Skip to main content

Finalize your Python environment

In this lesson, you’ll prepare your Python environment using the provided installation script.
This sets up a virtual environment, installs the correct version of PyTorch (with or without GPU support), and pulls in all required dependencies.

If you're in the rell/ folder, move up and then into the Python folder:

cd ../python

Alternatively, from the root of the project:

cd python

Make sure you're inside the python/ folder for the next steps.

Run the install script​

The install script handles everything:

  • Creates a vector_demo_env virtual environment (if not already created)
  • Detects your platform and GPU availability
  • Installs PyTorch with the correct backend (CUDA, MPS, or CPU-only)
  • Installs all project dependencies listed in requirements.txt

💡 If venv is not available on your system, you may need to install it.
On Ubuntu/Debian: sudo apt install python3-venv

To run the installer:

python3 install.py

Once it finishes, you can activate the virtual environment:

source vector_demo_env/bin/activate

Update your .env file​

Open the python/.env file and update it with your deployment values:

# A list of Postchain node URLs (must be valid JSON)
NODE_URL_POOL='["https://node0.testnet.chromia.com:7740", "https://node1.testnet.chromia.com:7740", "https://node2.testnet.chromia.com:7740", "https://node3.testnet.chromia.com:7740"]'

# The blockchain RID of your deployed chain
BLOCKCHAIN_RID=<your-blockchain-rid>

# Your private key (for signing transactions)
PRIV_KEY=<your-private-key>

# 💰 Box office threshold for vectorization (movies below this are skipped)
BOX_OFFICE_THRESHOLD=100_000_000

# 🧠 Embedding model selection
EMBEDDING_MODEL=sentence-transformers/all-mpnet-base-v2
  • The PRIV_KEY is the private key you generated earlier using chr keygen
  • The EMBEDDING_MODEL must match the model you selected in Lesson 1
  • The BOX_OFFICE_THRESHOLD controls how many movies are embedded and stored on-chain:
    • A higher value includes only top-grossing movies
      → Fewer vectors = faster embedding (text-to-vector conversion)
    • A lower value includes more titles
      → More vectors = slower embedding, especially on CPU

If you're on CPU only, consider a higher threshold (e.g. 500_000_000) for speed.
On GPU, you can lower it — or set it to 0 to embed everything (which may take ~15–20 minutes).
The default (100_000_000) is a good balance for quick results on GPU.

What’s next?​

With your Python environment finalized and .env fully configured, you’re ready to start working with your data.