Agent Execution Steps while using Langflow API

[raprajapati1208]

Hi community,

I am building a flow which has agents having access to sub-agents which use different tools to execute its task. I will be using the API to run the flow.
I want my end-users to see what actions/steps the agent takes to complete the task just like we can see in the Playground. For example, Steps like Executing Agent_1, Executing Tool 1. Is this possible?

Regards,
Ranjan

[dosubot[bot]]

Hello, @raprajapati1208! I'm here to help you with any issues or questions you have. Just let me know how I can assist you!

Langflow's API does provide a built-in feature to track and expose the execution steps of agents and sub-agents. This is achieved through the AgentAsyncHandler class, which logs various stages of execution, such as on_chain_start, on_tool_start, on_tool_end, on_agent_action, and on_agent_finish. These methods use a log_function to record details about each step, allowing you to track and log the execution process comprehensively ^[1].

Additionally, Langflow's API handles agent execution and logging of actions by emitting tokenized responses through a socket service. It logs the action's thought process and sends it as a stream of tokens, which can be displayed to end-users ^[2].

For more detailed logging or tracing, you can integrate Langflow with Langfuse, which provides detailed production traces and insights into quality, cost, and latency ^[3].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other}