> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tacetai.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Going to production

> A short checklist before decisions start running on real traffic.

## Set a timeout and a fallback

Decide what happens if an answer doesn’t arrive in time. Usually that’s the same path as a low-confidence answer: send it to a person or a safe default.

```python theme={null}
from tacet import Tacet, TacetError

client = Tacet(timeout=1.0, max_retries=1)

try:
    result = client.decide(model="tacet-alpha", state=state, questions=questions)
except TacetError:
    return ask_a_human(state)
```

## Log every decision

Store the request id, the top answer, and its confidence. You’ll need them to audit decisions, tune thresholds, and spot drift.

## Pin the model

Use a dated model id in production, such as `tacet-alpha-2026-10`, so answers don’t shift when we release an update. Use `tacet-alpha` to always get the latest.

## Keep people in the loop

For decisions that seriously affect someone, such as money, health, or access, route low-confidence cases to a person and let people appeal automatic ones.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.