Success

Extract invoice data — 42 documents

run_4p8r2w1 retry

Agent

DataAgent

Sonnet 5

Duration

38.8s

14 steps

Tokens

11,240 in · 3,182 out

Cost

$0.08

Triggered by

API

POST /v1/runs · key ak_7f2…

Span

Tokens

Time

DataAgent

—

38.8s

Plan extraction strategy

1.8K

2.7s

storage.list_documents

—

610ms

Classify document types

2.7K

3.4s

ocr.extract_text

—

12.2s

ocr.health_check

—

520ms

ocr.extract_text#2

—

6.4s

Normalize invoice fields

4.5K

4.8s

schema.validate

—

720ms

ReviewAgent

—

4.9s

Review flagged outliers

3.1K

3.5s

db.upsert_invoices

—

1.1s

notify.slack

—

590ms

Summarize batch result

2.3K

900ms

storage.write_report

—

120ms

Inspector

tool.callfailed

ocr.extract_text

Latency

12.2s

Offset

+7.0s

Tokens

—

Model

—

ToolExecutionError

ocr.extract_text timed out after 12000ms (upstream 502 from ocr-gateway)

ToolExecutionError: upstream returned 502
    at OcrClient.extractText (services/ocr/client.ts:118:11)
    at async ToolRunner.invoke (runtime/tools/runner.ts:64:20)
    at async AgentLoop.step (runtime/agent/loop.ts:203:16)
  caused by: GatewayTimeout: queue depth 4812 exceeded admission threshold
    at ocr-gateway/admission.go:71

{
  "batch_id": "inv-2026-08-B7",
  "language": "en",
  "timeout_ms": 12000
}

Output

No output — the span failed.

Tool schema

ocr.extract_text

Run OCR over a batch of document blobs and return text layers.

batch_idstringreqStorage batch identifier

languagestringISO 639-1 hint for the OCR engine

timeout_msintegerPer-document timeout

0:38.8 / 0:38.8