The compact Reachy Mini is designed for accessibility and engagement. | Supply: Hugging Face
This week, Hugging Face launched an agentic toolkit that lets anybody construct a working app for Reachy Mini, its open-source desktop robotic. The toolkit works in beneath an hour, with out writing a single line of code. As an alternative, an AI agent writes all of the code.
To make use of the toolkit, customers describe the conduct they’d wish to see in plain English, and the agent writes, assessments, and ships the code to the robotic.
“For 60 years, robots have been constructed by roboticists. As of right now, they are often constructed by anybody,” mentioned Clément Delangue, co-founder and CEO of Hugging Face. “When the software program is open-source, and an AI agent can write the code, the gating that used to return from technical data simply disappears.”
For the whole historical past of robotics, three issues stood between an thought and a working robotic: experience, costly {hardware}, and weeks of integration work. Hugging Face mentioned it goals to break down all three. An AI agent replaces the experience. The {hardware} is a low-cost, open-source desktop robotic anybody can purchase. The combination is a one-click move on a web site thousands and thousands of builders already use.
Hugging Face is an open platform for AI builders, usually referred to as “the GitHub of AI.” Hundreds of thousands of builders and tens of hundreds of corporations use it to share AI fashions, datasets, and purposes. In 2024, Hugging Face acquired Pollen Robotics, the French maker of the Reachy line of open-source robots.
Meet considered one of Hugging Face’s early makes use of
Joel Cohen is a 78-year-old retired advertising govt within the Raleigh-Durham space who runs CEO peer teams. He’s not a developer, and he has by no means labored in robotics. Cohen additionally colorblind and wears listening to aids. Hugging Face mentioned it took him over two weeks to assemble his Reachy Mini Lite (it often takes round 3 hours).
Then, Cohen constructed an app. He created a voice-controlled AI co-facilitator for the CEO peer teams he runs on Zoom. Reachy Mini sits on his desk. The robotic is 11 in. (27.9 cm) tall and 6.3 in. (16 cm) vast, and it weighs a mere 3.3 lb. (1.5 kg).
When Cohen says, “Hey, Reachy,” it wakes up, listens, and responds. It has a character, which Cohen calls his “VP of future pondering.” The system additionally has 4 facilitation modes, a financial institution of over 60 questions, and greets every of his 29 group members by title.
Mid-session, it may possibly hot-seat a member, push again on a surface-level reply, generate a contemporary query on the spot, or summarize the important thing themes earlier than closing.
“I constructed this by describing what I wanted in plain English. Claude wrote the code,” mentioned Cohen. “No SDK. No robotics background. No developer expertise.”
App retailer for Reachy Mini is now open
Reachy Mini’s apps reside on the Hugging Face Hub, which is searchable, forkable, and installable with one click on. If you happen to see an app you want, you’ll be able to duplicate it, ask the AI agent to alter it, and publish your model. The firm mentioned it takes simply minutes to create a brand new app.
As well as, each Hugging Face app additionally runs in a browser-based simulator, so anybody can play with its catalog with out proudly owning the {hardware}.
The Reacy Mini App Retailer has greater than 200 apps and is offered now. Customers can construct their very own apps by asking a machine studying intern or their favourite agent. Among the apps embody:
- Joel’s Co-Facilitator — voice-controlled session companion for CEO peer teams
- Language tutor — listens to the consumer to assist modify talking accents
- Emotional Injury Chess — performs chess and reacts to each transfer, dropping its head on a blunder (“Oh no! Large mistake!”) and cheering on a profitable mixture
- Reachy Cellphone House — anti-procrastination mode detects when customers decide up their telephones and calls them again to work
- Purple Mild, Inexperienced Mild — the Squid Sport kids’s model, with the robotic enjoying the doll
- F1 race commentator — calls Components 1 races as they occur, reside from the desktop
- Cook dinner assistant — walks by a recipe step-by-step, hands-free
- Coding instructor — teaches children to program in a simplified scripting language
- An workplace receptionist that Hugging Face co-founder and CEO, Clem Delangue, inbuilt beneath two hours
