Automation

[MashrafeeAryan]

Summary

This pull request introduces the initial runtime architecture for Friday.

The current implementation establishes the basic execution loop:

User Goal
    ↓
Take Screenshot
    ↓
Send Goal + Screenshot to Gemini
    ↓
Gemini Generates Python
    ↓
Execute Python
    ↓
Take New Screenshot
    ↓
Repeat Until Complete

Changes

• Added model.py

  • Handles communication with the Gemini API.
  • Maintains the AI conversation.
  • Returns generated Python code.

• Added computer.py

  • Captures screenshots.
  • Executes the Python code generated by the AI.

• Added prompt.py

  • Defines the system prompt that instructs Gemini how to control the computer.
  • Establishes the behavior and rules for the AI.

• Added main.py

  • Implements the main execution loop.
  • Connects the model, computer, and prompt components.

Notes

This is the first working architecture for Friday. The focus of this pull request is simplicity. Friday is responsible for providing the AI with the current computer state and executing the generated Python code, while the AI performs all reasoning and decides how to accomplish the user's goal.

Future pull requests will expand the computer interaction capabilities, improve prompt quality, and make the execution loop more robust.