chushulsuri/hermaquette — explained in plain English
Analysis updated 2026-05-18
Describe a physical object in plain language and get back a colored, printable 3D model.
Automatically check and repair a generated 3D model so it can actually be manufactured.
Get a real vendor price quote for a 3D printed object and process payment through Stripe.
Track an order after payment and get automated delivery quality checks from photos.
| chushulsuri/hermaquette | 0xradioac7iv/tempfs | 7vignesh/pgpulse | |
|---|---|---|---|
| Stars | 0 | 0 | 0 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | hard | moderate | moderate |
| Complexity | 5/5 | 3/5 | 4/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Needs Docker Compose, a git submodule, and several third party API keys (fal.ai, Stripe, OpenAI, Sculpteo).
Hermaquette is a hackathon project that turns a written description of a physical object into a colored, 3D printable model, using a team of three AI agents built on a framework called Hermes to do the work from start to finish. You type a sentence describing a non electronic object, and the system researches it, creates concept images, builds a full color 3D model, checks and repairs the model so it can actually be printed, gets a real price quote from a manufacturing vendor, takes a payment through Stripe, and manages what happens after the order is placed. The three agents each have a role. The first, called Hermaquette, is the orchestrator that understands the request, generates concept images, and hands off work to the other two. The second, called Sculptor, builds the full 3D colored model using a service called fal.ai, then runs a repair loop until the design is manufacturable, or marks it as blocked if it cannot be fixed. The third, called Follow-up, tracks the order after payment, checks delivered photos against the design using GPT vision, and drafts any reprint or refund request for a human to approve, since nothing gets sent automatically. The process runs in stages: describing the object, picking a generated concept, building and repairing the 3D model, getting a vendor quote plus a service fee, and finally paying through Stripe in test mode with a human approving any vendor spending. Every step in between is logged in a SQLite database that acts as the single source of truth for the whole order. The project is built with a Next.js web app, a Python service for mesh repair, and a Hermes agent service, all run together using Docker Compose, with API keys needed for the various third party services it depends on. The project's own code is released under the MIT license, though it also includes a separate open source component under its own upstream license. It was built for the Hermes Hackathon 2026.
An AI agent pipeline that turns a text description into a priced, printable 3D model, handling design, payment, and order tracking end to end.
Mainly TypeScript. The stack also includes Next.js, TypeScript, Python.
The project's own code can be used freely including commercially, but a bundled submodule keeps its own separate license.
Setup difficulty is rated hard, with roughly 1h+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.