revgate-python 0.1.0b2

Proprietary SDK for Revgate services

Revgate SDK

The Revgate SDK provides a simple interface for tracing, billing, and observing your AI applications. It supports both manual event tracking and automatic instrumentation for popular AI libraries.

1. SDK Structure

The SDK is organized into the following modules:

• revgate: Top-level package exposing the main Revgate client and key functions.
• revgate.client: Contains the Revgate class for managing the connection and accessing sub-clients (e.g., billing).
• revgate.tracing: The core tracing module.
  • revgate.tracing.core: Handles OpenTelemetry (OTEL) integration and span management.
  • revgate.tracing.events: Defines custom event tracking (e.g., billing events).
  • revgate.tracing.autoinstrumentation: Provides automatic instrumentation for AI libraries.
• revgate.logger: Centralized logging configuration.

2. Usage

Installation

Basic Installation (Core) Lightweight install with just the core dependencies.

pip install revgate-python

With Specific Integrations Install support for specific AI libraries to enable auto-instrumentation.

# For specific providers
pip install "revgate-python[openai]"
pip install "revgate-python[anthropic,langchain]"
pip install "revgate-python[google-genai]"

# For all supported integrations
pip install "revgate-python[all]"

Development Setup For contributors or local development, install in editable mode with all features.

# From the root of the repository
pip install -e ".[all]"

Auto-Instrumentation

Revgate can automatically trace calls to supported AI libraries like Google GenAI, OpenAI, Anthropic, and LangChain. This captures spans, prompts, completions, and token usage without manual instrumentation.

Example with Google GenAI:

import os
from revgate import auto_instrument, Revgate

# 1. Initialize Auto-Instrumentation
# This must be called before using the AI library.
# Supported libraries: "google-genai", "openai", "anthropic", "langchain"
auto_instrument(["google-genai"])

# 2. Use the AI library as usual
from google import genai

client = genai.Client(api_key=os.environ["GEMINI_API_KEY"])
response = client.chats.create(model='gemini-1.5-flash').send_message("Hello!")

print(response.text)
# A trace is automatically generated and sent to the configured OTEL collector.

Manual Event Tracking (Billing)

For custom business logic or billing events that are not captured by auto-instrumentation, you can use track_event.

Example:

from revgate import Revgate

client = Revgate()

# Track a billable event
client.billing.track_event(
    name="llm_call",
    metadata={
        "model": "gemini-1.5-flash",
        "tokens": 150,
        "cost": 0.00015,
        "currency": "USD",
        "user_id": "user_123"
    }
)

# Track a signal (non-billable)
client.billing.track_event(
    name="tool_execution",
    metadata={
        "tool": "search_db",
        "status": "success"
    }
)

Configuration

The SDK uses environment variables for configuration:

• REVGATE_ENABLED: Enable/disable tracing and event tracking (default: true).
• REVGATE_BASE_URL: Revgate base URL (default: http://dev.revgate.com).
• REVGATE_API_KEY: Your Revgate API key.
• REVGATE_OTEL_EXPORTER_TYPE: OTLP exporter type - proto or json (default: proto).
• REVGATE_OTEL_EXPORT_MODE: Span export mode - batch or simple (default: batch).
• REVGATE_ENABLE_CONSOLE_LOGGING: Enable console span exporter for debugging (default: false).
• REVGATE_LOG_LEVEL: Logging level (default: INFO).

3. Development & Release

Local Development

To set up the environment for development:

# Install in editable mode with all dependencies
pip install -e ".[all]"

Building the Package

To build the distribution files (wheel and source archive) locally:

1. Install build tools:

   pip install build
   

2. Run the build:

   python -m build
   

   This creates a dist/ directory containing the .whl and .tar.gz files.

3. Install the built wheel:

   pip install dist/*.whl
   # Note: Validates that the package installs correctly.
   

Publishing (Manual)

To verify the upload or publish manually:

1. Install Twine:

   pip install twine
   

2. Upload:

   twine upload dist/*