Quality Docs Validator

Catch potential inconsistencies between PFMEA and Control Plan — before technical review

A local-first Python CLI for quality, process and manufacturing engineers. Compare two .xlsx workbooks, surface severity-classified potential findings, and produce Markdown or JSON reports without sending your files to a cloud service.

Quick start · See a real example · What it checks · Limitations · Architecture

---

See it in action

Run the bundled synthetic example containing deliberately seeded gaps:

git clone https://github.com/migmcc/quality-docs-validator.git
cd quality-docs-validator
pip install -e .
qdv pfmea-control-plan \
  --pfmea examples/pfmea.xlsx \
  --control-plan examples/control-plan.xlsx \
  --out report.md

Documented sample terminal output (abridged):

Potential findings
  UNMATCHED_PROCESS_STEP                 warning    Op 40
  SPECIAL_CHARACTERISTIC_NOT_CONTROLLED  critical   Op 20
  MISSING_REACTION_PLAN                  critical   Op 20
  WEAK_DETECTION_METHOD                  warning    Op 20
  HIGH_SEVERITY_WEAK_CONTROL             warning    Op 20
  MISSING_CONTROL                        critical   Op 30

Score: 43/100    Verdict: FAIL
Critical: 3      Warnings: 3

Read the full committed sample report →

Important: Findings are potential inconsistencies for an engineer to judge. This is not a regulatory or standards compliance certification and does not replace human technical review.

The problem

A PFMEA and its Control Plan can look complete when reviewed separately while disagreeing between documents: an unmatched process step, a special characteristic not reflected in controls, or no reaction plan where severity is high. Those gaps deserve attention before an audit or submission.

Who it's for

Quality / Process / Manufacturing Engineers, supplier-quality teams, internal auditors and APQP/PPAP teams working with Excel-based documents.

What it validates

The current module is a PFMEA ↔ Control Plan consistency checker.

Capability	How it works
Excel input	Reads PFMEA and Control Plan .xlsx workbooks, with a recommended template and common header aliases.
Matching	Matches normalised operation IDs and process steps.
Six checks	Flags unmatched steps, missing controls, uncontrolled special characteristics, missing reaction plans and two warning-level weak-method heuristics.
Scoring	Severity-weighted 0–100 score with PASS / PASS-WITH-WARNINGS / NEEDS-REVIEW / FAIL verdicts.
Reports	Rich terminal summary plus Markdown or JSON, suitable for review and automation.

Read the exact finding triggers and scoring rules →

Quick start

Requirements: Python 3.12+ and a local checkout of this repository.

pip install -e .
qdv pfmea-control-plan \
  --pfmea examples/pfmea.xlsx \
  --control-plan examples/control-plan.xlsx \
  --out report.md

The bundled examples are synthetic and intentionally contain a gap. To regenerate them, run python examples/generate_examples.py.

JSON output for automation

qdv pfmea-control-plan \
  --pfmea examples/pfmea.xlsx \
  --control-plan examples/control-plan.xlsx \
  --format json --out report.json

The JSON report includes metadata, verdict, score, summary and findings.

Select sheets in multi-sheet workbooks

qdv pfmea-control-plan \
  --pfmea pfmea.xlsx --pfmea-sheet "PFMEA" \
  --control-plan control-plan.xlsx --control-plan-sheet "Control Plan" \
  --out report.md

If a specified sheet does not exist, the error lists available sheet names.

What it does not validate (by design)

• Formats other than .xlsx; CSV is not currently supported.
• Arbitrary column mapping beyond the recommended template and supported aliases.
• Other document pairs (Process Flow, Work Instructions, PPAP, 8D, SPC, MSA).
• Web interface, ERP / MES / QMS integration, PDF / OCR or AI-based interpretation.
• Regulatory conformity, audit certification or replacements for engineering judgement.

See scope, decisions and roadmap.

How it works

PFMEA.xlsx + Control Plan.xlsx → operation matching → six checks → scoring → terminal / Markdown / JSON

The rules and scoring are explicitly documented and testable rather than based on an LLM. For implementation details, see architecture and finding reference.

Get involved

• Explore the repository
• Open an issue
• Contribute

Licensed under MIT.