AI-Powered PDF Report Generation Workflow

This project is an automated AI pipeline designed to ingest complex inspection and thermal PDF reports and synthesize a professional, structured Detailed Diagnostic Report (DDR). Video Link: https://www.loom.com/share/fd7af0c5083e478ebb8da5c8e2b03f05

🚀 Overview

The workflow extracts text and images from multiple sources, processes them through a high-reasoning LLM (Google Gemini 2.5 Flash), and generates a client-ready PDF with:

• Professional Metadata: Site Address, Structure Type, Age, etc.
• Color-Coded Checklists: High-fidelity tabular assessments with status-based colors (Red/Yellow/Brown).
• Contextual Image Mapping: Photos are automatically placed alongside relevant observations.
• Summary Matrix: A final table mapping impacted areas to their probable root causes.

🛠️ Performance & Stability Features

• Dimensional Filtering: Automatically discards tiny/noise PDF artifacts (<150x150 pixels) to ensure clean reports.
• Auto-Retries: Integrated with tenacity for exponential backoff retries against transient API 503 errors.
• Strict Schema Enforcement: Uses Pydantic to guarantee 100% valid JSON responses from the AI.
• Intelligent Word Wrapping: Custom horizontal space boundary logic inside the FPDF engine to prevent rendering crashes.

📁 Project Structure

• app.py: FastAPI backend and API routes.
• frontend/: Web interface (HTML/CSS/JS).
• main.py: CLI orchestration script.
• pdf_processor.py: PDF extraction logic (text/images).
• llm_analyzer.py: Gemini 2.5 Flash integration.
• report_generator.py: PDF layout engine.
• Sample Report.pdf & Thermal Images.pdf: Input files.
• Main_DDR_Output.pdf: Final generated output.

⚙️ Setup & Execution

1. Prerequisites: Python 3.10+ and a Google Gemini API Key.
2. Installation:

   python -m venv venv
   .\venv\Scripts\activate
   pip install -r requirements.txt

3. Environment: Create a .env file and add:

   GEMINI_API_KEY=your_key_here

4. Run:
   • Web Interface (Recommended):

     uvicorn app:app --reload

     Access at: http://127.0.0.1:8000/
   • CLI Mode:

     python main.py

🎥 Submission Requirements

• Loom Video Link: Watch Video
• GitHub Repository Link: jayeshpandey01/DDR-Reports
• Project Demo Link: http://127.0.0.1:8000/ (Local Dev)

---

Built within the 24-hour AI Workflow Challenge.

D D R - R e p o r t s