Skip to main content

​
optimize_response

Decode a provider reply that was previously compressed with Kaizen, restoring the full JSON payload for downstream processing.

​
Parameters

NameTypeRequiredDescription
ktofstr✅KTOF response string returned by your provider (e.g., completion.output_text).
optionsDecodeOptions❌Control indentation or strictness during decode.

​
Code example

async with KaizenClient.from_env() as client:
    optimized_request = await client.optimize_request({"prompt": prompt})
    provider_response = openai_client.responses.create(
        model="gpt-4o-mini",
        input=optimized_request["result"]
    )
    decoded = await client.optimize_response({"ktof": provider_response.output_text})
    print(decoded["result"])

​
Response example

{
  "operation": "optimize.response",
  "status": "ok",
  "result": {
    "messages": [
      {"role": "assistant", "content": "Here is your decoded reply."}
    ],
    "metadata": {"source": "openai.responses"}
  }
}

​
Errors

  • 400 → ktof missing or malformed.
  • 404 → Referenced payload ID expired (if using stored IDs instead of inline strings).

​
Notes

  • Use this method together with optimize_request when building middleware for provider SDKs.
  • If you only need raw decompression without request/response semantics, use decompress.