AutoChain

Large language models (LLMs) have shown huge success in different text generation tasks and enable developers to build generative agents based on natural language objectives.

However, most of the generative agents require heavy customization for specific purposes, and adapting to different use cases is sometimes overwhelming using existing tools and framework. As a result, it is still very challenging to build a customized generative agent.

In addition, evaluating such agents powered by LLMs by trying different use cases under different potential user scenarios is a very manual and expensive task.

AutoChain takes inspiration from LangChain and AutoGPT and aims to solve both problems by providing a lightweight and extensible framework for developers to build their own conversational agents using LLMs with custom tools and automatically evaluating different user scenarios with simulated conversations. Experienced user of LangChain would find AutoChain is easy to navigate since they share similar but simpler concepts.

The goal is to enable quick user experiments of generative agents, knowing users would make more customizations as they are building their own agent.

If you have any question, please feel free to reach out to Yi Lu yi.lu@forethought.ai

Features

• 🚀 lightweight and extensible generative agent pipeline.
• 🔗 agent that can use different custom tools and support OpenAI function calling
• 💾 simple memory tracking for conversation history and tools' outputs
• 🤖 automated agent multi-turn conversation evaluation with simulated conversations

Setup

Quick install

pip install autochain

Or install from source after cloning the repo

cd autochain
pyenv virtualenv 3.10.11 venv
pyenv local venv

pip install .

Set PYTHONPATH and OPENAI_API_KEY

export OPENAI_API_KEY=
export PYTHONPATH=`pwd`

Run your first conversation with agent interactively

python autochain/workflows_evaluation/conversational_agent_eval/change_shipping_address_test.py -i

Example usage

If you have experiences with LangChain, you already know 80% of the AutoChain interfaces.

AutoChain aims to make creating a new customized agent very straight forward with as few concepts as possible.
Read about more example usages.

The most basic example uses default chain and ConversationalAgent:

from autochain.chain.chain import Chain
from autochain.memory.buffer_memory import BufferMemory
from autochain.models.chat_openai import ChatOpenAI
from autochain.agent.conversational_agent.conversational_agent import ConversationalAgent

llm = ChatOpenAI(temperature=0)
memory = BufferMemory()
agent = ConversationalAgent.from_llm_and_tools(llm=llm)
chain = Chain(agent=agent, memory=memory)

print(chain.run("Write me a poem about AI")['message'])

User could add a list of tools to the agent similar to LangChain

tools = [Tool(
    name="Get weather",
    func=lambda *args, **kwargs: "Today is a sunny day",
    description="""This function returns the weather information"""
)]

memory = BufferMemory()
agent = ConversationalAgent.from_llm_and_tools(llm=llm, tools=tools)
chain = Chain(agent=agent, memory=memory)
print(chain.run("What is the weather today")['message'])

AutoChain also added supports for function calling for OpenAI model. It extrapolates the function spec in OpenAI format without user explicit instruction, so user could follow the same Tool interface.

llm = ChatOpenAI(temperature=0)
agent = OpenAIFunctionsAgent.from_llm_and_tools(llm=llm, tools=tools)

Check out more examples under autochain/examples and workflow evaluation test cases which can also be run interactively.

How does AutoChain simplify building agents?

AutoChain aims to provide a lightweight framework and simplifies the building process a few ways comparing with other existing frameworks

1. Easy prompt update
   Prompt engineering and iterations is one of the most important part of building generative agent. AutoChain makes is very obvious and easy to update prompts and visualize prompt outputs. Run with -v flag to output verbose prompt and outputs in console.
2. Up to 2 layers of abstraction
   Since this goal of AutoChain is enabling quick iterations, it chooses to remove most of the abstraction layers from alternative frameworks and make it easy to follow
3. Automated multi-turn evaluation
   The most painful and uncertain part of building generative agent is how to evaluate its performance. Any change for one scenario could cause regression in other use cases. AutoChain provides an easy test framework to automatically evaluate agent's ability under different user scenarios.

Read more about detailed components overview

Workflow Evaluation

It is notoriously hard to evaluate generative agents in LangChain or AutoGPT. An agent's behavior is nondeterministic and susceptible to small changes to the prompt or model. It is very hard to know if agent is behaving correctly under different scenarios. The current path for evaluation is running the agent through a large number of preset queries and evaluate the generated responses. However, that is limited to single turn conversation, general and not specific to tasks and expensive to verify.

To effectively evaluate agents, AutoChain introduces the workflow evaluation framework which simulates the conversation between an generative agent and simulated users with LLM under different user contexts and desired outcomes of the conversation. This way, we could add test cases for different user scenarios and use LLMs to evaluate if multi-turn conversations reached the desired outcome.

Read more about our evaluation strategy.

How to run workflow evaluations

There are two modes for running workflow tests. Interactively or running all test cases. For example in autochain/workflows_evaluation/conversational_agent_eval /change_shipping_address_test.py, there are already a few example test cases.

Running all the test cases defined in the test:

python autochain/workflows_evaluation/conversational_agent_eval/change_shipping_address_test.py

You can also have an interactive conversation with agent by passing the interactive flag -i:

python autochain/workflows_evaluation/conversational_agent_eval/change_shipping_address_test.py -i

More explanations for how AutoChain works? checkout components overview