Inspiration
Across maker studios, community art collectives, and educational workshops, there is an accelerating movement toward circular, non-toxic creativity: crafting organic crayons, artisanal watercolors, and soft pastels from local agricultural waste (discarded yellow onion skins, avocado pits from restaurants, pruned woad leaves, and walnut husks) and natural earth clays.
However, artisanal makers quickly run into a costly roadblock: natural materials are chemically unpredictable. A raw lot of onion skins might retain 8% moisture one week and 18% after a rainstorm. This uncalibrated moisture drastically skews dry pigment yields. In the workshop, when wax-to-pigment ratios are off by even a few percent, whole batches of expensive cosmetic-grade carnauba wax ($45/kg) and filtered beeswax ($30/kg) seize, crack, or fail to lay down on paper—ruining hundreds of dollars in raw materials and wasting hours of labor.
Makers cannot solve this by feeding proprietary craft formulas, supplier manifests, or color-mixing curves into public cloud AI chatbots. Craft studios need 100% on-premises data sovereignty, reliable local storage, and instant offline computation right where the materials are weighed and melted. When we saw the UGREEN Hackathon 2026 on AI NAS, the opportunity became crystal clear: turn the UGREEN AI NAS into an autonomous, on-device formulation engine that eliminates recipe guesswork and prevents ruined batches right inside the studio.
What it does
BioChroma NAS Formulator is an autonomous on-device AI Skill for UGOS Pro and the Uliya AI Assistant. It transforms the UGREEN AI NAS into an active studio production assistant that converts local biomass inventory into standardized, non-toxic artistic supplies with zero cloud dependency. Key capabilities include: 1. Dynamic Moisture Compensation: When a maker specifies an inventory lot (e.g., onion_batch_01), the skill reads the recorded moisture percentage and dynamically calculates the effective extraction yield: Yieldeff=Cyield×(1−100Moisture%) It then outputs the exact weight of raw biomass to harvest. 2. Stoichiometric Binder Ratios: Computes precision mass breakdowns for 3 craft mediums: ◦ Organic Crayons: 35% bio-pigment, 45% beeswax, 10% carnauba wax, 10% plant oil (melt window: 65°C–72°C). ◦ Artisanal Watercolors: 50% bio-pigment, 35% gum arabic, 10% raw honey, 5% distilled water (mull at 20°C–25°C). ◦ Soft Drawing Pastels: 60% bio-pigment, 25% precipitated calcium carbonate, 15% methylcellulose binder. 3. Mandatory Human-in-the-Loop Safety Gate: Before writing any file or updating historical records, the skill pauses and displays a complete Bill of Materials (BOM) review card in the Uliya chat interface. Disk writes occur only after the artisan explicitly approves. 4. Structured Batch Production Sheets: Emits printable Markdown production records to /volume1/BioChroma/batches/ complete with physical studio checklists, thermal safety parameters, and QC sign-offs. 5. Knowledge Base Archival: Automatically indexes batch numbers, yields, and lot IDs into the local SQLite knowledge base via knowledge-agent, allowing artisans to ask Uliya natural-language questions like "How much onion skin did we need for our October crayon run?"
How we built it
BioChroma NAS was developed strictly under the Uliya Skill Development Standard and Capability Catalog v2.0.0: • Pure Markdown Orchestration (SKILL.md): The skill operates as a high-level orchestrator without executing unauthorized host commands. It delegates system actions cleanly to registered UGOS expert agents: ◦ file-agent (ug-filesystem, ug-filemgr): Reads local inventory manifests and manages file storage. ◦ document-agent (ug-document): Transforms formulation data into structured Markdown batch sheets. ◦ knowledge-agent (ug-knowledge): Indexes completed batch logs for semantic on-device search. • Biochemical Formulation Engine: We researched and modeled extraction coefficients and thermal processing limits for 7 raw materials: yellow onion skins, avocado seeds, madder root, woad/indigo leaves, black walnut hulls, yellow ochre, and red ferruginous clay. • Automated Verification & Static Compliance: ◦ Created check-skill.mjs to audit package naming, frontmatter schemas, balanced code fences, required headings, and link targets (0 Errors, 0 Warnings). ◦ Implemented unit test suite (test_biochroma.py) validating all mathematical stoichiometry and moisture guards. • Publication-Grade PDF Generation: Built a custom vector PDF compilation script using pycairo (build_proposal_pdf.py) to generate a 2-page, 443-word proposal document matching all 10 official Devpost questionnaire items.
