Documents and events
What the agent reads, which events it extracts, how sources are ranked, and what to do about screenshots.
Document types
| Type | Formats | Extracted |
|---|---|---|
| Invoice / statement | PDF (text, scanned), XLSX, CSV | container, charge type, period, days, rate, tier, amount, currency, invoice number |
| B/L, sea waybill, booking | PDF, EDI 301/310 | container, ISO type/size, POL/POD, Incoterm, haulage (carrier/merchant), parties, vessel/voyage |
| Terminal / depot events | EDI 315/322, portal CSV, screenshot | event code, timestamp, location, source |
| EIR | photo, PDF | gate-in/out timestamp, container, chassis, seal, damage |
| Customs / government | PDF, email | hold placed, released, authority |
| Carrier notices | PDF, email | roll, change of destination, availability |
| Contracts / tariffs | clauses indexed by topic with page references | |
| Correspondence | EML, MSG, pasted | claims, concessions, evidence cited, dates, parties |
Event vocabulary
Discharge · Available for pickup · Hold placed / released (customs, terminal, carrier) · Gate-out full · Trucker handover (out) · Delivered · Empty available · Trucker handover (in) · Gate-in empty · Invoice issued.
Source ranking
When sources disagree: contract-named source of record → EDI event → signed EIR → portal export → email statement → screenshot. The file shows all timestamps found and marks which one the computation used and why.
Screenshots
Accepted. OCR extracts container, event and timestamp; the file carries a provenance note. Where possible, request the EDI event or EIR before sending the dispute.
Container numbers
Validated with the ISO 6346 check digit; transpositions are flagged and the shipment list searched for the intended number.
Limits
200 MB per file; 2 GB per batch. Statements over 1,000 lines are split. Password-protected PDFs must be unlocked first.