Qvidence

Quantum + Evidence — bio/clinical evidence that survives the quantum transition.

---

What is Qvidence?

AI-native bio pipelines emit clinical, genomic, patient-stratification, and manufacturing data at high velocity. This data has long-lived value, which makes it a direct target of the harvest-now, decrypt-later threat model — while regulators (FDA real-time monitoring, EU AI Act, regional sovereignty) demand stronger evidence and stricter routing.

Qvidence binds three previously-separate domains into one operating system:

• Bio / Clinical — hospital, CRO, wearable, lab, and manufacturing evidence;
• PQC / Cybersecurity — every record sealed in a post-quantum capsule with a hash chain;
• Regulatory governance — consent-aware jurisdictional routing, zero-knowledge style audit, IP escrow, incident replay.

It is built end-to-end with the PG / PGF AI-native design framework.

---

Architecture (8 components)

#	Component	Module
1	Clinical Data Ingest Layer	qvidence/ingest.py
2	Crypto-Inventory Scanner	qvidence/crypto_inventory.py
3	PQC Evidence Capsule	qvidence/capsule.py
4	Consent & Jurisdiction Router	qvidence/consent.py
5	Zero-Knowledge Audit Interface	qvidence/zk_audit.py
6	Regulator Adapter Layer	qvidence/regulator.py
7	Pharma / IP Escrow Module	qvidence/escrow.py
8	Incident Replay Engine	qvidence/incident.py

Plus shared kernel core.py (types, ULID-shaped IDs, errors, clock pair) and façade app.py.

End-to-end flow:

caller
  │  record_consent(subject, scope, jurisdictions)
  │  ingest(source, raw, subject, scope, intended_jurisdictions, meta)
  ▼
QvidenceApp ─┬─▶ build_envelope ─▶ router.evaluate ─▶ build_capsule ─▶ store.append
             │   accumulator.append(digest)
             │   emit ingest / capsule_built / route_decision events
             │
             │  regulator_export(jurisdiction, report_type, subject, scope)
             ├─▶ re-evaluate consent at export time
             │   inclusion proofs ─▶ adapter.shape ─▶ structural-scan-no-bytes
             │
             │  revoke_consent(subject, scope)
             ├─▶ next regulator_export ⇒ ConsentDenied
             │
             │  replay(subject)
             └─▶ deterministic time-ordered events

For the full design, see docs/Qvidence-TechnicalSpecification.md and the PG/PGF artifacts under .pgf/.

---

Quick start

Stdlib-only — no third-party packages required for the prototype.

git clone https://github.com/sadpig70/Qvidence.git
cd Qvidence

# run tests
python -m pytest -q
# 28 passed in 0.14s

Minimal usage:

import sys; sys.path.insert(0, "src")
from qvidence.app import QvidenceApp

app = QvidenceApp()

subject = "patient-001"
scope   = "trial:NCT-Q01"

app.record_consent(subject, scope, jurisdictions=["FDA", "EMA"])

ingest = app.ingest(
    source="hospital",
    raw=b"vital-signs:hr=72,bp=118/76",
    subject_id=subject,
    scope=scope,
    intended_jurisdictions=["FDA", "EMA", "KR-MFDS"],
    meta={"study": "NCT-Q01", "site": "S001"},
)
print(ingest["allowed_sinks"])     # ['fda.realtime', 'ema.dossier']

fda = app.regulator_export("FDA", "realtime_monitoring", subject, scope)
print(fda["schema"], fda["merkle_root"])

app.revoke_consent(subject, scope)

