AI Procurement Readiness Tool

Singapore government agencies evaluate vendor AI systems against 11 AI Verify principles.

Project Status

Phase 1: Framework Extraction ✅

• Extracted 90 process checks from official AI Verify GitHub
• Fixed P9 Accountability parsing issue
• Deliverable: framework_structure_complete.json

Phase 2: Question Bank Construction ✅

• Generated 90 assessment questions (Q001-Q090)
• Each question mapped 1:1 to framework check (PID)
• Standardized scoring: Have=1.0, Partial=0.5, Gap=0.0
• Deliverable: question_bank.json

Phase 3: Scoring & BigQuery Schema ⏳

• BigQuery schema for assessments
• Scoring algorithm (Have/Partial/Gap → readiness %)
• Audit trail logging

Phase 4: LangGraph Workflow ⏳

• Assessment state machine
• Question-by-question flow
• HITL review gate
• Gemini integration (narrative generation)

Phase 5: Word Report Generation ⏳

• Readiness scorecard
• Principle-by-principle summary
• Professional Word document output

Phase 6: Streamlit UI ⏳

• Web interface for assessors
• Evidence file upload
• Real-time scoring display

Tech Stack

• Python 3.11+
• LangGraph (workflow orchestration)
• Google Gemini API (AI)
• Google BigQuery (audit trail, asia-southeast1)
• python-docx (Word report generation)
• Streamlit (web UI)

How to Run Phase 2

python src/phase2_question_bank/generate_questions.py

Generates question_bank.json with 90 assessment questions.

Project Structure

src/
├── phase1_framework/
│   ├── extract_aiverify_framework.py
│   └── __init__.py
└── phase2_question_bank/
    ├── question_bank.py
    ├── generate_questions.py
    └── __init__.py

docs/
└── PHASE_2_COMPLETE.md

README.md
.gitignore

Governance

• Framework extracted from official IMDA AI Verify source
• All questions mapped to framework checks (audit trail)
• Scoring standardized across all questions
• Every assessment logged to BigQuery (immutable append-only)

Author

AIknowlah - Learning AI Governance through hands-on project work

Singapore | 2026

Save the file.

---

## **Step 4: Create `.gitignore`**

Create a new file called `.gitignore` in your root folder:

Generated outputs

question_bank.json *.pyc pycache/ .env venv/ .DS_Store