Skip to main content

Getting Started

Let's install VectorWave and create your first vectorized function. It only takes 5 minutes.

1. Prerequisites​

  • Python: 3.10~3.13
  • Docker: Required to run the Weaviate database.
  • (Optional) OpenAI API Key: Required when using AI auto-documentation and high-performance embedding.

2. Installation​

Library Installation​

Install the VectorWave package using pip in your terminal.

pip install vectorwave

Running Weaviate Database​

VectorWave uses Weaviate as its backend. You can use the docker-compose file included in the project or run it using the command below.

# Download and run vw_docker.yml
curl -O [https://raw.githubusercontent.com/cozymori/vectorwave/main/test_ex/vw_docker.yml](https://raw.githubusercontent.com/cozymori/vectorwave/main/test_ex/vw_docker.yml)
docker-compose -f vw_docker.yml up -d

Environment Variable Setup (.env)​

Create a .env file in the project root and add the following settings. We recommend huggingface mode which doesn't require a separate API key for local testing.

# .env file content
WEAVIATE_HOST=localhost
WEAVIATE_PORT=8080
WEAVIATE_GRPC_PORT=50051

# Vectorizer strategy selection: huggingface (Local CPU, Free)
VECTORIZER="huggingface"
HF_MODEL_NAME="sentence-transformers/all-MiniLM-L6-v2"

# Optional: When using OpenAI (VECTORIZER="openai_client")
# OPENAI_API_KEY="sk-..."

3. Quickstart​

Now let's write your first VectorWave script.

DB Initialization and Function Definition​

Create a quickstart.py file and write the code below.

import time
from vectorwave import vectorize, initialize_database

# 1. Database Connection and Schema Initialization
# Only needs to be called once when the script starts.
initialize_database()

# 2. Apply @vectorize to function
@vectorize(
    search_description="Process user payment and return receipt.", # Description for search
    team="billing",  # Custom tag
    priority=1       # Custom tag
)
def process_payment(user_id: str, amount: int):
    print(f"Processing payment for {user_id}...")
    time.sleep(0.1) # Simulate processing time
    return {"status": "success", "receipt_id": f"rcpt_{amount}"}

# 3. Execute function (Automatically stored)
if __name__ == "__main__":
    print("Function Calling...")
    result = process_payment("user_123", 5000)
    print(f"Result: {result}")

Check Execution Result​

Run the script you wrote.

python quickstart.py

Running it will automatically perform the following operations:

  1. The source code and description of the process_payment function are stored in the VectorWaveFunctions collection.
  2. The function execution record (inputs user_123, 5000 and return value) is stored in the VectorWaveExecutions collection.

Now this data is searchable and can be utilized for RAG or testing later.