You can pass an explicitDocumentation Index
Fetch the complete documentation index at: https://docs.opper.ai/llms.txt
Use this file to discover all available pages before exploring further.
model on every call, or you can hand off the choice to Opper and just say what you’re optimizing for. Hints are how you do the second thing.
prefer: pick by goal, not by name
prefer | What Opper picks |
|---|---|
cheap | The lowest-cost model that can handle your task. |
fast | The model with the lowest latency. |
quality | The most capable model available, regardless of cost. |
balanced (default) | A reasonable trade-off across the three. |
Other hints
Standard generation parameters work too.| Hint | What it does |
|---|---|
temperature | 0 to 2. Lower is more deterministic. Default depends on the model. |
top_p | 0 to 1. Nucleus sampling. |
max_tokens | Caps the output length. |
reasoning_effort | low, medium, high. Only used by reasoning models. |
Python
How hints interacts with model and Route
There are three layers of model selection, and the most specific one wins:
- Explicit
modelparameter on the call. Always wins, as long as Comply allows it. - Route rule for the scope. Used if the call didn’t pass
model. hints.preferplus the Comply allowlist. Used if neither of the above is set.
hints and model. The prefer hint is ignored when model is set, but other hints (temperature, top_p, etc.) still apply.
What’s next
Calls
The basics of the JSON API.
Route
Set a default model at the scope level.
Models
Browse what
prefer picks from.