Desktop Pet Agent

This PySide6 desktop companion keeps a LangChain-driven duck named Pixel floating on your desktop. The translucent duck glides left and right with a looping GIF, and a click on the pet pops open the chat window where you can trade messages powered by LangChain. The codebase targets Python 3.14+ and leans on its modern typing + dataclass ergonomics.

Getting started

1. Install dependencies

   uv pip install -r <(uv pip compile pyproject.toml)

   or with regular pip:

   pip install -e .

2. Configure your Grok credentials

   Create a .env file (or export env vars) with the deployment you want the agent to call:

   GROK_BASE_URL=https://example.com
   GROK_AUTH_TOKEN=sk-your-token

   The PySide agent and the CLI tester both load these values automatically via python-dotenv.

3. Run the pet

   python main.py

   A frameless duck overlay will appear immediately. Click the duck any time you want to summon or re-focus the chat window.

Testing the Grok API

Populate .env with your deployment details:

GROK_BASE_URL=https://example.com
GROK_AUTH_TOKEN=sk-your-token

Then exercise the endpoint:

python scripts/test_grok_api.py "Tell me a cat fact" --model grok-beta

Add --raw to inspect the full JSON response, or adjust the --endpoint, --temperature, and timeout flags as needed.

Modular layout

• d1.prompts.SYSTEM_PROMPT centralizes the pet tone so both local and remote LLMs stay in sync.
• d1.models.ChatModelSettings + d1.models.create_chat_model() decide between langchain-openai and the offline LocalPetChatModel.
• d1.agents.PetAgent is a dataclass that wires prompts, history, and the runnable chain. PetAgent.reset() now returns Self for fluent usage.
• d1.ui.DuckOverlayWindow renders the animated desktop pet, while d1.ui.DesktopPetWindow and d1.ui.AgentWorker own the chat experience. d1.app.run_app() wires the overlay click signal to the chat window.

Feel free to remix the prompt, drop in another LangChain-compatible model, or customize the PySide widgets to make Pixel your own.