Zhongheng-Cheng
commented
3 months ago - replaced old LLM script with new script.
- got rid of llama-index.
- only used Gemini models to process image/video/audio and generate text responses.
- added image/video analysis, returning json format diagnose.
- added audio transcripting, returning text format transcript.
- manually track multimodal chat history.
- updated web ui