How nicholaschen09/font-llm works
font-LLM is a prototype web app that accepts an image of typography and returns ranked font matches with visual evidence, alternatives, and next steps. The main workflow runs from image selection and preview in the browser, through a validated API upload and Gemini vision analysis, back to a rendered results view. The API also supports configurable model fallback. The README describes future OCR and glyph matching as possible growth, not current implemented components; the graph reflects the sampled runtime code.
Main entry points: route.ts Analysis endpoint · page.tsx Image upload
Components 7
Web interface
- Image upload
app/page.tsx - Results view
app/page.tsx
Analysis API
- Analysis endpoint
app/api/analyze-font/route.ts - Image validation
app/api/analyze-font/route.ts - Model fallback
app/api/analyze-font/route.ts
Other components
- Designer
- Gemini vision API
Connections 8
- Designer → Image upload: selects image
- Image upload → Analysis endpoint: POST image
- Analysis endpoint → Image validation: validates upload
- Analysis endpoint → Model fallback: runs analysis
- Model fallback → Gemini vision API: tries models
- Gemini vision API → Model fallback: returns analysis
- Analysis endpoint → Results view: returns response
- Results view → Designer: shows matches