This is a GPT tool designed to provide a detailed analysis of selected code snippets in a Git project, similar to how git log and git blame work. It utilizes the existing git project, GitHub and GPT to aggregate and summarize the current state of the selected code, its history, changes in its meaning, related commits, pull requests, and details of contributors who have worked on it. 🧐
Installation
Search for "GPT Investigate" on the VS Code extension marketplace.
Features
🔍 GPT Investigate: Right-click on a selected code snippet in a Git project. Select "🔍 GPT Investigate". The extension will generate a concise detailed analysis of the code and its history in the project.
Requirements
Git is required to be installed and accessible through your PATH. The extension will attempt to run git commands in the workspace directory.
Limitations
It will only work in the existing workspace which must be initialized as a git repo
It hasn't been tested with Enterprise GitHub
There is no way to swap out OpenAI for a local LLM
Extension Settings
This extension contributes the following settings:
gptInvestigate.openAI.apiKey: Sets the API Key for OpenAI. It can also be set using the OPENAI_API_KEY environment variable to protect your key from being included in your workspace settings.
gptInvestigate.git.accessToken: Personal access token for GitHub to fetch PR information related to a commit. This is optional according to GitHub documentation, but you'll be rate limited sooner without a PAT set. It can also be set using the GIT_PAT environment variable to protect your key from being included in your workspace settings.
gptInvestigate.openAI.temperature: Sets the temperature for the model. Default 0.
gptInvestigate.openAI.modelName: Name of the OpenAI model to use. Default gpt-3.5-turbo. gpt-4 will provide the best results but can be costly. If you're running into token limitations, try using gpt-3.5-turbo-16k.
gptInvestigate.openAI.maxTokens: Sets the maxTokens for the API. Default -1, which is a smart token calculation from langchain library.
