- Access audit logs: see who changed what in your organisation
- Export logs: create an export job for your request logs
- Download exported logs: get the finished file
Have a list of IDs, such as from a data pack? See Look up by request ID for audit log entries and Export specific logs by ID for logs.
1. Access audit logs
Audit logs record administrative changes: who made a change, to which resource, and when.Access in the app
Go to Admin Settings > Audit Logs. Audit logs are collected automatically across all workspaces. See Audit Logs for what each entry contains.Filters
Narrow the results with any of these filters:
In the app, use the filters on the Audit Logs page.
Access with the API
CallGET /audit-logs. start_time, end_time, and organisation_id are required. start_time and end_time are ISO 8601 timestamps.
Results are paginated. Use
current_page and page_size (maximum 100) to page through them.
Look up by request ID
To fetch specific audit log entries from a list of request IDs, usePOST /v1/audit-logs/bulk instead of paging through a time range. Use an organisation admin API key.
In the app, filter by Request ID and paste the ID.
With the API, send the request IDs in batches:
- Send 500 to 2,500 request IDs per call.
- If the final batch has fewer than 500 IDs, set
last_pagetotrue. - The response contains the matching audit log entries and a
total. - IDs with no entry are left out. You only see entries for your own organisation. Duplicate IDs are ignored.
2. Export logs
Exports run as jobs. Each job writes your request logs for a time range, or for a list of log IDs, to a JSONL file, up to 50,000 logs per job.Export in the app
1
Open Exports
Go to Exports on the main sidebar and click Request Data.
2
Configure the export
Set the Time Range, the Logs Limits (maximum number of logs or a custom offset), and the Requested Fields to include.
3
Request and start the export
Click Request Export. The job appears in the exports list with the status Draft. Click Start to run it.
Export with the API
Create the export job, then start it. The job runs asynchronously.1
Create the export
POST /v1/logs/exports with a time range and the fields to export. The response contains the export id.2
Start the export
Open a single log in the app
To open one log by its ID, use this URL pattern:Export specific logs by ID
To get the full request and response for a list of log IDs, create an export withfilters.ids. Use an organisation admin API key.
- Include 1 to 50,000 IDs per export. For a longer list, create more than one export.
- Use
idsonly. Don’t combine it with a date range or other filters. - You can’t change the IDs on an export after creating it. Create a new export.
- Export by ID returns logs even if they are older than your plan’s log retention period.
- If your organisation blocks fields such as
requestorresponsefrom exports, creating the export is rejected. Check what is allowed withGET /v1/logs/exports/field-restrictions. - Export by ID is for logs only.
Enterprise and self-hosted organisations
Enterprise logs are stored only in your own log store. If the IDs you have belong to logs stored on Portkey SaaS, add the query paramsaas-logs-only=true when creating the export:
3. Download exported logs
Download in the app
1
Wait for the export to finish
Open Exports on the main sidebar. The status changes to Success when the job is done. If it shows Failure, click Start Again to retry.
2
Download the file
Click Download next to the completed export. The file is in JSONL format, one log entry per line.
Download with the API
1
Check the export status
GET /v1/logs/exports/{export_id}. Check about every 30 seconds until status is success.2
Get the download URL
GET /v1/logs/exports/{export_id}/download returns a signed URL. Use an organisation admin API key.3
Download the file
The signed URL expires quickly, so download the file right away. If the link has expired, request a new one.
Other data
- Hook results: results stored with your logs are included in log exports. For hook results that aren’t in logs, email datasupport@portkey.ai.
- Fine-tuning runs: list jobs with
GET /v1/fine_tuning/jobsand open one withGET /v1/fine_tuning/jobs/{job_id}.

