Guides
Intelligence Trace
Intelligence Trace
See how Aivana handled a request, without exposing how it works inside.
Intelligence Trace explains how Aivana handled a request: how much intelligence it applied, how many independent perspectives contributed, and whether live information was used. It's a customer-safe summary. It doesn't reveal which models ran or how Aivana decides.
Where to find it
- In the Playground, the Intelligence Trace panel shows the trace for each answer, automatically.
- In the API, set intelligence_trace to true. The response's trace field carries it; a stream carries it in trace events.
- In Usage & Insights, find a run by the request_id from its response.
Summary
| Field | Description |
|---|---|
| intelligence_level | How much intelligence Aivana applied to the request: Efficient, Advanced or Premium. Explained below. |
| route | The kind of path the request took, in plain terms. |
| perspectives | How many independent perspectives contributed to the answer. |
| complexity | A short description of how demanding Aivana judged the request to be, when it reports one. It is for display: don't branch on its wording. |
| fresh_data | Whether live web information was used. See Web Search. |
| sources_used | How many live sources were used, when web search ran. |
| elapsed_ms | Time taken, in milliseconds. |
The trace also has status (running, completed or failed), a short list of steps, each with a title, detail and status, and why_this_route, a plain-language explanation of the path taken.
Intelligence level and effort
intelligence_level is the outcome: how much intelligence Aivana applied to this request. It is always one of three names. Efficient suits straightforward requests, while Advanced and Premium apply progressively more depth to harder ones. Usage & Insights uses the same three names for the routing mix across a period.
effort is the input you control: a ceiling on how much intelligence Aivana may apply (auto, low, medium or high; the Playground's Effort setting is the same parameter). The level is what Aivana did within that ceiling. With the default, auto, Aivana judges from the request alone, and a lower effort bounds that judgement rather than replacing it. The two are not a lookup table: a given effort doesn't promise a given level, or a particular number of perspectives. How Aivana decides isn't part of the trace.
Example
curl https://api.aivana.ai/v1/generate \
-H "X-API-Key: $AIVANA_API_KEY" \
-H "Content-Type: application/json" \
-d '{"prompt": "Should we enter the EU market in 2027?", "intelligence_trace": true}'Built to be shown
Everything in the trace is safe to display to your own users. Use it to build trust in an answer or to debug a request; quote the request_id when you contact support.
Intelligence Trace explains an individual request. Usage & Insights summarizes behavior and economics across a period you choose, and its benchmark comparison is made on the whole period, not as a savings claim about any one run.
Learn more
Try the Playground
Experiment with Aivana MMI — no account or code required.
Parameters
Every option the Generate API accepts.
Quickstart Guide
Get your key and make your first request in minutes.
Still have questions? We are here to help!