Yeqi Huang

I tried to bind the extension with different shortcuts, but they all faild. I guess you forgot to bind the action in background.js. Hmm, however, I do not found `open-extesion`...

Interesting findings on M2

5

I tested the latency on M2 Macbook Air. When I tested in Asahi Linux, the performance was not so reasonably.  ```csv ,,,,,,, 67.14945294999993,,,,,,, 67.34945848333324,67.49946433333325,,,,,, 67.41613864999997,67.23281358333334,67.31615300000004,,,,, 178.53200283333322,176.4820619666668,176.41543756666658,174.71548903333348,,,, 177.74884061666643,177.46554656666677,174.28226821666678,174.03230284999995,38.24977780000004,,, 173.86567576666653,173.11570996666654,172.48240873333353,176.34908294999994,39.799796883333315,39.91646433333331,, 176.5824546166666,179.0991352166667,174.7825140666666,175.28253513333323,40.1998201833333,40.199821316666636,39.266493283333304,...

Here's result on [i5-8600T](https://www.intel.com/content/www/us/en/products/sku/129938/intel-core-i58600t-processor-9m-cache-up-to-3-70-ghz/specifications.html)  Min latency: 32.9ns ±0.0 cores: (5,2) Max latency: 35.8ns ±0.0 cores: (1,0) Mean latency: 34.2ns ```csv ,,,,, 35.829233900000006,,,,, 34.96898663333334,34.5728319,,,, 35.10521058333333,35.07162216666667,33.275935483333335,,, 34.02019273333333,33.00294551666667,34.16384906666668,33.78305628333333,, 33.94283296666667,35.17520166666666,32.87415983333332,34.19390786666667,33.41371766666667, ```

Here's some data from [Intel Xeon CPU E5-2695](https://ark.intel.com/content/www/us/en/ark/products/91316/intel-xeon-processor-e52695-v4-45m-cache-2-10-ghz.html?ui=BIG) Command: ```bash core-to-core-latency --csv 30000 1000 > result.csv ``` Min latency: 35.2ns ±0.0 cores: (35,31) Max latency: 161.3ns ±2.3 cores: (20,13) Mean...

Redo action

2

Hmm, I love this extension. And I want to know is there a redo action if I shut some windows by mistakes? (I know there is a shortcut in the...

这个项目的意义并不是否定 GPT-4o 这一产品，事实上，我也承认，GPT-4o 是我使用过用户体验最好的多模态模型。但是，我们必须承认的是，GPT-4o 中，o 所代表的 Omnipotent 并不是真的那么强大。 首先，项目用的都是几个月之前的 API，目的是证明我们完全可以用多模型组合的方案实现多模态效果。 其次，博客里给出了可以达到实时效果的并行方案，如果不使用 API 调用，部分环节自己部署，可以达到实时交互效果。 最后，我希望用这个项目重新解放大家的想象力，项目的每一个环节都可以替换成大家自己的模型，也可以对整个工作流程提出定制化的玩法。 代码： https://github.com/Chivier/easy-gpt4o/ 实现细节和后续思考： [easy-gpt4o - 200行代码实现 GPT-4o 的效果](https://blog.chivier.site/2024-05-15/2024/200-%E8%A1%8C%E4%BB%A3%E7%A0%81%E5%A4%8D%E5%88%BB%E4%B8%80%E4%B8%AA-GPT-4o/)