# subsequent export now raises ConsentDenied
events = app.replay(subject)
print([e["kind"] for e in events])

---

Security model

Threat	Mitigation
Harvest-now, decrypt-later on TLS / DB / backup	Every record sealed in a PQC capsule (KEM + AEAD + signature).
Tampered storage	Capsule digest binds AAD + nonce + ciphertext + tag + encapsulated key; signature over digest; hash chain over prev_digest.
Mock provider in production	QVIDENCE_REQUIRE_REAL_PQC=1 blocks both build and verify against any provider whose name starts with mock-.
Raw payload exfiltration via regulator submission	RegulatorExportService runs a structural scanner; adapter shapes contain only digests / proofs / Merkle root / timestamps.
Replay of escrow release signature	Canonical message includes at_bucket = floor(at / 60).
Stale consent at export	Consent re-evaluated at export time, not at capsule-build time.
Inventory leakage	crypto_inventory reads filename + ≤ 256-byte sidecar manifest only — never opens secret bytes.
Logging payloads via incident events	EventLog.append rejects non-string and oversized detail values.

⚠ Mock cryptography. This prototype ships MockPqcProvider (mock-pqc-v0) built from HKDF-SHA256 + HMAC-SHA256 + a sha256-stream AEAD. It is not real PQC. Production deployments must register a PqcProvider backed by ML-KEM-768 / ML-DSA-65 (e.g. liboqs-python). The env gate QVIDENCE_REQUIRE_REAL_PQC=1 exists to prevent accidental promotion of mock artifacts.

---

Repository layout

.
├── README.md                                # this file
├── LICENSE
├── assets/
│   └── banner.svg
├── docs/
│   └── Qvidence-TechnicalSpecification.md   # full technical spec
├── .pgf/                                    # PG/PGF design + verification artifacts
│   ├── DESIGN-Qvidence.md
│   ├── REVIEW-Qvidence.md
│   ├── WORKPLAN-Qvidence.md
│   ├── status-Qvidence.json
│   ├── VERIFY-Qvidence.md
│   └── REPORT-Qvidence.md
├── src/qvidence/
│   ├── __init__.py
│   ├── core.py
│   ├── capsule.py
│   ├── consent.py
│   ├── ingest.py
│   ├── crypto_inventory.py
│   ├── zk_audit.py
│   ├── regulator.py
│   ├── escrow.py
│   ├── incident.py
│   └── app.py
└── tests/
    ├── conftest.py
    ├── test_core.py
    ├── test_capsule.py
    ├── test_consent.py
    ├── test_zk_audit.py
    ├── test_crypto_inventory.py
    ├── test_escrow.py
    └── test_e2e.py

---

PG / PGF conformance

Qvidence was specified, planned, executed, and verified using the PG (PPR/Gantree Notation) and PGF (PPR/Gantree Framework) AI-native skills.

• Gantree depth = 3 (limit 5); no (decomposed) split required.
• Every leaf passes the 15-minute / single-responsibility atomic-node test.
• @dep: graph is a DAG; topologically equivalent to the actual import graph.
• Acceptance criteria are inline (# acceptance_criteria:) and traced 1:1 to test cases in .pgf/VERIFY-Qvidence.md.

PGF artifacts:

Artifact	Purpose
DESIGN-Qvidence.md	Gantree + PPR.
REVIEW-Qvidence.md	3-perspective design review (feasibility / risk / architecture).
WORKPLAN-Qvidence.md	12 work nodes, DAG, POLICY block.
status-Qvidence.json	done = 12 / 12, tests = 28 / 0.
VERIFY-Qvidence.md	3-perspective cross-verification.
REPORT-Qvidence.md	Final report.

---

Roadmap (out of prototype scope)

1. Real PqcProvider (liboqs-python ML-KEM-768 + ML-DSA-65 or AWS KMS PQ hybrid).
2. Real ZK verifier (SNARK/STARK over capsule digest + processing-step transcript).
3. Persistence: append-only WAL store, atomic head fsync, restart recovery.
4. HTTP surface: thin FastAPI binding over QvidenceApp; auth via JWT-PQC hybrid.
5. Source connectors: HL7/FHIR ingest, MES batch ingest, wearable streaming.
6. Regulator schema fielding: align FDA / EMA / IRB shapes with current submission specs.
7. Multi-region: per-region capsule stores with cross-region Merkle anchoring.

---

License

See LICENSE.