2 curated items with deep-dive analysis
@tom_doerr (Tom Dörr)
REST API for sending WhatsApp messages
Stats: 811 likes | 85 RTs Link: https://x.com/tom_doerr/status/2007693904456335449
whatsapp-api is a Node.js-based REST API that enables programmatic WhatsApp message sending by wrapping WhatsApp Web. It sidesteps the official WhatsApp Business API (which requires approval, costs money, and has message template restrictions) by automating the web client directly.
The architecture splits into two main processes: the REST API server handling HTTP requests, and a WhatsApp Web session manager that maintains the authenticated connection. Authentication happens via QR code scanning—the same flow you'd use to link WhatsApp Web manually.
This is essentially web automation packaged as a developer-friendly API. It's the kind of project that fills a gap for small-scale automation, notification systems, or prototyping before committing to official API costs.
Key Features:
- REST API interface: Simple HTTP endpoints for sending messages programmatically
- QR code authentication: Links to your WhatsApp account via the standard WhatsApp Web flow
- Heroku-ready: Built with deployment on Heroku in mind for quick hosting
- TypeScript codebase: Type-safe implementation for easier maintenance
- Insomnia collection included: Pre-built API request collection for testing
Technical Stack:
- Written in TypeScript with Node.js
- Uses WhatsApp Web automation under the hood
- Includes Docker/Heroku deployment configs
For developers building notification systems, chatbots, or automation workflows, the official WhatsApp Business API is often overkill—expensive and bureaucratic for simple use cases. This project offers a pragmatic alternative for personal/small-scale use: connect your own WhatsApp number and send messages via HTTP requests.
Caveat: This relies on unofficial WhatsApp Web automation, so it's subject to breaking if WhatsApp changes their web client. Not suitable for production business use where you need reliability and ToS compliance.
GitHub: https://github.com/felipeDS91/whatsapp-api (324 stars)
@GalaxeaDynamics (Galaxea Dynamics)
We just open-sourced G0 Plus VLA model & launched "Pick Up Anything" demo. See our robot perform diverse real-world tasks through pure language. No specialized training needed. That's zero-shot embodied intelligence.
Stats: 125 likes | 34 RTs Link: https://x.com/GalaxeaDynamics/status/2007665494841799124
G0 Plus is Galaxea Dynamics' latest Vision-Language-Action (VLA) model that enables robots to perform manipulation tasks from natural language instructions without task-specific fine-tuning. The "Pick Up Anything" demo showcases zero-shot generalization—point at an object, describe it in plain English, and the robot executes the grasp.
The architecture follows a dual-system design that separates high-level reasoning from low-level motor control. This mirrors cognitive science models of System 1/System 2 thinking, allowing the VLA to leverage pretrained vision-language understanding while outputting precise action trajectories.
What makes this release notable is the accompanying Galaxea Open-World Dataset—a diverse robotics dataset available in LeRobot format on HuggingFace. This isn't just a model drop; it's a full training pipeline with fine-tuning code and real-robot inference examples.
The zero-shot claim is significant. Most robotic manipulation systems require extensive demonstration data per task. G0 Plus aims to transfer internet-scale vision-language knowledge directly to embodied control, reducing the data collection burden for new environments.
Key Features:
- Zero-shot manipulation: Execute novel pick-and-place tasks from language alone, no per-task training required
- Dual-system VLA architecture: Separates semantic reasoning from action generation for better generalization
- Open-World Dataset: Large-scale manipulation dataset in LeRobot format for community fine-tuning
- Full inference pipeline: Real-robot deployment code included, not just model weights
- Multi-task pretrained: Single model handles diverse manipulation primitives
The VLA space is heating up (OpenVLA, RT-2, π0), but most releases are research artifacts. Galaxea providing inference code, fine-tuning scripts, and training data in a standardized format (LeRobot) makes this immediately usable. If you're building robotic systems and tired of collecting thousands of demos per task, zero-shot VLAs are the path forward—and G0 Plus is now the most accessible entry point.
GitHub: https://github.com/OpenGalaxea/GalaxeaVLA (360 stars)
| # | Author | Type | Topic |
|---|---|---|---|
| 1 | @tom_doerr | github | REST API for sending WhatsApp messages https://t.... |
| 2 | @GalaxeaDynamics | github | We just open-sourced G0 Plus VLA model & launc... |
Generated: 2026-01-04T21:20:01.913176