🧠 Lucid-VSX: Local Ollama Integration for VS Code

This project provides a Copilot-like experience within VS Code using a local or network Ollama API. The extension supports both Chat Participant (chat) and inline code completion (ghost text) features.

🚀 Quick Start

1. Install dependencies:

npm install

2. Compile:

npm run compile

3. Development / Debug (Extension Development Host):

• After opening the project in VS Code, press F5. A new VS Code window (Extension Development Host) will open with the extension loaded.

4. Configure settings (in development host or normal VS Code settings):

{
  "lucid.ollamaEndpoint": "http://<OLLAMA_HOST>:11434",
  "lucid.ollamaApiKey": "llm-...",
  "lucid.ollamaExtraHeaders": { "X-Request-Source": "post_text_script" },
  "lucid.enableInlineCompletion": true,
  "lucid.logUnmaskedHeaders": false,
  "lucid.enableStreamingStatus": false
}

5. Using the extension in a normal VS Code window (packaging):

npm run compile
npm install -g vsce
vsce package
# install the generated .vsix
code --install-extension lucid-vsx-x.x.x.vsix

▶️ Run / Test

• Open the Chat panel in the development host (F5) window and chat with @lucid.
• To test code completion, type in any code file; if lucid.enableInlineCompletion is enabled, ghost text suggestions should appear.

🔐 Environment Variables and Headers

• OLLAMA_EXTRA_HEADERS: Additional headers in JSON format. Example:

export OLLAMA_EXTRA_HEADERS='{"X-Request-Source":"post_text_script"}'

• OLLAMA_API_KEY: API key (can also be provided via lucid.ollamaApiKey setting):

export OLLAMA_API_KEY='tn-llm-...'

The extension automatically adds the Content-Type: application/json header (if not specified in settings) and sends the API key with the X-API-Key header.

🌊 Stream / Chunk Test

Responses from Ollama can be in chunks (NDJSON or line-based). You can simulate streaming by running a test server within the project:

// tiny-stream-server.js
const http = require("http");
http
  .createServer((req, res) => {
    res.writeHead(200, { "Content-Type": "application/json" });
    res.write(
      JSON.stringify({
        choices: [{ message: { role: "assistant", content: "Hello" } }],
      }) + "\n"
    );
    setTimeout(() => {
      res.write(
        JSON.stringify({
          choices: [
            {
              message: {
                role: "assistant",
                content: " How can I help you?",
              },
            },
          ],
        }) + "\n"
      );
      res.end();
    }, 500);
  })
  .listen(8089);

// run: node tiny-stream-server.js

You can test the extension with F5 by running this server and setting lucid.ollamaEndpoint to http://localhost:8089; incoming chunks will be displayed instantly.

📦 Packaging & Deployment

• Create a .vsix with vsce package and install it with code --install-extension.
• If you want to publish to the Marketplace, you can use vsce publish (update package.json metadata before publishing).

🐞 Debugging

• View logs in the Extension Development Host console: Help → Toggle Developer Tools or Debug Console.
• If the server returns a JSON parse error, the extension automatically adds Content-Type: application/json; if you still get an error, check the endpoint path and expected body format.

🛡️ Security Notes

• If you set lucid.logUnmaskedHeaders to true, sensitive headers (e.g., X-API-Key) will be clearly visible in the logs. Keep it disabled in Production.