Team Name

NT -Road Riders

Team Members

Project Description

📊 PBS What-if Simulator + Vetted Chat

Deterministic, retrieval-only budget simulator with trust scoring, audit trails, and Streamlit UI.

This tool allows analysts to explore the 2024–25 Australian Government Portfolio Budget Statements (PBS) and procurement data in a controlled, auditable way.

It combines:

- What-if simulation: project procurement changes into PBS budget impact.

- Vetted chatbot: retrieval-only analyzers with no hallucinations.

- Trust scoring: freshness, coverage, consistency, backtesting.

- Audit log: JSONL file of all chat interactions for compliance.

- Evidence tables & provenance: every answer includes supporting data lineage.

🚀 Features

• PBS Integration

  • Auto-downloads the official PBS 2024–25 program expense CSV from data.gov.au.
    
  • Normalizes headers and numbers across encodings.
    

• Procurement Data

  • Uses sample procurement dataset (synthetic if none provided).
    
  • Daily transactions across ICT, Construction, Consulting, Office Supplies, Travel.
    

• What-if Simulator

  • Apply shocks: category-specific changes, inflation, savings.
    
  • Compare scenario vs PBS baseline with variance.
    

• Vetted Chatbot (Retrieval-Only)

  • Supports vetted queries:
    
  • Budget outlook vs PBS
    
  • Vendor outliers
    
  • Procurement red flags
    
  • Rejects unsupported or low-evidence queries.
    

• Trust Scoring

  • Composite score based on:
    
  • Data freshness
    
  • Coverage vs elapsed months
    
  • Cross-check consistency
    
  • Statistical strength
    
  • Forecast backtest accuracy
    

• Audit Trail

  • All chat answers logged in audit_logs_chat.jsonl.
    
  • Includes query, analyzer intent, trust metrics, provenance fingerprints.
    

📦 Installation

```bash

Clone the repository

git clone https://github.com/your-org/pbs-what-if-simulator.git
cd pbs-what-if-simulator

Create a virtual environment (recommended)

python3 -m venv .venv
source .venv/bin/activate # or .venv\Scripts\activate on Windows

Install dependencies

pip install -r requirements.txt