Searchable PDF Service

Convert scanned PDFs to natively searchable PDFs using the Docstrange API (Nanonets). The service extracts text with word-level bounding boxes and embeds invisible text at the correct positions so you can search and copy text using native PDF viewers.

Project structure

• backend/ – FastAPI service and CLI
• frontend/ – React UI for upload, processing, and PDF viewing

Features

• Upload scanned PDFs (max 5 pages for sync processing)
• Extract text + word-level coordinates via Docstrange API
• Embed invisible searchable text layer at correct positions
• Web UI: drag-and-drop upload, output PDF viewer, search (Ctrl+F)
• CLI for batch processing

Prerequisites

• uv (Python package manager)
• Python 3.10+
• Node.js 18+
• Nanonets/Docstrange API key

Setup

Backend

cd backend
uv sync
cp .env.example .env
# Edit .env and set NANONETS_API_KEY=your_api_key_here

Frontend

cd frontend
npm install

Running

Option 1: Full stack (recommended)

Terminal 1 – backend:

cd backend
uv run uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

Terminal 2 – frontend (proxies API to backend):

cd frontend
npm run dev

• Frontend: http://localhost:5173
• API docs: http://localhost:8000/docs

Option 2: Backend only

cd backend
uv run uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

Frontend

The UI provides:

• Left panel: Upload zone (drag & drop or click to browse), Process button
• Right panel: Output PDF viewer and Download button

Use Ctrl+F (Cmd+F on Mac) to search within the embedded PDF viewer.

To point the frontend at a different API URL, set VITE_API_URL when building, or use the built-in dev proxy (defaults to http://localhost:8000).

API

POST /process

Upload a PDF and receive a searchable PDF.

Request: multipart/form-data with file = PDF

Response: Searchable PDF file (application/pdf)

Status codes:

• 200: Success, returns PDF
• 400: Invalid file type, too many pages (>5), or extraction failed
• 401: Invalid API key
• 413: File too large (default max 10MB)
• 429: Rate limit exceeded

GET /health

Readiness check. Returns 503 if NANONETS_API_KEY is not configured.

Configuration

Variable	Required	Default	Description
NANONETS_API_KEY	Yes	-	Docstrange/Nanonets API key
MAX_FILE_SIZE	No	10485760	Max file size in bytes (10MB)
MAX_PAGES	No	5	Max pages (Docstrange sync limit)

CLI Utility

cd backend
uv run searchable-pdf document.pdf -o output.pdf

Tests

cd backend
uv sync --extra dev
uv run pytest tests/ -v

Example (curl)

curl -X POST http://localhost:8000/process \
  -F "[email protected]" \
  -o searchable-document.pdf

AWS Deployment (App Runner)

The project includes a Dockerfile and GitHub Actions workflow for deploying to AWS App Runner.

Local vs production

Environment	API key source
Local	NANONETS_API_KEY env var (from .env or shell)
Production	AWS Secrets Manager (never in config or code)

One-time setup

1. ECR, IAM roles – Already created for this account.

2. For production: Store the API key in Secrets Manager:

   NANONETS_API_KEY=your_key ./deploy/create-secret.sh
   # Saves the ARN – use it for production deploy

3. GitHub Actions uses OIDC (no AWS keys). Optionally add APP_RUNNER_SERVICE_ARN secret for auto-deploy on push.

Deploy

Local/dev (env var – for first-time service creation):

NANONETS_API_KEY=your_key ./deploy/deploy.sh

Production (Secrets Manager):

NANONETS_SECRET_ARN=arn:aws:secretsmanager:... ./deploy/deploy.sh --production

Automatic: Push to main triggers build and push to ECR (and deploy if APP_RUNNER_SERVICE_ARN is set).

Local Docker test

docker build -t searchable-pdf .
docker run -p 8000:8000 -e NANONETS_API_KEY=your_key searchable-pdf
# Open http://localhost:8000