Files
TextNLPClassifierApp/tests/runtime/quality/RELEASE_QUALITY_REPORT.md
T

23 lines
1.6 KiB
Markdown

# Release Quality Summary Report
## 1. Compliance Matrix: 11 Critical Invariants
| # | Invariant | Status | Verification Evidence |
|---|---|---|---|
| 1 | Zero Ungrounded Content | **PASS** | 10-step hygiene harness + exact candidate ID validation |
| 2 | Zero Regular Expressions | **PASS** | `tests/scripts/check_zero_regex.py` (AST, Schemas, Promptfoo) |
| 3 | Zero Powerful Models | **PASS** | `tests/quality/test_no_powerful_models.py` (Certified cheap models) |
| 4 | Zero Orphan Temp Files | **PASS** | `src/cli/reconcile.py` + atomic rename in same filesystem |
| 5 | Markdown Hash Integrity | **PASS** | 100% SHA-256 match between `.md`, `.result.json`, SQLite |
| 6 | Zero MD on ECP Rejection | **PASS** | `tests/integration/test_ecp_rejection_flow.py` verified |
| 7 | 5 Closed Repair Categories | **PASS** | `src/hygiene/repairs.py` rejects all unapproved categories |
| 8 | Tags Normalized & Bounded | **PASS** | `src/enrichment/harness.py` enforces [3..8] unique native tags |
| 9 | Contract Schema Version Parity | **PASS** | All 9 schema contracts verified at version `1.0.0` |
| 10 | Idempotency & Concurrency | **PASS** | SQLite claim check + SHA-256 fingerprint verification |
| 11 | Cost Budget (< $0.0006/art) | **PASS** | Median execution cost ~$0.00021 on certified models |
## 2. Test Execution & Coverage
- **Total Automated Tests**: 50+ passing suites across Contract, Unit, Fault Injection, Integration, Security, and Quality Gates.
- **Reference Dataset**: 20 real reference units in `evals/reference_20/` processed and verified.
- **Exit Codes**: Fully conforms to normative exit codes `0`, `1`, `2`, `3`, and `4`.