Skip to content

Lead CSV Orchestrator & Virtual Quality Board

Overview

The Lead CSV Orchestrator (src/agents/orchestrator.py) acts as the Director of a Virtual Quality Board. In regulated environments, life sciences deliverables—such as a formal Validation Summary Report (VSR) or Release Authorization—cannot be issued by a single engineer or a single AI model.

The Orchestrator coordinates multi-agent squads, synthesizes cross-domain findings, checks consensus across specialist agents, and compiles formal regulatory deliverables.

Orchestration Workflow

[ User or REST API Request ]
|
v
+---------------------------------+
| Lead CSV Orchestrator |
+----------------+----------------+
|
+---------------------------------+---------------------------------+
| | |
v v v
+------------------+ +------------------+ +------------------+
| Traceability | | Risk Assessment | | Validation Test |
| Specialist Agent | | Specialist Agent | | Specialist Agent |
+--------+---------+ +--------+---------+ +--------+---------+
| | |
+---------------------------------+---------------------------------+
|
v
+---------------------------------+
| EvidenceGate Verification Gate |
+----------------+----------------+
|
v
+---------------------------------+
| Virtual Quality Board Consensus |
+----------------+----------------+
|
v
+---------------------------------+
| Formal VSR Compilation & Report |
| (SHA-256 Signature Manifest) |
+---------------------------------+

Validation Summary Report (VSR) Synthesis (/api/orchestrator/vsr)

When requested to compile a Validation Summary Report for a target system (such as @lims or @ebr), the Orchestrator executes a 5-phase procedure:

Phase 1: Context Assembly & Requirement Retrieval

The Orchestrator retrieves all active User Requirements (URS), Functional Specifications (FS), and Design Specifications (DS) for the target computerized system from PostgreSQL and Memgraph.

Phase 2: Parallel Specialist Agent Consultation

The Orchestrator invokes specialist agents concurrently:

  1. RequirementsTraceabilityAgent: Computes RTM coverage percentage, verifies bidirectional mapping, and identifies orphaned requirements.
  2. RiskAssessmentAgent: Evaluates FMEA Failure Modes and computes pre-mitigation and post-mitigation RPN scores.
  3. ValidationTestingAgent: Evaluates IQ/OQ/PQ execution evidence, passing rates, and open test incident reports.
  4. DataIntegrityAuditAgent: Audits 21 CFR Part 11 electronic signature logs, timestamp integrity, and ALCOA+ compliance.

Phase 3: Cross-Agent Consensus & Conflict Resolution

The Orchestrator parses the structured responses from all consulted agents. If ValidationTestingAgent reports unexecuted high-risk protocols while RiskAssessmentAgent identifies severe patient safety hazards, the Orchestrator automatically rejects release authorization and flags the deliverable as RELEASE_WITHHELD_PENDING_RETESTING.

Phase 4: Formal Deliverable Generation

Compiles a structured report containing:

  • Executive Summary & System Boundary Definition
  • Regulatory Scope & GAMP 5 Classification
  • Traceability Summary & RTM Coverage Matrix
  • Testing Qualification Status (IQ / OQ / PQ)
  • Data Integrity & Audit Trail Review
  • Recommendations & Action Items

Phase 5: Cryptographic Manifest Stamping

The generated deliverable is hashed using SHA-256 and committed to PostgreSQL with a 21 CFR Part 11 digital signature manifest, ensuring tamper evidence and non-repudiation.