Access chat History from a tool

[darthShana]

I have a tool that does search from an inventory of products/services available in a store
however instead of the agent having to create that query for me
I wanted to get better performance by taking the entire conversation . So my tool looks like

class RecommendServiceInput(BaseModel):
    chat_history: str = Field("the chat history between an ai and a visitor")

def recommend_service(chat_history: str):

My tool can then run a customized few shot template to extract the query from.. and them run it
Kinda like self query but wanting control over how the query is extracted from the conversation history .

However: The above approach is inconsistent.

• Some time the chat history is a sting, sometimes its a list
• sometimes it contains just the human messages, some tiem human and ai messages
• sometime it doesnt have the most recent human hemmage some times it does

Is there a more reliable way for me to access the chat history inside a tool?
Or should i be doing this outside a tool and have it ready and a separate variable?

Any guidance would be greatly appreciated..

[hinthornw]

Does this help? https://langchain-ai.github.io/langgraph/how-tos/pass-run-time-values-to-tools/

I think it will depend on the next release 0.0.10 to be done today

[darthShana]

Yus !!!
love it