UniCraft AI AgentInstall

Summary

Send a chat message to a configurable UniCraft AI agent and receive a response, optionally leveraging conversation memory, tool integrations, and multi-modal input and output.

Inputs

• Agent Type (required) — Select the type of AI agent to create (e.g., Conversational, Tool-Enabled, Smart Routing, Custom).
• Input Message (required) — The user message or prompt to send to the AI agent. If the node input contains fields like message, text, prompt, query, or messages array, the last message is used automatically.
• System Message — Defines the AI agent's behavior and personality via a system prompt.
• Memory Configuration — Configure the conversation memory handling to maintain context with options such as buffer, window, summary, token buffer, or none.
• Model Configuration — Configure the AI model selection mode (Smart Routing or Manual), model choice (if manual), temperature, maximum tokens, and smart routing strategy.
• Multi-Modal Configuration — Configure input and output types for multi-modal processing, including text, images, audio, video, documents, code, or markdown, and upload associated media files as Base64 encoded data.
• Tools Configuration — Enable or disable tool usage for the agent, and define custom tools with names, descriptions, JSON parameter schemas, and API endpoints.
• Advanced Configuration — Advanced options including maximum iterations for tool-enabled agents, execution timeout in seconds, enable streaming responses, and option to return intermediate agent execution steps.

Output shape

A single JSON object representing the agent response, including AI-generated message content, any tool calls made during processing, usage statistics, model info, cost, and metadata about agent type, conversation memory, and multi-modal content if applicable.

The output includes the assistant's reply under 'response.content', any tool calls under 'response.tool_calls', conversation memory configuration if enabled, and optionally media data when multi-modal output is requested. Errors during processing are thrown unless 'continueOnFail' is enabled. The output corresponds to the input item and may include intermediate steps if configured.

Examples

Example 1: Basic conversational AI chat

Set Agent Type to Conversational, supply a user message, and optionally customize system message and model parameters. The agent replies with generated text maintaining conversation memory as configured.

Example 2: Tool-enabled agent with external function calls

Enable Tools, define tool properties and endpoints, select Tool-Enabled Agent type. The agent can call defined tools during chat processing with iteration and timeout limits.

Example 3: Multi-modal input/output

Configure inputType and outputType for multi-modal capabilities such as image or audio generation. Upload media files encoded in Base64 as inputs for processing along with text messages.