For AI agents: the complete documentation index is available at https://aioproxy.dev/llms.txt, the full documentation bundle is available at https://aioproxy.dev/llms-full.txt, and this page is available as Markdown at https://aioproxy.dev/guide/configuration/transforms.md.
  • English
  • Request & Response Transforms

    Transforms allow you to adjust incoming requests, inject default hyperparameters, or modify response payloads without modifying client code.

    config.jsonc
    {
      "router": {
        "models": {
          "gpt-5": {
            "params": {
              "temperature": 0.7,
              "max_tokens": 4096,
            },
            "overrides": {
              "openai": {
                "model": "gpt-5-turbo",
              },
            },
          },
        },
      },
    }

    Practical Scenarios

    1. Parameter Clamping: Ensure consistent temperature or enforce context length limits across different clients.
    2. Model Rewriting: Expose a unified model name (gpt-5) externally, while mapping to provider-specific model IDs internally (gpt-5-2026-preview).
    3. Response Headers: Inject custom debugging headers or trace IDs into client responses.