HarShar017/AI_Legislative_Analyzer

AI Legislative Analyzer

FastAPI + React application for analyzing legislative and policy documents using chunking, compression, and LLM-powered insight extraction.

Current Stack

• —Backend: FastAPI, Uvicorn
• —LLM: Google Gemini (google-generativeai)
• —Compression: ScaleDown API with fallback compression
• —Parsing: pypdf for PDF extraction
• —Frontend: React 18 + Vite + Tailwind CSS + Framer Motion + Lucide React + Axios

Repository Structure

Legislative-Analyzer-GenAI/
├── app.py
├── pipeline.py
├── compressor.py
├── llm.py
├── parser.py
├── utils.py
├── test_pipeline.py
├── requirements.txt
├── .env
└── frontend/
    ├── package.json
    ├── vite.config.js
    ├── index.html
    └── src/
        ├── main.jsx
        ├── App.jsx
        ├── index.css
        └── components/
            ├── Navbar.jsx
            ├── Hero.jsx
            ├── AnalysisCard.jsx
            ├── TextTab.jsx
            ├── FileTab.jsx
            ├── StatsRow.jsx
            ├── ResultsCard.jsx
            └── InsightSection.jsx

Prerequisites

• —Python 3.9+
• —Node.js 18+
• —npm 9+

Environment Variables

Create .env in the repository root:

GEMINI_API_KEY=your_gemini_api_key
SCALEDOWN_API_KEY=your_scaledown_api_key

Notes:

• —GEMINI_API_KEY is required for LLM insights.
• —If SCALEDOWN_API_KEY is missing, the app still works using fallback compression.

Installation

1) Backend setup

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2) Frontend setup

cd frontend
npm install
cd ..

Running the Project

Run backend and frontend in separate terminals.

Terminal 1: Backend

source .venv/bin/activate
python app.py

Backend runs at:

• —API base: http://127.0.0.1:8000
• —API docs: http://127.0.0.1:8000/docs
• —Health: http://127.0.0.1:8000/health

Terminal 2: Frontend

cd frontend
npm run dev

Frontend runs at:

• —App: http://localhost:5173

Vite proxy forwards:

• —/analyze -> http://127.0.0.1:8000/analyze
• —/health -> http://127.0.0.1:8000/health

API Usage

POST /analyze

Accepts either:

• —file (multipart upload; PDF or text)
• —text (form field)

Example (file):

curl -X POST http://127.0.0.1:8000/analyze \
  -H "Content-Type: multipart/form-data" \
  -F "file=@sample.pdf"

Example (raw text):

curl -X POST http://127.0.0.1:8000/analyze \
  -F "text=Your policy or bill text here"

Example response:

{
  "status": "success",
  "summary": "Key Changes:\n- ...",
  "num_chunks": 4,
  "chunk_stats": {
    "original_total_tokens": 8400,
    "compressed_total_tokens": 3100
  }
}

Implementation Flow

1) Input parsing (app.py, parser.py)

• —Detects PDF vs text input
• —Extracts and cleans text
• —Rejects empty/unreadable input

2) Chunking (utils.py)

• —Splits text into chunks (default 2000 chars)
• —Uses overlap (default 200 chars)
• —Tries sentence/word boundaries

3) Compression (compressor.py)

• —Attempts ScaleDown compression
• —Falls back to deterministic local compression if API is unavailable

4) Insight generation (llm.py)

• —Sends compressed chunk text to Gemini
• —Enforces strict JSON response shape
• —Formats into 5 sections:
• —Key Changes
• —Who Is Affected
• —Financial Impact
• —Timeline
• —Risks and Concerns

5) Aggregation and refinement (pipeline.py)

• —Deduplicates and prioritizes lines
• —Re-structures final output into sectioned bullets
• —Returns chunk stats and final summary

Frontend Features

• —Two input modes: text and file upload
• —Animated tab switching and transitions
• —Analyze button with loading state
• —Error handling UI
• —Parsed result rendering by sections
• —Stats cards:
• —Chunks Processed
• —Original Tokens
• —Compressed Tokens
• —Compression Rate

Testing

Run the sample pipeline test:

source .venv/bin/activate
python test_pipeline.py

This runs the end-to-end backend pipeline using the sample legislative text in test_pipeline.py.

Build Frontend

cd frontend
npm run build
npm run preview

Troubleshooting

Missing Gemini key

Symptom: warnings about Gemini model not available and weak/no insights. Fix: set GEMINI_API_KEY in .env.

ScaleDown unavailable

Symptom: compression still works but uses fallback behavior. Fix: set SCALEDOWN_API_KEY in .env.

Empty PDF extraction

Symptom: "contains no extractable text" error. Fix: use text-based PDFs (scanned/image PDFs need OCR, not implemented here).

Port conflicts

• —Backend port is configured in app.py (8000)
• —Frontend dev server defaults to 5173

Important Notes

• —Main user interface is the React app served by Vite at http://localhost:5173 during development.
• —Backend remains API-first and serves /analyze, /health, and Swagger docs.