autopilot

autopilot is a free, local AI code generation extension for Visual Studio Code.

• Cross-platform compatibility
• Incredibly fast
• Understands context

autopilot is a proof of concept.

Usage

This project utilizes the Ollama self-hosted API for managing and communicating with large language models.

1. Download and install Ollama for your operating system:
   • Windows
   • MacOS
   • Linux: curl -fsSL https://ollama.com/install.sh | sh
2. Test Ollama first with your intended LLM to make sure that all model files are downloaded and prepared, and that interactions are working. This can be done by running the ollama run codegemma command on your system (with codegemma being your public LLM of choice).
3. Install the Visual Studio Code extension (this repository!)
4. Navigate to the source file you wish to generate code for, and position the cursor on a line where code should be generated.
5. Press Ctrl + Shift + P to open the Command Palette, then run the command Autopilot: Generate code to generate code at the cursor's position.

Settings

• Context Size - Defines the number of lines (before and after) the cursor position to include in the context for the LLM prompt to generate code.
  • Type: integer
  • Default value: 50
  • Identifier: autopilot.contextSize
• Model Name - This is the name of the Ollama model that will be interacted with to perform generation requests.
  • Type: string
  • Default value: llama3
  • Identifer: autopilot.modelName
• Prompt Type - This is the format of the prompt that will be generated, which is required when interacting with LLMs such as codegemma with nonstandard formatting.
  • Type: enum (of values generic, codegemma)
  • Default value: generic
  • Identifier: autopilot.promptType
• URL - This is the API endpoint created by Ollama on your computer or on another server. This notably includes the API path /api/generate.
  • Type: string
  • Default value: http://localhost:11434/api/generate
  • Identifier: autopilot.url