JiMax Object LLM

Generate structured objects using AI SDK with internal persistent memory

JiMax Object LLM

AI-generated

Summary

Generates a structured object from a prompt using an OpenAI, Google Gemini, Anthropic, or OpenAI-compatible model. It can include OGG audio, starting messages, a system prompt, and optional session-based conversation memory.

Inputs

  • Provider and Model ID (required) — Choose a provider and enter its model ID. Supply the matching provider credentials; OpenAI-compatible credentials also provide a base URL.
  • Prompt (required) — Text sent as the current user message.
  • Audio URL (OGG) — Optional URL to an OGG audio file. The node fetches it and sends it with the prompt.
  • Zod Schema (required) — Zod expression defining the shape of the generated object; the variable z is available.
  • System Prompt — Optional system message prepended to the conversation.
  • Initial Messages (JSON) — Optional JSON array of starting messages. Each must have a role of user, assistant, or system and string content.
  • Use Memory — Enables persistent conversation history stored in jimax_memory.json in the process working directory.
  • Session ID — Required when memory is enabled; selects the conversation history to load and update.
  • Memory Window — When memory is enabled, sets how many previous interactions to include. The node loads up to twice this number of stored messages.
  • Save in Background — When memory is enabled, starts saving history without waiting for the save to finish.
  • Clear Memory — When memory is enabled, deletes the selected session history and returns a success message without generating a model response.

Output shape

One output item per input item, each containing the generated object as its JSON data.

The model is asked to produce an object matching the Zod schema. With Continue On Fail enabled, failed items instead contain an error field; clearing memory returns a status and message item for that input.

Links

Discussion