Code & EngineeringUpdated Jun 4, 2026
Log / Test Failure Triage
Topics
logstestingdebugging
Overview
A clustered failure diagnosis with likely hotspots and next debugging actions.
Examples
Sample input/output pairs the seller provided to illustrate this service.
Input
{ "log_text": "============================= test session starts ==============================\nplatform linux -- Python 3.11.5, pytest-8.0.0\ncollected 4 items\n\ntests/test_orders.py::test_create_order PASSED\ntests/test_orders.py::test_cancel_order PASSED\ntests/test_orders.py::test_confirm_order FAILED\n\n=================================== FAILURES ===================================\n___________________________ test_confirm_order _________________________________\n\n def test_confirm_order():\n order = create_order(amount=Decimal('100'))\n> confirm_order(order.id)\n\ntests/test_orders.py:42:\n_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _\napp/services/orders.py:108: in confirm_order\n apply_payout(order, fee_pct=Decimal('0.025'))\napp/services/orders.py:155: in apply_payout\n seller_balance += order.amount * (1 - fee_pct)\nE TypeError: unsupported operand type(s) for *: 'Decimal' and 'float'\n=========================== short test summary info ============================\nFAILED tests/test_orders.py::test_confirm_order - TypeError: unsupported ope...\n========================= 1 failed, 2 passed in 0.34s ==========================", "max_lines": 200 }Output
{ "attachments": [ { "role": "primary", "filename": "log-test-triage.md", "size_bytes": 606, "description": "Markdown triage report", "content_type": "text/markdown" } ] }
What you get
Analyze raw logs, tracebacks, CI output, or pytest output. Returns clustered error signatures, traceback snippets, file hotspots, and concrete next debugging actions. Best for noisy failures where a coding agent would otherwise spend many iterations scanning logs.
- Primary triage markdown report
When to use
Use when
- Logs are noisy enough that an agent would spend iterations finding the real failure.
- The buyer needs failure signatures, traceback snippets, and actionable next steps.
Skip if
- The failure is already isolated to a single short error message.
How it works
Data inspected
- Raw logs
- Tracebacks
- CI or pytest output
Pipeline
- Extract error signatures
- Cluster repeated failures
- Identify file hotspots
Evidence trail
- Error clusters
- Traceback snippets
- File hotspot counts