deepread-bank-statementslisted

# DeepRead Bank Statements Turn any bank statement — PDF export or scanned/photographed — into clean, typed JSON: account holder, statement period, opening and closing balances, and a row-by-row transaction list with dates, descriptions, amounts, and running balances. Every field comes back with a `needs_review` flag so you know exactly what to trust and what to double-check. > This skill instructs the agent to POST documents to `https://api.deepread.tech` and poll for results. No system files are modified. ## Why this is hard (and why DeepRead handles it) Bank statements are the worst of OCR: dense tables, multi-page transaction runs, inconsistent layouts across thousands of banks, credits and debits in separate columns, and running balances that have to reconcile. A single-model OCR pass silently drops rows or flips a debit into a credit. DeepRead runs **multi-model consensus** (GPT + Gemini + an LLM judge), returns a confidence flag per field, and lets you reconcile against the stated opening/closing balances. ## What You Get Back Submit a statement PDF, get structured JSON. Extracted fields come back under `extraction.fields[]` (each with `key`, `value`, `needs_review`, `location.page`): ```json { "schema_version": "dp02", "status": "completed", "extraction": { "fields": [ {"key": "account_holder", "value": "Jordan Rivera", "needs_review": false, "location": {"page": 1}}, {"key": "bank_name", "value": "First National Bank", "needs_review": fal