Image Pipeline Setup — March 5, 2026

What

End-to-end image generation pipeline for all WiderWings projects.
Generate AI images → auto-resize for purpose → upload to Supabase Storage → get embeddable URLs.

Location

• Skill: ~/clawd/skills/image-pipeline/
• Generator: ~/clawd/skills/image-pipeline/scripts/generate.py
• Uploader: ~/clawd/skills/image-pipeline/scripts/upload.py

Providers

• Primary: OpenAI gpt-image-1 (ready, API key in env)
• Secondary: Nano Banana 2 / Gemini 3.1 Flash Image (needs inference.sh account)

Supabase Storage Buckets Created

• MedSchools: images bucket (public) at ohkdiwblocbrcfrhxeyg
• WiderWings: images bucket (public) at bmwbwyptqiiqolgvbhsw

Agents with Access

• Kai (frontend) — blog heroes, thumbnails, OG images for UI
• Maya (content) — blog post images, social sharing images
• Sam (designer) — design assets, avatars, icons

Purpose Presets

• og: 1200×630 PNG (social share)
• blog-hero: 1200×630 + 400×210 thumb WebP
• thumbnail: 300×300 + 600×600 WebP
• avatar: 128×128 + 256×256 WebP
• icon: 64×64 + 128×128 PNG
• full: original size WebP

Pending

• Nano Banana 2: needs inference.sh account login (infsh login)
• Could add Hedge-specific image needs later