Commit History
-
fix: Add image placeholder to prompt for multimodal model
Danmoreng3308917 -
fix: Pass dummy 1x1 image to ImageToText processor
Danmorengefe7a41 -
fix: Correct ONNX wasmPaths CDN URL
Danmoreng862fb05 -
fix: Use ImageToText model for text-only chat
Danmoreng73f6156 -
fix: Revert model ID to Ministral-3B and re-add quantization config
Danmorengb10d0f1 -
feat: Refactor to simple HTML/JS chat application
DanmorengRefactored the application from a React-based webcam captioning tool to a simple, vanilla JavaScript and HTML chat application. - Removed all React components, hooks, and dependencies. - Replaced the video captioning UI with a text-based chat interface. - Created a new src/model.js to encapsulate the model loading and inference logic from Transformers.js. - Implemented a new src/main.js to handle the chat UI, user input, and communication with the model. - Updated package.json to remove React/TypeScript dependencies and simplify scripts for a Vite-based HTML/JS project. - Added a .gitignore file.a3a2aed