watchtower-sdk 0.1.1

Python SDK for Watchtower AI - Data Drift & Quality Monitoring

Watchtower AI SDK

The official Python client for Watchtower AI.

Watchtower AI is a comprehensive observability platform for your data pipelines and machine learning models. It helps you detect data drift, monitor data quality, and evaluate LLM performance in real-time.

---

🚀 Key Features

• Data Drift Detection: Automatically detect distribution shifts in your tabular data (KS tests, PSI, etc.).
• Data Quality Monitoring: Validate schema, check for nulls, duplicates, and ranges.
• LLM Observability: Monitor token usage, toxicity, and response quality for your GenAI applications.
• Async Logging: Low-latency, non-blocking API calls designed for high-throughput production environments.
• Cloud Ready: Optimized for deployment on Render, AWS, and other cloud providers.

---

📦 Installation

pip install watchtower-sdk

---

⚙️ Configuration

The SDK is designed to be Zero-Config in production. It automatically reads from environment variables.

Environment Variables (Recommended)

Set these in your local terminal or your cloud provider's dashboard (e.g., Render Environment Variables):

# Your Project API Key (Required)
export WATCHTOWER_API_KEY="your_project_api_key_here"

# Your Watchtower Backend URL (Required for Cloud)
# Defaults to http://localhost:8000 if not set
export WATCHTOWER_API_URL="https://your-watchtower-app.onrender.com"

---

🛠️ Usage Examples

1. Monitoring Tabular Data (ML Models)

Perfect for validating training data or inference batches.

import pandas as pd
from watchtower.monitor import WatchtowerInputMonitor

# Initialize (reads env vars automatically)
monitor = WatchtowerInputMonitor(project_name="Credit Scoring v1")

# Load your production data
df = pd.read_csv("inference_batch.csv")

# Log data to Watchtower
# This sends the data asynchronously to your backend
response = monitor.log(df)

print(f"Logged {len(df)} rows. Status: {response}")

2. Monitoring LLMs (GenAI)

Track prompts, responses, and token usage.

from watchtower.llm_monitor import WatchtowerLLMMonitor

# Initialize
llm_monitor = WatchtowerLLMMonitor(project_name="Customer Support Bot")

# Log an interaction
llm_monitor.log(
    input_text="How do I reset my password?",
    output_text="You can reset your password by clicking 'Forgot Password' on the login page.",
    metadata={
        "model": "gpt-4",
        "latency_ms": 450,
        "user_id": "user_123"
    }
)

---

☁️ Cloud Deployment (Render/Railway)

If you are hosting your own Watchtower Backend on Render:

1. Deploy Backend: Follow the Deployment Guide.
2. Get URL: Copy your backend URL (e.g., https://my-app.onrender.com).
3. Set Env Var: Set WATCHTOWER_API_URL to that URL in your production environment.
4. Run: Your code works exactly the same as it did locally!

---

🛡️ License

This project is licensed under the MIT License.