Unify LLM icon

Unify LLM

LangChain-style orchestration and multi-provider intelligence via Unify LLM

Actions10

Orchestration → Route Only

AI-generated

Summary

Routes a prompt through a Bayesian-Pareto-Prim ensemble to select the best candidate model without generation.

Inputs

  • prompt (required) — Prompt or query text
  • systemPrompt — Optional system instruction
  • temperature — Sampling temperature for the routing request, 0-2, default 0.7
  • maxTokens — Maximum tokens for the routing request, default 512
  • candidateModels — Comma-separated models used by the combined router ensemble, default: gpt-4o-mini,claude-3-5-sonnet-20241022,gemini-1.5-pro,llama3.1

Output shape

Returned as a single item on the main output connection.

If continueOnFail is enabled, errors produce an item with json.error instead of throwing.

Examples

Example 1: Select best model from ensemble for downstream use.

Prompt: 'Summarize this text.' No system prompt. Default temperature and maxTokens. Default candidate models. Operation: Route Only. Resource: Orchestration. Result: json contains selectedModel (winning model name), voting (array of {router, selectedModel, error} per Bayesian/Pareto/Prim), and parsed candidateModels array.

Links

Discussion