Feature: Import SAM 3 Web Library & Basic Inference Tests

[blayyyyyk]

Description

As an ML engineer, I want to import the SAM 3 web-compatible library and run basic inference tests so that we can verify the model executes correctly in the browser environment.

Acceptance Criteria

• Select and configure the browser inference engine capable of running SAM 3 (e.g., ONNX Runtime Web).
• Import the base SAM 3 model weights into the web app's public assets.
• Write a test script to pass a dummy prompt (coordinate point or bounding box) to the model.
• Successfully log the inference output (mask array) to the browser console without throwing WebGL/WASM memory errors.

Implementation Notes & Resources

• Exporting the standard PyTorch weights to ONNX format will likely be required for efficient ONNX Runtime Web execution.
• Ensure the prompt logic handles SAM 3's specific required tokens/concept prompts so it doesn't fail on basic coordinate inputs.
• Resource: ONNX Runtime Web Documentation