LoomCycle LLM icon

LoomCycle LLM

Chat completion + embeddings via loomcycle's LLM gateway

LoomCycle LLM

AI-generated

Summary

Sends a conversation to LoomCycle's LLM gateway (POST /v1/_llm/chat, non-streaming) and returns the aggregated completion. Use it for chat completions without running an agent loop. The gateway picks the provider and model.

Inputs

  • Messages (required) — Ordered list of messages, each with a Role (user, assistant, system, tool) and Content text. At least one is required, or the node errors with "Chat requires at least one message."
  • Max Tokens — Additional field. Maximum tokens to generate. Sent as max_tokens (UI default 1024, minimum 1).
  • Temperature — Additional field. Sampling temperature from 0 to 2 (UI default 0.7).
  • Provider — Additional field. Routing hint that pins a provider such as anthropic or openai. On its own, the resolver picks the best model in that provider for the tier.
  • Model — Additional field. Routing hint that pins a model. With Provider it is an explicit pin. On its own, the resolver picks the provider that hosts it.
  • Tier — Additional field. Resolver dispatch tier. Per the field description, the gateway uses default when neither a pin nor a tier is given.
  • User ID — Additional field. Used for per-user quota tracking. If empty, it falls back to the credential's Default User ID. If that is also empty, the per-user cap is bypassed.
  • User Tier — Additional field. Per-user tier overlay. It takes precedence over Tier when set.

Output shape

One output item per input item. Its JSON is the gateway's aggregated chat response, which per the client docs includes the content blocks, usage counters, and the chosen provider and model.

The exact response field names are not defined in this source, so check a test run. Only role and content are sent per message, so tool definitions and tool calls cannot be passed from this node. Errors such as 401, 429 per-user quota, 503 resolver not configured, and 400 are wrapped as n8n API errors. With Continue On Fail, an item with an error message is returned instead.

Examples

Example 1: Summarize incoming text with a system prompt

Messages: a System message with instructions, then a User message with the text via expression. Additional Fields: Max Tokens 500, Temperature 0.2.

Example 2: Force a specific provider and model

Messages: one User message. Additional Fields: Provider = anthropic, Model = a model ID hosted by that provider.

Example 3: Apply per-user quotas

Messages: a User message. Additional Fields: User ID set to the end user's ID and User Tier set to their plan tier.

Links

Discussion