GDSS Maverick IMDB Auto-Fill

Rust backend plus Svelte frontend for extracting Item Master Database fields from product package images.

Backend

Pick one provider.

OpenRouter:

export AI_PROVIDER=openrouter
export OPENROUTER_API_KEY=sk-or-...
export OPENROUTER_MODEL=openai/gpt-4.1-mini

Optional OpenRouter attribution:

export OPENROUTER_SITE_URL=https://your-app.example
export OPENROUTER_APP_NAME="Maverick IMDB Auto-Fill"

OpenAI direct:

export AI_PROVIDER=openai
export OPENAI_API_KEY=sk-...
export OPENAI_MODEL=gpt-4.1-mini

Run locally:

cargo run -- serve --host 127.0.0.1 --port 8080

Health check:

curl http://127.0.0.1:8080/api/health

Extraction endpoint:

curl -X POST http://127.0.0.1:8080/api/extract \
  -F "images=@Hackathon Materials/product images/S221234199_550719011.jpg" \
  -F "images=@Hackathon Materials/product images/S221234199_550719013.jpg"

The backend uses:

• Tesseract for local OCR.
• zbarimg for local barcode scanning.
• OpenAI Responses API or OpenRouter Chat Completions for image + evidence extraction.
• Strict Structured Outputs for the 10 IMDB fields.

API docs:

• Responses API: https://platform.openai.com/docs/api-reference/responses/create
• Images and vision: https://platform.openai.com/docs/guides/images-vision
• Structured Outputs: https://platform.openai.com/docs/guides/structured-outputs
• OpenRouter Chat Completions: https://openrouter.ai/docs/api/api-reference/chat/send-chat-completion-request
• OpenRouter image inputs: https://openrouter.ai/docs/guides/overview/multimodal/image-understanding
• OpenRouter structured outputs: https://openrouter.ai/docs/guides/features/structured-outputs

Frontend

cd frontend
cp .env.example .env
npm install
npm run dev

Set frontend/.env:

VITE_API_BASE_URL=http://localhost:8080

Deploy

Backend:

• Render Docker web service.
• Use Dockerfile and render.yaml.
• For OpenRouter, set AI_PROVIDER=openrouter and add OPENROUTER_API_KEY as a secret env var.
• For OpenAI direct, set AI_PROVIDER=openai and add OPENAI_API_KEY as a secret env var.

Frontend:

• Vercel.
• Set VITE_API_BASE_URL to the Render backend URL.

Output Schema

The live app returns these 10 IMDB fields:

barcode
category_type
segment_type
manufacturer
brand
product_name
weight_and_unit
packaging_type
country_of_origin
marketing_messages

Unknown fields remain empty strings.