Iron Symptom Image Generation Task¶

Status: Ready to generate with MCP Playwright inside Docker container

Context¶

Generate 6 Iron deficiency symptom images from NIH factsheet using ChatGPT DALL-E. Using incremental workflow: Generate ONE → Verify → Proceed with remaining 5.

Environment¶

Tailscale Exit Node: Norway (sjs-pi4-02) - ACTIVE ✅
ChatGPT Login: Google OAuth (codiedev42@gmail.com)
Credentials: /gdrive/.nutrie-secrets
Mascot Reference: /work/nutri-e/website/images/nutri-e-mascot.png

6 Iron Symptoms from NIH Factsheet¶

Source: https://ods.od.nih.gov/factsheets/Iron-Consumer/

GI Upset - "GI upset"
Weakness - "weakness"
Tiredness - "tiredness"
Lack of Energy - "lack of energy"
Problems with Concentration and Memory - "problems with concentration and memory"
Reduced Ability to Fight Infections - "less able to fight off germs and infections"

Image Requirements¶

Format: Vertical 9:16 (1024x1536 pixels) for TikTok/social media

Character: Purple Nutri-E mascot (egg-shaped, big eyes, soft features)

Background: Orange gradient (brand style)

Text (MUST be clearly visible): - Top: "Iron Deficiency" - Middle: [Symptom name, e.g., "GI Upset"] - Bottom: "Source: NIH Office of Dietary Supplements"

Style: Cute 3D render, mascot fills 60-70% of frame

Workflow (CRITICAL - Incremental)¶

1. Generate GI Upset image first (TEST)
2. Verify quality:
   ✓ Mascot design matches reference
   ✓ All text clearly visible
   ✓ NIH citation present
3. If PASS → Generate remaining 5 symptoms
   If FAIL → Fix prompt, regenerate test

Output Directory¶

/tmp/iron-symptoms/ or /output/iron-symptoms/

Next Steps for Claude Code Inside Container¶

Use MCP Playwright tools to navigate to ChatGPT
Upload mascot reference image
Generate ONE test symptom (GI upset) with prompt template above
Verify output quality
Proceed with remaining 5 if test passes

Why This Approach¶

Cost Protection: 1 test image before committing to all 6
Quality Control: Catch issues early
Brand Consistency: Mascot reference ensures correct design