Skip to main content
Fetches logs for a deployed model. Use this command to debug issues or monitor model behavior in production.

​
Options

​
TEXT
Name of the remote in .trussrc.
​
TEXT
ID of the model.
​
TEXT
ID of the deployment.
​
Tail for ongoing logs. Streams new log entries as they arrive.
​
DATETIME
Start of the log time range (ISO 8601). No-timezone values are local. Defaults to a short look-back ending at —end. Window must be <= 7 days.
​
DATETIME
End of the log time range (ISO 8601). No-timezone values are local. Defaults to now. Window must be <= 7 days.
​
TEXT
Logs from a relative time ago until now, written as a number plus a unit of s, m, h, or d (for example 90s, 2h, or 3d). Max 7d. Cannot be combined with --start or --end.
​
debug | info | warning | error
Minimum log severity. Defaults to all lines. Any value returns lines at or above that severity and drops lines with no level.
​
TEXT
Case-sensitive substring that must appear in the message (repeatable).
​
TEXT
Case-sensitive substring that drops any line containing it (repeatable).
​
TEXT
RE2 regex matched against the message. Prefer —includes/—excludes.
​
TEXT
Only return logs emitted by this replica (5-char short ID).
​
TEXT
Only return logs tagged with this inference request ID.
​
humanfriendly | W | WARNING | I | INFO | D | DEBUG
default:"humanfriendly"
Logging verbosity. humanfriendly (default) is pretty-printed; INFO, DEBUG, WARNING produce structured logs.
​
Disable interactive prompts. Use in CI/automated contexts where stdin isn’t a TTY.
Example: Fetch logs for a specific deployment:
Stream logs in real-time:
Fetch logs from the last two hours:
Fetch logs for a specific time window:
Filter logs to errors only:

​
Rate limits

Baseten limits the deployment logs endpoint (POST /v1/models/{model_id}/deployments/{deployment_id}/logs) to 30 requests/second per API key. Interactive use, including --tail, stays well under that limit. Scripts that wrap truss model-logs in a tight poll loop can hit the limit and receive 429 Too Many Requests. Wait for the response’s retry_after value (in seconds) before retrying. For the full response shape and a retry example, see management API rate limits.