GigaChat AI icon

GigaChat AI

Chat with GigaChat AI models with full memory persistence and tool support

GigaChat AI

AI-generated

Summary

Send a user prompt to a selected GigaChat AI model and receive a conversational AI response with optional memory context and tool invocation support.

Inputs

  • modelId (required) — Select the GigaChat AI model and version to generate the response (loaded dynamically).
  • prompt (required) — Input text prompt or message to send to the AI assistant.
  • options.functionCall — Controls how the AI model handles function/tool calls: 'auto' lets the model decide, 'none' disables function calls.
  • options.maxIterations — Maximum number of consecutive tool invocations the AI model may perform to prevent infinite loops; defaults to 5.
  • options.maxTokensToSample — Limits the maximum tokens generated by the model in the response.
  • options.repetitionPenalty — Penalty factor to control word repetition in the AI output; 1.0 is neutral.
  • options.systemMessage — Optional system prompt defining the assistant's role and instructions.
  • options.temperature — Controls randomness of the response; higher values (0-2) produce more random output.
  • options.topP — Nucleus sampling parameter defining probability mass of tokens considered by the model (0-1).
  • simplifyOutput — If true, returns only the AI's text response; otherwise includes metadata such as token usage and iterations.
  • removeMarkdown — If true, strips Markdown formatting from the AI's text response.

Output shape

a list of JSON objects, one per input item, each containing the AI's response text and metadata

Each output item includes 'response' (string) with the final AI text output, 'model' used, 'usage' statistics with token counts, 'sessionId' for continuity, and 'iterations' of tool calls made. If 'simplifyOutput' is true, the output contains only 'response'. Handles memory context if connected and supports dynamic tool function calls during the chat interaction.

Examples

Example 1: Send a prompt to the GigaChat model to generate a conversational reply.

Set 'modelId' to 'GigaChat', enter your prompt text, optionally configure parameters such as temperature or maximum tokens, then execute the node. Connect memory input if conversation context persistence is desired. Enable tools input if you want the AI to invoke external functions dynamically.

Links

Discussion