Juan Manuel García

Crawley is an open-source crawling and scraping framework for Python, and the project I am best known for — around 190 stars on GitHub. It lets you crawl sites and pull structured data out of them through a small declarative API: fetch a page, query it with CSS or XPath selectors, and persist what you found through SQLAlchemy. I first wrote it on non-blocking I/O back when that meant eventlet and greenlets, and I have since rewritten it on asyncio and httpx with SQLAlchemy 2.x, a documentation site, and continuous integration.

PyFb is an open-source Python client for the Facebook Graph API, with around 70 stars on GitHub. Instead of handing back raw JSON dictionaries it returns Python objects, so reading a profile, a friend list, or a feed takes a couple of lines. It also covers the OAuth flow — the redirect, the code exchange, and the access token — and ships a worked example of adding Facebook login to a Django project with a few settings and two views.

A code generator that writes the SQL data access layer for a database so you do not have to, without adding an ORM as a dependency. It reads your tables and produces plain Python classes with the CRUD operations for each one, going through the DB-API directly. It exists for projects where SQLAlchemy or the Django ORM would be a heavy dependency for a handful of queries. An early version built its SQL by string formatting; when users pointed out that this allowed injection, I moved it to bound parameters and let the driver escape the values. Around 50 stars on GitHub.

A real-time chat room built on Node.js and Socket.IO, written as a readable example of pushing messages between browsers over WebSockets instead of polling for them. The server tracks the connected clients, broadcasts each message as it arrives, and handles joins and disconnects; the page updates with no reload. It is deliberately small — one server file and one page — so it can be read end to end and used as the starting point for real-time features in a bigger app.

A worked example of putting a Backbone.js front end on a Django back end through a Tastypie REST API. It shows the parts that are usually left as an exercise: exposing the models as REST resources, wiring Backbone models and collections to those endpoints, and getting create, update, and delete to round-trip cleanly between the browser and the server. I wrote it when single-page front ends on top of Django were new and that integration was the hard part of every project.

A routing plugin that replaces Django's urls.py with convention over configuration. It uses introspection to infer views from the file and class names in the project, so adding a view is enough to make it routable, and it renders templates by the same convention — there is no URL table to keep in sync by hand. It supports class-based views on Python 2.7 and 3.2+, installs as a normal package, and ships a test suite. I built it after maintaining URL files that were almost entirely repetition.

DeployCloud is a deployment platform I designed and run on my own servers: you give it a git URL or a directory and it gives back a running HTTPS app. It builds the image, issues the certificate, starts the release behind a reverse proxy, and keeps a preview environment per branch. It also covers what usually gets bolted on later — scheduled jobs, managed PostgreSQL, MySQL and Redis add-ons, off-site backups, metrics and alerts, one-off commands, rollbacks — and an automated reviewer that reads what strangers deploy on the free tier and blocks proxy nodes and miners. It ships a CLI published on npm and an MCP server, so an AI agent can deploy and operate apps on it without a browser.

A portfolio of Shopify apps I build, sell, and operate myself, from the store listing to the servers they run on: automatic product tagging, reporting and exports, scheduled publishing, metafield editing, discount rules, shipment tracking, and an embedded chat. Each one is a full product — OAuth install, billing, webhooks, background workers, and a theme or admin extension — and they run on infrastructure I keep myself, so I am also the person who gets the alert when a queue backs up or a disk fills. Working on them is what taught me Shopify's APIs, its rate limits, and what merchants actually complain about.