Debug with ExecLog

Use execution logs to find what went wrong

Every request is tracked in ExecLog. Use it to debug issues.

View Recent Executions

orpheus runs my-agent

Output:

REQUEST_ID                            STATE      DURATION  WORKER
a1b2c3d4-e5f6-7890-abcd-ef1234567890  completed  1.2s      my-agent-worker-1
b2c3d4e5-f6a7-8901-bcde-f12345678901  failed     0.8s      my-agent-worker-2
c3d4e5f6-a7b8-9012-cdef-123456789012  completed  2.1s      my-agent-worker-1

Filter by State

# Only failed requests
orpheus runs my-agent --state failed

# Only crashed (daemon died mid-request)
orpheus runs my-agent --state crashed

Request States

State	Meaning
queued	Waiting for worker
started	Currently executing
completed	Finished successfully
failed	Handler returned error
crashed	Daemon died during execution

Debug a Failed Request

1. Find the failed request:

   orpheus runs my-agent --state failed
   

2. Get details:

   curl "http://localhost:7777/v1/execlog?agent=my-agent&state=failed&limit=1"
   

3. Check the error message in response

Find Crashed Requests

Crashed = daemon died while request was running

orpheus runs my-agent --state crashed

Crashed requests are NOT auto-retried (safe-first design). You decide:

• Safe to retry? Run it again
• Has side effects? Investigate first

API Access

# Get last 10 executions
curl "http://localhost:7777/v1/execlog?agent=my-agent&limit=10"

# Filter by state
curl "http://localhost:7777/v1/execlog?agent=my-agent&state=failed"

What Gets Logged

Field	Description
request_id	Unique ID for this execution
agent_name	Which agent
worker_id	Which worker handled it
session_id	Session (if provided)
state	Current state
started_at	When execution began
completed_at	When execution ended
duration_ms	How long it took
error	Error message (if failed)

FAQ

Quick answers →