FISCUS AI Financial Workflow State Fixtures
Description
FISCUS AI Financial Workflow State Fixtures is a small, hand-authored reference dataset of 25 synthetic JSON Lines records for testing auditable, human-in-the-loop finance automation. It covers bank reconciliation, invoices, e-documents, expenses, collections, foreign-exchange rates, ledger workflows, organization authorization, idempotency, provider uncertainty, reversals, and reconstructible audit events. Each record separates an AI recommendation from a committed accounting fact and describes preconditions, transition guards, approval requirements, side-effect state, audit events, reversal paths, and expected user-visible status. The dataset is intended for examples, tests, documentation, and finance-control design discussions. No real customer, bank, invoice, tax identifier, personal data, production model output, provider secret, or live accounting record is included. It is not production-ready and does not provide accounting, tax, legal, or investment advice. Product context: https://fiscusai.org/ Canonical archive: https://doi.org/10.5281/zenodo.23269165
Files
Steps to reproduce
1. Download README.md and financial_workflow_state_fixtures.jsonl. 2. Verify the SHA-256 checksums: README.md = 5024b63275248491aee80ac00a98156f239205140990367fc91f2e3b77013a4f; JSONL = a4f618e6c8a01f93a78383142200ba89d1afca81bdba9f280f9053cb12ab1666. 3. Read the JSONL file line by line; each of its 25 lines is one independent valid JSON object. 4. Inspect the module, current state, proposed transition, guard, approval requirement, side-effect status, audit event, reversal path, and expected user-visible state. 5. Use the records only as synthetic reference fixtures. They are not executable production workflows and must not be treated as accounting, tax, legal, or investment instructions.