Flask Application Development

Instructions

Build well-structured Flask applications following these patterns.

Application factory:

• Define create_app(config_name=None) in app/__init__.py to construct the Flask instance
• Register extensions, blueprints, and error handlers inside the factory
• Return the configured app — this enables testing with different configurations and avoids circular imports

Blueprints:

• Group related routes into blueprints: api_bp = Blueprint('api', __name__, url_prefix='/api')
• Register blueprints in the factory: app.register_blueprint(api_bp)
• Each blueprint gets its own module with routes, models, and helpers

Request handling:

• JSON APIs: access request.json (parsed body), request.args (query params), request.form (form data)
• Validate input early; return structured error responses with appropriate HTTP status codes
• Use @app.errorhandler for consistent error formatting across the application

Templates:

• Jinja2 for server-rendered HTML: render_template('index.html', data=data)
• Use template inheritance ({% extends "base.html" %}) for consistent layouts
• Keep logic out of templates — prepare data in view functions

Configuration:

• app.config.from_object('config.ProductionConfig') for class-based config
• app.config.from_envvar('APP_SETTINGS') to override from environment
• Never commit secrets; load sensitive values from environment variables

Background jobs:

• APScheduler for in-process scheduled tasks: scheduler.add_job(func, 'interval', minutes=30)
• Initialize the scheduler inside the application factory; guard against double-start in reloader mode
• For heavy workloads, offload to Celery or RQ instead of in-process scheduling

Integration patterns (Patriot Agent reference architecture):

• Anthropic SDK: anthropic.Anthropic(api_key=key).messages.create() for AI-powered features
• Tavily: TavilyClient(api_key=key).search(query) for web search integration
• feedparser: feedparser.parse(url) for RSS/Atom feed ingestion
• Pattern: Flask serves the API, APScheduler runs periodic data collection, Anthropic processes content, results stored and served via endpoints