Skip to content

PYTHON_SETUP.md

Latest commit

 

History

History
138 lines (99 loc) · 3.6 KB

PYTHON_SETUP.md

File metadata and controls

138 lines (99 loc) · 3.6 KB

Python Environment Setup

Python Version Requirement

Agent Arena requires Python 3.11 (specifically tested with 3.11.9)

Many ML/AI packages (llama-cpp-python, torch, faiss, etc.) don't yet have pre-built wheels for Python 3.14. Using Python 3.11 ensures all dependencies install smoothly.

Installation Steps

1. Install Python 3.11

Download: Python 3.11.9 (64-bit)

During installation:

  • ✅ Check "Add Python 3.11 to PATH"
  • ✅ Check "Install for all users" (optional)

Verify installation:

py -3.11 --version
# Should output: Python 3.11.9

2. Create Virtual Environment

cd "c:\Projects\Agent Arena\python"

# Create venv with Python 3.11
py -3.11 -m venv venv

# Activate venv
venv\Scripts\activate

# Verify Python version in venv
python --version
# Should output: Python 3.11.9

3. Install Dependencies

Option A: Full Installation (Recommended)

Installs all dependencies including LLM backends, vector stores, etc.

pip install --upgrade pip
pip install -r requirements.txt

Note: This may take 5-10 minutes as some packages (torch, transformers) are large.

Option B: Minimal Installation (For IPC Testing Only)

If you just want to test the IPC system quickly:

pip install --upgrade pip
pip install -r requirements-minimal.txt

This installs only FastAPI, uvicorn, and essential packages. You'll need the full installation later for LLM functionality.

4. Verify Installation

# Test imports
python -c "import fastapi; import uvicorn; print('IPC dependencies OK')"

# Full test (only if you did full installation)
python -c "import torch; import transformers; import faiss; print('All dependencies OK')"

Running the IPC Server

cd "c:\Projects\Agent Arena\python"
venv\Scripts\activate
python run_ipc_server.py

You should see:

============================================================
Agent Arena IPC Server
============================================================
Host: 127.0.0.1
Port: 5000
Max Workers: 4
============================================================
INFO:     Started server process
INFO:     Uvicorn running on http://127.0.0.1:5000

Troubleshooting

"Python 3.11 not found"

  • Make sure you installed Python 3.11 from python.org
  • Try py --list to see available Python versions
  • If 3.11 doesn't appear, reinstall and check "Add to PATH"

"No module named 'fastapi'"

  • Make sure venv is activated: venv\Scripts\activate
  • Reinstall: pip install -r requirements.txt

"ERROR: Could not build wheels for llama-cpp-python"

"torch" installation is very slow

  • torch is ~2GB, be patient
  • Alternative: Use CPU-only version: pip install torch --index-url https://download.pytorch.org/whl/cpu

Port 5000 already in use

  • Change port: python run_ipc_server.py --port 5001
  • Update Godot IPCClient.server_url to match

Using Different Python Versions

If you need to keep Python 3.14 as default but use 3.11 for this project:

# Always use py -3.11 to create the venv
py -3.11 -m venv venv

# Once activated, the venv uses 3.11 automatically
venv\Scripts\activate
python --version  # Shows 3.11.9

Next Steps

After setup:

  1. Start IPC server: python run_ipc_server.py
  2. Open Godot and run scenes/ipc_test.gd
  3. Verify communication in console logs

See docs/ipc.md for detailed IPC documentation.