Skip to main content
You may optionally pass an extra header to disable template induction on a per-request basis.
curl ...
-H "Butter-Disable-Induction: true" \
...
Template induction runs asynchronously from request-time, so it doesn’t affect request latency.

Format

Butter-Disable-Induction expects the string value "true" to disable template induction for a request. Any other value (or omitting the header) will leave template induction enabled.

Combining with Manual Bindings

You can still use the Butter-Bindings header even when template induction is disabled. This allows you to manually specify templates while preventing automatic pattern detection:
curl -X POST https://proxy.butter.dev/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $OPENAI_API_KEY" \
-H "Butter-Auth: Bearer $BUTTER_API_KEY" \
-H "Butter-Bindings: {\"name\": \"Erik\"}" \
-H "Butter-Disable-Induction: true" \
-d '{"messages":[{"role":"user", "content":"Say hello to Erik"}],"model":"gpt-4o"}'