In today's era of rapid technological development, every breakthrough in the field of artificial intelligence can make waves in the technological circle. The cloud3.5 model released by foreign AI giants and shopic companies once again refreshed the entire technology circle. Why? That is, it can transform an ordinary computer into a real "Jarvis" with API and general artificial intelligence.
Now you just need to tell the computer a sentence as usual, such as "Find flights from Seattle to San Francisco next Tuesday to Thursday," and the computer will automatically open the flight search website and help you enter your address and information. search. The flight information you want, isn't this the real-life version of Iron Man Jarvis, it's amazing.
But how does such a powerful "Jarvis" do it? Fortunately, UNSHI has released its implementation source code. Let’s look at the specific steps.
The successful launch of the project is only the first step. What is more worthy of discussion is its implementation principle. The entire project uses a cross-platform desktop application development framework - ELE (there may be typos in the original text here, presumably it is a development framework such as Electron). We open the main folder in the SRC folder. Stored here is the main process code of the Electron application, which is mainly responsible for creating and managing windows. Stored in the render folder is the code for rendering the corresponding square in the upper left corner. The button inside is the "Let's go" (start task button) on the page.
When we click the start task button, the "lang agent" method is triggered (this may also be a specific term in the material, or it may be a spelling mistake such as "lang agent"), which makes the computer have Jarvis thinking. Core approach to capabilities. It first obtains the user's chat content, and then calls ENOPIC's latest cloud3.5 dual-network model, which has powerful human-computer interaction capabilities. Afterwards, the model learns the current user's screen content by taking screenshots, and the model returns the corresponding action instructions. These action instructions are very important. It was they who turned the computer into a working "JARVIS".
For example, the "moss move" command tells the computer that the mouse should move to the specified XY axis coordinate position; the "left-click jack" command tells the computer to right-click and drag the current content to the specified coordinate position. In addition, there are cursor position instructions, left mouse click instructions, right mouse click instructions, middle mouse click instructions, mouse double click instructions, keyboard key instructions, etc. After each instruction is executed, the screenshot will be retaken, and The cloud3.5 model is notified so that it can continue to issue instructions to complete the user's task until the final model determines whether the task is completed. Once completed, FINNISH will be returned. This type of instruction indicates that the task has been completed and the entire program will terminate.
How about it? Isn't it interesting? Interested students can download the code and develop their own Jarvis. If there is anything you don’t understand during the process, you can also write it in the comment area and discuss it together. I hope you can gain something from your exploration of programming, and don’t forget to give me a follow so that you can learn more about cutting-edge programming knowledge. See you next time.
What do you think of Jarvis' development? Come and share your thoughts in the comment section. If you think this article is good, please share it with your friends who are also interested in programming.
Share on Twitter Share on Facebook
Indie123
Comments
There are currently no comments