title	BMR Date & Sequence Validator
emoji	🛡️
colorFrom	gray
colorTo	green
sdk	gradio
sdk_version	5.29.0
app_file	app.py
pinned	true
license	apache-2.0
short_description	21 CFR Part 11 deterministic batch record date validator

🛡️ Validated BMR Micro-Agent

Batch Record Date & Sequence Validator — v1.0.0

A 100% deterministic, cryptographically-hashed, 21 CFR Part 11-compliant micro-agent
for validating date formats and chronological sequence in electronic Batch Records.

No LLMs. No hallucinations. No ambiguity. Just math.

---

What This Agent Does

In pharmaceutical manufacturing, an electronic Batch Record (eBR) is a legal document. Every date entry must:

1. Conform to ICH Q7 format — DD-MMM-YYYY (e.g., 10-MAY-2026)
2. Follow chronological order — no step can be dated before the preceding step (backdating indicator)

A single deviation triggers an FDA data integrity finding. This agent checks both rules deterministically in < 5 milliseconds.

---

Architecture

Input: BMR Text (plain-text / OCR)
         │
         ▼
┌─────────────────────────────────────────────────────────────┐
│  engine.py  (BatchRecordValidator)                          │
│                                                             │
│  PASS 1: Regex extraction → RULE_01_FORMAT (ICH Q7)         │
│          DD-MMM-YYYY strict? Valid calendar date?           │
│                                                             │
│  PASS 2: Chronological scan → RULE_02_SEQUENCE              │
│          Date[N] >= Date[N-1]? Otherwise → backdating FAIL  │
│                                                             │
│  SHA-256 hash of input → tamper-evident anchor              │
└─────────────────────────────────────────────────────────────┘
         │
         ▼
Output: Part11AuditTrail (Pydantic v2 JSON)
  ├── execution_id (UUID v4)
  ├── timestamp_utc (ISO 8601)
  ├── input_sha256 (tamper evidence)
  ├── status: PASS | FAIL
  ├── rules_evaluated[] (every finding with evidence)
  ├── summary (format_failures, sequence_failures)
  └── execution_metadata (runtime_ms, peak_ram_mb)

Zero external API calls. Zero ML inference. Zero hallucination surface.

---

Quick Start

# 1. Clone
git clone https://github.com/j-arndt/validated-bmr-micro-agent
cd validated-bmr-micro-agent

# 2. Install
pip install -r requirements.txt
pip install "gradio>=5.0.0"   # managed by sdk_version on HF Spaces; install manually for local use

# 3. Run engine (CLI)
python engine.py

# 4. Launch Gradio UI
python app.py
# → http://localhost:7860

---

Sample Output

{
  "schema_version": "1.0",
  "agent_name": "BMR-Date-Sequence-Validator",
  "agent_version": "1.0.0",
  "execution_id": "a3f7c12e-...",
  "timestamp_utc": "2026-05-13T16:00:00.000000+00:00",
  "input_sha256": "e3b0c44298fc1c149afb...",
  "status": "FAIL",
  "rules_evaluated": [
    {
      "rule_id": "RULE_01_FORMAT",
      "result": "FAIL",
      "evidence": {
        "location": "Line 3",
        "found_value": "05/11/2026",
        "expected_value": "DD-MMM-YYYY",
        "delta": "Format deviation - expected ICH Q7 standard"
      }
    },
    {
      "rule_id": "RULE_02_SEQUENCE",
      "result": "FAIL",
      "evidence": {
        "location": "Line 5",
        "found_value": "09-MAY-2026 (follows 11-MAY-2026 at Line 4)",
        "expected_value": ">= 11-MAY-2026",
        "delta": "-2 days - chronological break (possible backdating)"
      }
    }
  ],
  "summary": {
    "total_dates_found": 7,
    "format_failures": 2,
    "sequence_failures": 1
  },
  "execution_metadata": {
    "runtime_ms": 2.847,
    "peak_ram_mb": 38.21
  }
}

---

Validation Package

The /validation_package folder contains the regulatory trust layer that separates this from a bare script:

Document	Description	GAMP 5 Phase
IQ_Report.md	Installation Qualification — Python version, dep hashes, file integrity	IQ
OQ_Test_Matrix.csv	1,000 synthetic test cases (4 categories x 250 rows)	OQ
21CFR11_Addendum.md	Field-by-field FDA Subpart B requirement mapping + ALCOA+ assessment	OQ/PQ
generate_oq_matrix.py	Reproducible OQ matrix generator (deterministic seed)	OQ

Generate the OQ matrix locally:

python validation_package/generate_oq_matrix.py

---

OQ Test Categories

Category	Count	Description
A — True Negatives	250	Valid format + chronological sequences → PASS
B — Format Errors	250	MM/DD/YYYY, D-Mon-YY, slash separators, etc. → FAIL
C — Sequence Errors	250	Chronological breaks / backdated entries → FAIL
D — Edge Cases	250	Leap years (29-FEB-2024), impossible dates (31-FEB), same-day entries

---

Rules Reference

Rule ID	Name	Standard	Logic
RULE_01_FORMAT	Date Format Check	ICH Q7 / 21 CFR Part 11	Must match DD-MMM-YYYY with valid month abbreviation and valid calendar date
RULE_02_SEQUENCE	Chronological Sequence	GxP Data Integrity / ALCOA+	Date[N] must be >= Date[N-1]; any reversal is a backdating indicator

---

Design Philosophy

We do not use probabilistic generative models for deterministic compliance tasks.

Concern	LLM Approach	This Agent
Cost per execution	~$0.01–$0.10	~$0.000001
Hallucination risk	Present	Zero (deterministic)
Audit trail	None	Full 21 CFR Part 11 JSON
FDA defensibility	Unvalidated	IQ/OQ/PQ documented
Latency	2–30 seconds	< 5 ms
Offline capable	No	Yes

---

Requirements

pydantic>=2.9.0
psutil>=5.9.8

Gradio is managed via sdk_version: 5.29.0 on HuggingFace Spaces. For local use, install separately: pip install "gradio>=5.0.0"

Python >= 3.10 | No GPU required | Runs on any laptop or air-gapped GxP server

---

License

Apache 2.0 — see LICENSE

---

Built by Justin Arndt
18+ years GxP / CSV / QA | GSK background | Validated Micro-Agent Portfolio
GitHub