A newer version of the Gradio SDK is available: 6.19.0
metadata
title: CraftPilot
emoji: 🧶
colorFrom: yellow
colorTo: red
sdk: gradio
sdk_version: 6.16.0
app_file: app.py
python_version: '3.10'
pinned: false
license: mit
tags:
- build-small-hackathon
- backyard-ai
- off-the-grid
- llama-champion
- sharing-is-caring
- field-notes
- craft
- vision
- multi-agent
- llama-cpp
- minicpm
- hackathon
- track:backyard
- sponsor:openbmb
- achievement:offgrid
- achievement:llama
- achievement:sharing
- achievement:fieldnotes
datasets:
- skamathramesh/craftpilot-traces
CraftPilot — AI-Powered Craft Business Assistant
Upload a photo of your handmade craft. Get a complete marketplace listing with catalog data, product descriptions, social captions, and fair pricing.
Built for someone I know — a creative introvert who does crochet, painting, embroidery, and sewing but struggles to write product listings and price her work fairly.
Links
- Demo Video: Watch on Google Drive
- Social Post: LinkedIn
- Blog (Field Notes): CraftPilot Blog
How it works
- Upload a craft photo
- Select craft type
- (Optional) Add notes, material cost, time spent
- Click "Analyze My Craft"
The AI runs a multi-agent pipeline, all powered by a single model:
- Vision — describes your craft item in detail
- Cataloger — extracts structured metadata (category, materials, colors, tags)
- Copywriter — generates product title, descriptions, and Instagram captions
- Pricer — suggests fair pricing based on materials, labor, and market rates
Technical Details
- Single model: MiniCPM-V 2.6 (~8B) via llama.cpp — handles both vision and text
- No cloud APIs — everything runs locally
- Streaming results — pipeline streams as each agent completes
- Agent traces — full pipeline transparency in the Agent Traces tab
Badges
- Off the Grid — no cloud API calls
- Llama Champion — all inference via llama.cpp
- Sharing is Caring — agent traces uploaded to HF dataset
- Field Notes — blog post about the build
Built for the Build Small Hackathon 2026 — Backyard AI track.