Torrix Proxy icon

Torrix Proxy

Send LLM requests through Torrix to log tokens, cost, latency, and prompt traces. Works with OpenAI, Anthropic, Groq, Mistral, Ollama, and any OpenAI-compatible API.

Torrix Proxy

AI-generated

Summary

Sends each input item’s user message, and optional system prompt, to an LLM endpoint through Torrix’s /proxy endpoint. Torrix receives the provider URL and credentials plus optional run, session, and trace labels; the node returns each proxy response as an output item.

Inputs

  • LLM Provider URL (required) — The provider endpoint URL. If empty, uses the Torrix credential’s default provider URL; execution fails if neither is set.
  • Provider API Key — Provider key sent to Torrix as upstream authorization. If empty, uses the credential’s default provider API key; if both are empty, the upstream authorization header is omitted.
  • Model (required) — Model name included in the request body. If empty, uses the credential’s default model; execution fails if neither is set.
  • User Message — Text sent as the user message. Supports n8n expressions to reference input data.
  • System Prompt — When non-empty, adds a system message before the user message.
  • Run Name — Optional label sent to Torrix for the run.
  • Session ID — Optional identifier sent to Torrix to group conversation turns.
  • Trace ID — Optional identifier sent to Torrix to group calls in an agent run.
  • Additional Request Fields — Adds max_tokens (default 1024) and temperature (default 1) to the request body. Temperature supports values from 0 to 2.

Output shape

One output item per input item, each containing the Torrix proxy response in its json field.

The node sends a POST request to the Torrix credential’s base URL with /proxy appended. The request body contains the model, messages, and configured additional fields.

Links

Discussion