OpenAIFunctionsAgent vs OpenAIMultiFunctionsAgent

[aleksanderhan]

It's not really clear to me whats the difference. What makes them different?

[dosubot[bot]]

Answer generated by a 🤖

Answer

Hello,

The OpenAIMultiFunctionsAgent and OpenAIFunctionsAgent are both classes in the LangChain codebase that are designed to interact with language models using OpenAI's function powered API. However, based on the context provided, it seems that the OpenAIFunctionsAgent class is not defined in the repository, so I can't provide specific details about it.

The OpenAIMultiFunctionsAgent class, on the other hand, is a multi-action agent that has several specific functionalities. Here are some of them:

1. get_allowed_tools: This method returns a list of the names of the tools that the agent has access to.

2. validate_llm: This is a root validator that checks if the language model is an instance of ChatOpenAI. If not, it raises a ValueError.

3. validate_prompt: This is another root validator that checks if agent_scratchpad is one of the variables in the prompt. If not, it raises a ValueError.

4. input_keys: This property returns a list of input keys. In this case, it returns ["input"].

5. functions: This property returns a list of tools formatted for OpenAI functions. It also includes a custom tool that bundles the action name and the action because some actions could have the same schema, and without this, there is no way to differentiate them.

6. plan and aplan: These methods take in intermediate steps, callbacks, and other user inputs, and return an action specifying what tool to use. They also format the intermediate steps into a list of messages to send to the language model for the next prediction.

7. create_prompt: This class method creates a prompt for the agent.

8. from_llm_and_tools: This class method constructs an agent from a language model and tools.

You can find the code for the OpenAIMultiFunctionsAgent class in the libs/langchain/langchain/agents/openai_functions_multi_agent/base.py file.

As for the specific use cases, the OpenAIMultiFunctionsAgent can handle tool invocation, message parsing, message conversion, message formatting, prompt creation, agent construction, and early stopping scenarios. You can find more details about these use cases in the context provided above.

I hope this helps! If you have any more questions or need further clarification, feel free to ask.

Sources

• libs/langchain/langchain/agents/openai_functions_agent/base.py
• libs/langchain/langchain/agents/openai_functions_multi_agent/base.py
• libs/langchain/langchain/agents/agent.py

---

This response is meant to be useful, save you time, and share context. 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.