Chutes AI Agent icon

Chutes AI Agent

AI Agent that works exclusively with Chutes Chat Model. Generates an action plan and executes it. Can use external tools.

Chutes AI Agent

AI-generated

Summary

Process input prompts through a connected Chutes Chat Model to generate AI-driven responses, optionally using external tools, memory, and output parsers to enhance the interaction.

Inputs

  • promptType (required) — Choose the source for the user's prompt either automatically from a connected Chat Trigger node input field named 'chatInput' (default) or manually define a static or expression-based prompt.
  • text (required) — The actual prompt text sent to the AI agent. Defaults to reading 'chatInput' from the incoming data when promptType is 'auto', or a user-defined string when promptType is 'define'.
  • hasOutputParser — Enable if you require the output to be formatted according to a connected output parser node.
  • options — Optional parameters including system message instructions to the AI, maximum agent iteration count, and whether to return intermediate reasoning steps.

Output shape

a list of JSON objects each containing the AI agent's output response, the prompt used, and optionally intermediate reasoning steps

The output includes the final AI-generated response under 'output', the prompt under 'prompt', and if enabled, 'intermediateSteps' detailing each tool call and observation. Errors during processing are returned per item if 'Continue On Fail' is set in the node.

Examples

Example 1: Simple prompt to the AI without tools or memory connected.

Set promptType to 'auto' to use connected chat trigger input, or 'define' to enter a static prompt text. Connect a Chutes Chat Model node to the 'Chat Model' input. Leave tools, memory, and output parser inputs unconnected. Optionally customize systemMessage and maxIterations in options.

Example 2: Use external tools to augment AI agent capabilities.

Connect one or more AiTool nodes implementing specific functions to the 'Tool' input. The node will call these tools as directed by the AI's planning, iterating up to maxIterations. Optionally enable 'Return Intermediate Steps' to debug agent decisions.

Example 3: Maintain conversational context with memory.

Connect an AiMemory node to supply and save conversation history. The agent prepends memory history to each prompt, allowing context-aware interaction during the session.

Links

Discussion