apo
Demo · read-only
Sign in
Agent Testing
Tasks
Runs
Schedules
Observability
Traces
Toggle Sidebar
demo
Runs
Runs
demo-bch
real-agent/documents/data-extraction
demo-run_3
Failed
real-agent/documents/data-extraction
·
openai/gpt-4o-mini
·
cli
·
8w ago
Details
▸
Task
Run
Trace home
Delete
40%
pass rate
2
passed
·
3
failed
·
5
checks
17.9s
duration
$0.002153
9.3k tok
—
reasoning
Checks
5
Conversation History
Deliverables
Trace home
2
/5 passed
3 failed
Click to expand
✓
used-extraction-workflow
✗
extracted-all-key-fields
expected expected finding: invoice number findings include invoice number; expected expected finding: invoice date findings include invoice date; expected expected finding: due date findings include due date; expected expected finding: tax amount findings include tax amount; expected expected finding: total findings include total; expected expected finding: seller name findings include seller name; expected expected finding: buyer name findings include buyer name; expected expected finding: seller email findings include seller email; expected expected finding: payment terms findings include payment terms
✗
findings-grounded-in-invoice
Value 1 indicates successful entity extraction with accurate capture of dates, names, emails, and amounts, which suggests that these values can be traced to the invoice document content. However, Value 2 highlights a discrepancy between the calculated total and the provided subtotal in the invoice, indicating a potential issue with the accuracy of the amounts. Since the instruction requires that all values must be traceable and accurate, the discrepancy in Value 2 suggests that not all values meet this criterion.
✗
extraction-is-thorough
Value 1 indicates successful entity extraction with various details captured, but it does not mention the specific secondary details required by the instruction, such as the early-pay discount, contact person, account manager, or line items. Value 2 discusses a discrepancy in totals but does not provide any of the required secondary details either. Therefore, neither value meets the criteria for a PASS.
✓
invoice-file-present