BIS Compliance Engine

Appearance

Step 11: Architectural Gaps & Production Roadmap ​

This document outlines the known architectural gaps and technical debt in the current 21-BIS-Compliance-Engine implementation, ranked by business and engineering criticality, along with concrete remediation proposals.

Prioritized Gaps & Fix Suggestions ​

1. Automated Document Verification via OCR (Critical) ​

  • Current State: Users manually check compliance artifacts via checkboxes in the UI ([App.jsx:L119-L134](file:///Users/avishjain/avishjain9-repos/21-bis-compliance-engine/bis-frontend/src/App.jsx#L119-L134)).
  • Risk: An auditor or user can check all boxes without providing authentic legal proof, resulting in unverified "Audit Ready" determinations.
  • Fix Suggestion: Add a multipart file upload route POST /audit/verify-document that uses Tesseract OCR or Google Document AI to parse PDF calibration certificates, extract NABL accreditation numbers, check expiration dates against the current date, and verify validity automatically.

2. Multi-Candidate Standard Retrieval ($k > 1$) (High) ​

  • Current State: Search query hardcodes index.search(query_vec, k=1) in [app.py:L90](file:///Users/avishjain/avishjain9-repos/21-bis-compliance-engine/app.py#L90), returning only the single nearest standard.
  • Risk: Complex industrial products (e.g., an electric vehicle that requires both AIS 156 for its battery and AIS 004 for electromagnetic compatibility) cannot be audited comprehensively in a single session.
  • Fix Suggestion: Upgrade retrieval to $k=3$, calculate relative confidence percentages, and allow the user to select the primary standard or run a joint composite audit pack.

3. Persistent Database & Audit History (High) ​

  • Current State: All state is ephemeral in client memory. Refreshing the browser resets the chat stream and audit packs.
  • Risk: Quality directors cannot review historic factory audits or track readiness score progression over time.
  • Fix Suggestion: Integrate SQLite (for single-node edge deployments) or PostgreSQL with SQLAlchemy/Alembic migrations, storing audit records, user IDs, timestamped findings, and exported report links.

4. Authentication, Authorization & Multi-Tenancy (Medium) ​

  • Current State: Endpoints are public and unauthenticated.
  • Risk: Competitors or unauthorized users on local networks can execute unlimited queries.
  • Fix Suggestion: Implement OAuth 2.0 with JWT access and refresh tokens, supporting role-based access control (RBAC) with roles: Auditor, FactoryManager, and Guest.

5. Production Hardening: Dynamic CORS & Rate Limiting (Medium) ​

  • Current State: CORS allows all origins (*) in [app.py:L18](file:///Users/avishjain/avishjain9-repos/21-bis-compliance-engine/app.py#L18); frontend hardcodes API_URL = "http://127.0.0.1:8000".
  • Risk: Potential for CSRF abuse from third-party sites and deployment friction when hosting frontend and backend on different hostnames.
  • Fix Suggestion: Restrict CORS to explicit allowed origins from environment variables, externalize VITE_API_URL, and install slowapi to enforce rate limits of 30 req/min per IP.

6. Branded PDF Audit Certificate Generation (Low) ​

  • Current State: Export produces raw RFC 4180 CSV files ([app.py:L148](file:///Users/avishjain/avishjain9-repos/21-bis-compliance-engine/app.py#L148)).
  • Enhancement: Add a styled PDF generator using WeasyPrint or ReportLab to output formatted audit certificates with official seal watermarks, tabular scorecards, and cryptographic QR code verification.