Challenges we ran into
1. Non-Linear Moisture Attenuation: Raw agricultural waste varies widely in water content. Calculating the exact biomass to harvest required deriving a robust moisture-adjusted yield curve so that dry pigment mass remains consistent regardless of whether biomass arrives at 8% or 18% moisture.
2. Strict Orchestration Constraints: Under UGREEN's Uliya Skill Standard, custom scripts, daemons, and direct binaries are prohibited in submission packages. Adapting our logic into pure Markdown instructions that delegate to file-agent, document-agent, and knowledge-agent required meticulous mapping against Capability Catalog v2.0.0.
3. Designing a Foolproof Safety Gate: Artisanal studios cannot afford accidental file overwrites or premature execution. Implementing the synchronous GATE: REQUIRE_USER_APPROVAL ensures that every formulation must be reviewed and confirmed by the human maker before any file is saved.
4. Adhering to the 500-Word Proposal Cap: The hackathon rules enforce a strict 500-word limit across 10 detailed questions. Condensing our technical architecture, user flow, hardware testing strategy, and risk mitigation into 443 substantive words demanded multiple rounds of precise editing.
Accomplishments that we're proud of
• 100% Static Check Compliance: Verified with check-skill.mjs with 0 errors and 0 warnings, adhering strictly to Uliya Skill Development Standard v2.0.0.
• Tangible Studio ROI: Saves an estimated $350 to $600 per production run by eliminating scrapped batches of costly beeswax and carnauba wax.
• Pioneering Circular Bio-Manufacturing on NAS: The first AI Skill on an AI NAS platform connecting local household/agricultural waste streams with precision, non-toxic art supply manufacturing.
• Complete Data Sovereignty: Zero cloud telemetry, zero external API keys, zero subscription costs. The entire workflow operates offline on the local studio network.
What we learned
• The NAS as an Intelligent Studio Hub: An AI NAS is vastly more powerful than a simple backup drive—it functions as an always-on, intelligent studio appliance that indexes workshop inventories, orchestrates physical manufacturing workflows, and protects intellectual property.
• The Value of Declarative Orchestration: The Uliya Skill Standard demonstrates that separating high-level agentic orchestration from low-level system execution creates safer, more auditable, and modular AI capabilities.
• Cross-Disciplinary Synergy: Combining organic chemistry heuristics with on-device local language models unlocks practical solutions for sustainable makers that neither could achieve alone.
What's next for BioChroma NAS Formulator
• Round 2 Testing on UGREEN AI NAS iDX6011 Pro: Upon advancing to Round 2, we will deploy the skill to our assigned remote hardware to benchmark local NPU inference latency (targeting <500ms prompt-to-sheet generation) and validate NVMe/SATA RAID storage throughput.
• Automated Visual QC with photo-agent (ug-photo): Integrate computer vision models on the NAS to inspect macro photos of poured crayon sticks and dried watercolor pans, automatically grading pigment dispersion, surface blooming, and color saturation.
• IoT Studio Sensor Integration: Connect studio dehydrators and drying racks via local webhooks so real-time moisture sensor readings automatically update /volume1/BioChroma/inventory.csv.
• Community Recipes on UGREEN Skills Hub: Enable artisans to package non-proprietary bio-pigment recipes (e.g. "Avocado Terracotta" or "Woad Indigo") to share with the global Uliya community.
Built With
- antigravity
- gemini
- uliya-skill-kit-0.7.7-en
