# RFQ review demonstration

This is a synthetic demonstration of a native n8n file workflow. It reads a text PDF, a catalogue CSV and a separate approved-price CSV, then writes an Excel review workbook. It does not connect to an ERP or send quotations.

The visible sample contains nine lines: two accepted and seven exceptions. The workbook has one Review sheet with an explicit status and reason for every line. A separate eight-line fixture and a four-line supplemental fixture were also executed and inspected. Those tests cover decimal quantities, invalid quantities, unknown codes, duplicate references, unit differences, price mismatches, missing prices and literal formula-looking text. Missing, invalid and blank PDFs produced failed executions in the independent checks.

The layout is deliberately narrow: one title line beginning RFQ synthetic fixture v, then numbered rows separated by vertical bars. Columns are line reference, code, description, quantity, unit and optional requested price. The approved-price CSV supplies the price used for accepted rows. Any supplied requested price must agree. Limits are two PDF pages, 50 RFQ lines and 200 rows in each reference CSV. Other document layouts require a parser change and new validation. These fixtures do not establish performance on customer documents.

Files in the zip retain the rfq-demo and rfq-acceptance directory structure. No runtime packages, credentials or private logs are included. The tested runtime was n8n 2.39.1 beta on Node 24.18.0. Configure an appropriate supported n8n environment before using the workflow. The standalone workflow.json has portable /files paths; change the three read paths and the write path for your own instance and use its documented permitted-file settings.

The Python build-native.py and run-native.py files reproduce the tested local CLI setup when n8n is installed under rfq-demo/runtime. They use Python's standard library. The build command generates paths for that checkout; the run command records input, workflow and fresh output hashes. It does not turn an old workbook into a passing run when execution fails.

The workbook is a review artifact. Exceptions remain for a person to resolve. It makes no product substitution, pack conversion, currency conversion or production update. Totals are withheld pending review. A paid test requires representative files and separately agreed scope, acceptance checks and delivery terms.

Small Fixes Lab