_tool",
    name="langchain assistant e2b tool",
    instructions="You are a personal math tutor. Write and run code to answer math questions.",
    tools=tools,
    model="gpt-4-1106-preview",
    as_agent=True
)

This will allow you to reuse an existing assistant with the specified assistant_id ^[1][2].

To continue talking to Dosu, mention @dosu.