GoogleAI-agent

This Capstone Project is part of the 5-Day AI Agents Intensive Course with Google (Nov 10 - 14, 2025) and is open to all participants. The project lets you apply what you’ve learned during the course by building AI agents.

Supply Chain LLM ETL Pipeline

Automated pipeline that extracts tables from PDFs/CSVs using Gemini, detects data quality errors, performs LLM-based root cause analysis, auto-fixes valid issues, and generates an RCA PDF + audit logs.

---

🚀 Features

• PDF/CSV table extraction using Gemini API (with fallback to mock extractor)
• Data quality validations (missing values, invalid dates, quantity errors)
• LLM-based reasoning for Root Cause Analysis
• Auto-fixing when safe
• Generates:
  • final_cleaned_output.csv
  • rca_report.pdf
  • audit_log.json

---

📦 Installation

pip install -r requirements.txt

🔑 Setting API Key

Set environment variable:

Linux/macOS: export GOOGLE_API_KEY="your-key"

Windows PowerShell: $env:GOOGLE_API_KEY="your-key"

▶️ Running the Pipeline

python main.py --input data/sample_input.pdf

---

📁 Output

After running, the folder will contain:

• cleaned_output.csv
• rca_report.pdf
• audit_log.json

---

🧩 Fallback Mode

If no Gemini API key is available, the pipeline runs in MOCK mode, useful for:

• Developers
• Testing ETL flows
• CI/CD pipelines

---