[教學] 使用 LM Studio 大語言模型幫助 XCode 編碼工作

javacomhk

1.  首先根據上一篇教學 看看所需硬件要及安裝 LM Studio and 大語言模型 例如 DeepSeek-R1

2.  在 LM Studio 將下載的模型 Load 入RAM 後及在第二個的 Developer Tab 開啟 server 如下


3. 開啟 Terminal 打入以下 code 以測試  LM Studio server

1. curl -X POST http://localhost:1234/v1/chat/completions \
   
2. -H "Content-Type: application/json" \
   
3. -d '{
   
4.   "messages": [
   
5.     { "role": "user", "content": "Write a hello world program in swift" }                        
   
6.   ]
   
7. }'

複製代碼

4. 測試 server 成功後便可以在 Xcode 創建 Xcode Source Editor Extension 步驟如下 (reference from ChatGPT)
Develop an Xcode Source Editor Extension
a. Create a New Project:

Open Xcode and create a new macOS App project (e.g., XcodeLLMExtension).
b. Add a Source Editor Extension Target:

In your project, add a new target by selecting File > New > Target.
Choose Source Editor Extension under the macOS section and name it (e.g., LLMCodeAssistant).
c. Configure the Extension:

In the extension's Info.plist, set the following keys:
NSExtensionPointIdentifier to com.apple.xcode.source-editor
NSExtensionPrincipalClass to $(PRODUCT_MODULE_NAME).SourceEditorCommand
d. Implement the Command:

In SourceEditorCommand.swift, implement the perform function to interact with LM Studio's API.

1. import XcodeKit
   
2. 
   
3. class SourceEditorCommand: NSObject, XCSourceEditorCommand {
   
4.     func perform(with invocation: XCSourceEditorCommandInvocation, completionHandler: @escaping (Error?) -> Void) {
   
5.         // Retrieve selected code
   
6.         guard let selection = invocation.buffer.selections.firstObject as? XCSourceTextRange else {
   
7.             completionHandler(nil)
   
8.             return
   
9.         }
   
10. 
    
11.         let selectedLines = invocation.buffer.lines
    
12.             .subarray(with: NSRange(location: selection.start.line, length: selection.end.line - selection.start.line + 1))
    
13.             .compactMap { $0 as? String }
    
14.             .joined(separator: "\n")
    
15. 
    
16.         // Prepare the prompt for LM Studio
    
17.         let prompt = "Analyze and optimize this Swift code:\n\n\(selectedLines)"
    
18. 
    
19.         // Call LM Studio API
    
20.         callLMStudioAPI(prompt: prompt) { response in
    
21.             if let response = response {
    
22.                 // Insert the response into the editor
    
23.                 invocation.buffer.lines.add("\n// LLM Suggestion:\n\(response)")
    
24.             }
    
25.             completionHandler(nil)
    
26.         }
    
27.     }
    
28. 
    
29.     private func callLMStudioAPI(prompt: String, completion: @escaping (String?) -> Void) {
    
30.         let url = URL(string: "http://localhost:1234/v1/chat/completions")!
    
31.         var request = URLRequest(url: url)
    
32.         request.httpMethod = "POST"
    
33.         request.addValue("application/json", forHTTPHeaderField: "Content-Type")
    
34. 
    
35.         let body: [String: Any] = [
    
36.             "messages": [["role": "user", "content": prompt]],
    
37.             "temperature": 0.7
    
38.         ]
    
39.         request.httpBody = try? JSONSerialization.data(withJSONObject: body)
    
40. 
    
41.         let task = URLSession.shared.dataTask(with: request) { data, _, _ in
    
42.             if let data = data,
    
43.                let json = try? JSONSerialization.jsonObject(with: data) as? [String: Any],
    
44.                let choices = json["choices"] as? [[String: Any]],
    
45.                let message = choices.first?["message"] as? [String: Any],
    
46.                let content = message["content"] as? String {
    
47.                 completion(content)
    
48.             } else {
    
49.                 completion(nil)
    
50.             }
    
51.         }
    
52.         task.resume()
    
53.     }
    
54. }

複製代碼

Ensure that the callLMStudioAPI function matches the API specifications provided by LM Studio.

5. Enable and Test the Extension

a. Enable the Extension:
Build and run your macOS app.
Navigate to System Settings > Privacy & Security > Extensions and enable your Xcode extension.

b. Test in Xcode:
Open a Swift file in Xcode.
Select the code you want to analyze or refactor.
Go to Editor > LLMCodeAssistant to invoke the extension.
The AI-generated suggestions will appear as comments in your code.

By following these steps, you can harness the power of LM Studio within Xcode, enhancing your coding experience with AI-driven insights and assistance.


6 Possible Enhancements
Add different commands for:
Code explanation
Bug detection
Autocompletion
Use LLM fine-tuning for specific Swift best practices.
Implement a UI dialog for user input before sending a request.