{agent_scratchpad} with create_react_agent ?

[mrctito]

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

config={"configurable": {"thread_id": "1"}}
result = await react_agent.ainvoke(inputs, config=config, debug=True)

Description

Hello @dosu,

Should I use the {agent_scratchpad} key with create_react_agent?

If so, how?

I am not sure how to pass variables to the ainvoke method of create_react_agent:

config={"configurable": {"thread_id": "1"}}
result = await react_agent.ainvoke(inputs, config=config, debug=True)

Thank you.

System Info

langchain==0.2.6
langchain-community==0.2.6
langchain-core==0.2.10
langchain-experimental==0.0.62
langchain-openai==0.1.13
langchain-qdrant==0.1.0
langchain-text-splitters==0.2.1

[dosubot[bot]]

Yes, you should use the {agent_scratchpad} key with create_react_agent. The prompt must include the agent_scratchpad key to contain previous agent actions and tool outputs as a string.

To pass variables to the ainvoke method of create_react_agent, you need to include the required input keys in the dictionary you pass to the invoke method. Here is an example:

from langchain import hub
from langchain_community.llms import OpenAI
from langchain.agents import AgentExecutor, create_react_agent
from langchain_core.prompts import PromptTemplate

# Define the prompt template
template = '''Answer the following questions as best you can. You have access to the following tools:

{tools}

Use the following format:

Question: the input question you must answer
Thought: you should always think about what to do
Action: the action to take, should be one of [{tool_names}]
Action Input: the input to the action
Observation: the result of the action
... (this Thought/Action/Action Input/Observation can repeat N times)
Thought: I now know the final answer
Final Answer: the final answer to the original input question

Begin!

Question: {input}
Thought:{agent_scratchpad}'''

prompt = PromptTemplate.from_template(template)
model = OpenAI()
tools = [...]  # Define your tools here

# Create the agent
agent = create_react_agent(model, tools, prompt)
agent_executor = AgentExecutor(agent=agent, tools=tools)

# Invoke the agent with the required variables
result = await agent_executor.ainvoke({
    "input": "hi",
    "agent_scratchpad": "",  # Initial empty scratchpad
    "tools": "tool descriptions here",
    "tool_names": "tool1, tool2"
}, config={"configurable": {"thread_id": "1"}}, debug=True)

print(result)

In this example, the ainvoke method is called with a dictionary containing the input, agent_scratchpad, tools, and tool_names keys. The agent_scratchpad is initially empty but will be populated with previous actions and observations as the agent operates ^[1][2][3][4].

To continue talking to Dosu, mention @dosu.