How do I see AI API call history in Cognocient?
A searchable, filterable log of every AI API call passing through the Cognocient proxy — model, tokens, cost, latency, and attribution headers. The fastest way to verify attribution is working, debug unexpected costs, and confirm a new feature is being tracked correctly.
The API Call Log (Dashboard → Live Calls) shows every AI API call passing through the Cognocient proxy — model, tokens, cost, latency, and attribution headers — paginated 50 rows at a time, newest first. Use it to verify attribution is working, debug unexpected costs, or confirm a new feature is being tracked correctly.
What each row shows
| Field | Description |
|---|---|
| Timestamp | Time of the call |
| Model | Provider and model name (e.g., gpt-4o, claude-sonnet-4-6) |
| Feature | Value of X-Cost-Feature, or "—" if missing |
| Workload | Value of X-Cost-Workload (e.g., agentic-read), shown as a badge |
| Dept | Value of X-Cost-Department |
| User | Value of X-Cost-User |
| Prompt / Compl | Input and output token counts |
| Cost | Cost in USD for this call |
| Latency | Response time in milliseconds |
| Status | A colored dot — green for 2xx, red for 4xx/5xx, amber otherwise. Hover for the raw status code. |
Filtering the call log
The filter bar has four controls:
- Search Model — free-text match against the model name
- Feature — dropdown of every feature tag seen in your data
- Department — dropdown of every department tag seen in your data
- Date Range — Last 24 hours / 7 days / 30 days / 90 days
There is no status, user ID, session ID, or cost-range filter in the UI today — if you need to isolate calls by one of those fields, export to CSV and filter there, or use the session_id query parameter directly against the /api/calls endpoint.
Common uses for the call log
Verify attribution headers after deployment — After shipping a new feature or updating header values, filter by feature name and refresh to confirm the new feature name is appearing on incoming calls.
Debug unexpected costs — If your dashboard shows a cost spike but you're not sure which feature caused it, set the date range to the spike window and filter by feature to see exactly which model and call pattern drove the increase.
Confirm proxy is routing correctly — After applying a routing rule (e.g., redirect gpt-4o → gpt-4o-mini for sentiment analysis calls), filter to that feature and check the Model column shows the cheaper model for matching calls.
Investigate errors — Rows with a red status dot are failed calls (4xx/5xx). Hover to see the exact status code. Frequent failures from one feature often indicate rate limiting, prompt errors, or context-window violations — all of which waste money without producing a usable response.
Exporting call logs
Click Export CSV to download the currently filtered view (respects your Feature, Department, Model, and Date Range filters). Use this for:
- Feeding into internal data warehouses (Snowflake, BigQuery)
- Custom compliance reports
- Providing call-level detail to enterprise clients in chargeback reports
Call log retention follows your plan — see Data Security for details.
Related articles