Personal Copilot

This tool is an experimental extension for the Visual Studio Code IDE. It is designed for developers and operations professionals. The extension helps with generating code, finding issues and vulnerabilities, summarizing code, creating documentation, and doing git code reviews before committing changes.

The extension uses different APIs for large language models (LLMs). Currently, it only works with locally hosted open-source APIs like Llama2 and Mistral. This gives the extension a wide range of features and functions. Since the LLMs are hosted locally, it works even when you are not connected to the internet.

Please note that this is an experimental version resulting from my learning and experimentation with Local LLMs. Therefore, you should use it at your discretion and at your own risk.

Features

• Code Generation
• Code Documentation
• Code Summary
• Find Code Problems and Vulnerabilities

Prerequisites

Configuring the Local LLM, which can be easiliy done by following my article on medium: https://medium.com/@uppadhyayraj/exploring-local-large-language-models-and-associated-key-challenges-93618691fc9b

On highlevel following are the exact requirements to getting started:

• LM Studio Installed on machine
• In adition to model mentioned in the article above, Download and configure "TheBloke/CodeLlama-7B-Instruct-GGUF" model in LM Studio for peronal copilot
• Last additional step in LM studio is to start the local server, here you have select the <-> icon on the left side bar and then on new page click "Start Server" button. It will start the server and end point will be shown on screen (default is http://localhost:1234/v1)

Configuring the Extension

Open VS Code settings, and search for "Personal-Copilot", to find the settings required for this extension. Key settings are:

• Provider (default Local)
• LLM Enpoint; Default configured with LM Studio Endpoint (http://localhost:1234/v1)
• Default Promopts for each tasks and can be updated if required

How to use the Personal Copilot?

• After done with Prerequisites and Configurations as given above, open your code project/workspace.
• Open the code file, where you want help from your Personal Copilot
• Select the code snippet in the file in VS Code, right click then you will be seeing the options starting with Personal-CoPilot
• Select the operation you want help with e.g. "Personal-CoPilot: Refactor the code"
• It will open the CoPilot window on left side in VS Code, with the suggested code
• Click on the code block appearing on the Personal CoPilot window, it will replace the code in code file.
• Similarly you can try with other options

Known Issues/Limitations

Developed this project while experimenting with Local LLM, still work for feedback, Azure Repo review is having some trouble. With keeping these context, use it with at your risk :) .

• Only Eabled with Local LLMs, new verisons will be coming with ChatGPT/Azure etc.
• Required intallation and configuration for Local LLMs

Release Notes

1.0.0

Hope you will like it!