Hi everyone,
I'm currently using an AI agent node within callin.io that utilizes a Postgres database for its memory. This agent has the capability to use tools, such as a "contact buyer" node, which facilitates sending messages to users (like buyers) throughout the workflow.
Problem:
The "contact buyer" node action, which is an intermediate step, is not being logged in the Postgres memory. Consequently, the agent is unable to recall previously sent messages.
Inquiry:
Is there a suggested method for storing these intermediate tool usages (e.g., messages sent to users) in the agent's memory (Postgres database)? This would enable the agent to refer back to them later in the workflow.
Has anyone else set up a similar system or have advice on best practices for logging tool usage in agent memory within callin.io?
Thanks!
Hey there,
Could you please share your workflow? This will help us provide more effective suggestions on what to add and where.
You can utilize Langsmith for tracking the AI processes within callin.io.
Refer to this link for detailed instructions on integrating callin.io with Langsmith. You'll find step-by-step guidance on how to implement Langsmith into your callin.io workflows.
Here is the workflow. Essentially, the model utilizes the think tool and the respond to user tool. However, at the end, none of the tool calls are recorded in the memory. (Refer to the image below, where the prompt was: “what’s the meaning of life, use the think tool and then send the user a message”)
Thanks. However, even in Nate’s video, the tool calls are not being saved in Supabase. In fact, the tool_calls
key is empty (at 9:40) when he utilized the Supabase tool.
If you want to see all the steps the model is performing, enable Return Intermediate Steps
docs