Inspect usage
The chat header shows context and token information and estimated cost. The sidebar also summarizes recent AI usage across features.
These are estimates. Your provider’s billing records are authoritative, and a Notely license does not include provider credit. Image generation, automatic chat summaries, context compaction, document summaries, transcript formatting, Draft with AI, Builder tests, and repeated automated runs can create additional usage.
Enabling notifications for a schedule adds result checking, which can use up to two additional AI requests per run with Bedrock or OpenRouter.
Keep requests manageable
- Use a focused prompt and attach only relevant material. Break a large task into steps with review points.
- Review context and maximum-output settings in Chat → General and agent overrides. Increasing a limit does not give a model capabilities beyond the provider’s supported limits.
- Use supported prompt caching where appropriate and inspect usage after testing a recurring schedule.
When requests stop or are throttled
- For rate limits, wait and reduce concurrency or schedule frequency.
- For a context limit, start a focused conversation with a short summary and the key sources.
- For an output limit, ask for a smaller result or adjust the applicable output setting within the provider’s limits. Avoid repeatedly retrying without reviewing the error.