JiMax Text LLM

Generate text using AI SDK with internal persistent memory

JiMax Text LLM

AI-generated

Summary

Generates a text response for each input item using the selected AI provider and model. Optionally includes starting messages and session-based conversation history, and can clear a session’s saved memory.

Inputs

  • Provider — Select OpenAI, Google (Gemini), Anthropic, or an OpenAI-compatible endpoint; configure the matching credentials.
  • Model ID (required) — Model identifier passed to the selected provider.
  • Prompt (required) — User message sent to the model.
  • System Prompt — Optional system message prepended to the conversation.
  • Initial Messages (JSON) — Optional JSON array of starting messages, each with role user, assistant, or system and string content. Defaults to an empty array.
  • Use Memory — When enabled, loads and saves conversation history using a Session ID. Memory is stored in jimax_memory.json in the process working directory.
  • Session ID (required) — Session key used to load, save, or clear conversation history; shown when Use Memory is enabled.
  • Memory Window — Number of previous interactions to include from saved history; each interaction is stored as a user and assistant message. Defaults to 5.
  • Save in Background — When enabled, starts saving the updated memory without waiting for the save to finish. Otherwise, waits for the save.
  • Clear Memory — When enabled with Use Memory and a non-empty Session ID, deletes that session’s stored history and returns a success message instead of generating text.
  • Debug History — When enabled, adds the messages sent to the model to the output as _history.

Output shape

One output item per input item, containing a response string; when Debug History is enabled, it also contains _history, the messages sent to the model.

Clearing memory returns {status: "success", message: ...} for that item instead of a response. With Continue On Fail enabled, errors are returned as {error: ...} items; otherwise they are thrown. Saving memory is capped at the most recent 100 messages per session.

Links

Discussion