revenium-middleware-perplexity 0.1.1

A Python library that meters Perplexity AI usage to Revenium.

[!IMPORTANT]

📦 This package has moved to the unified Revenium SDK

revenium-middleware-perplexity is now part of revenium-python-sdk — a single package that meters OpenAI, Anthropic, Google, Ollama, LiteLLM, Perplexity and fal.ai.

Please migrate:

pip install revenium-python-sdk

This package is no longer actively developed; maintenance and new features continue in the consolidated SDK. 📖 Docs & source: https://github.com/revenium/revenium-python-sdk#readme

---

Revenium Middleware for Perplexity (Python)

A lightweight, production-ready middleware that adds Revenium metering and tracking to Perplexity AI API calls in Python.

Features

• Automatic Metering - Tracks all API calls with detailed usage metrics
• Streaming Support - Full support for streaming responses
• Dual SDK Support - Works with both OpenAI SDK and native Perplexity SDK
• Custom Metadata - Add custom tracking metadata to any request
• Trace Visualization - Built-in support for distributed tracing
• Production Ready - Battle-tested and optimized for production use

Quick Start

Installation

Choose the installation that matches your SDK preference:

# For OpenAI SDK users (using Perplexity via OpenAI client)
pip install revenium-middleware-perplexity[openai]

# For native Perplexity SDK users
pip install revenium-middleware-perplexity[perplexity]

# For both SDKs (recommended for flexibility)
pip install revenium-middleware-perplexity[all]

Basic Usage

Option 1: Using OpenAI SDK

Perfect if you're already using OpenAI SDK or want maximum compatibility:

import os
from dotenv import load_dotenv
from openai import OpenAI

# Load environment variables
load_dotenv()

# Import Revenium middleware (this automatically patches OpenAI)
import revenium_middleware_perplexity

# Create OpenAI client with Perplexity base URL
client = OpenAI(
    api_key=os.getenv("PERPLEXITY_API_KEY"),
    base_url="https://api.perplexity.ai"
)

# Make a chat completion request
response = client.chat.completions.create(
    model="sonar",
    messages=[
        {"role": "user", "content": "What is the capital of France?"}
    ]
)

print(response.choices[0].message.content)
# Usage data automatically sent to Revenium!

Option 2: Using Native Perplexity SDK

Perfect if you prefer the official Perplexity SDK:

import os
from dotenv import load_dotenv
from perplexity import Perplexity

# Load environment variables
load_dotenv()

# Import Revenium middleware (this automatically patches Perplexity)
import revenium_middleware_perplexity

# Create Perplexity client
client = Perplexity(
    api_key=os.getenv("PERPLEXITY_API_KEY")
)

# Make a chat completion request
response = client.chat.completions.create(
    model="sonar",
    messages=[
        {"role": "user", "content": "What is the capital of France?"}
    ]
)

print(response.choices[0].message.content)
# Usage data automatically sent to Revenium!

Note: Both approaches work identically! The middleware automatically detects which SDK you're using and applies the appropriate wrapper.

Environment Variables

Create a .env file in your project root:

# Perplexity Configuration
PERPLEXITY_API_KEY=pplx_your_perplexity_api_key

# Revenium Configuration
REVENIUM_METERING_API_KEY=hak_your_revenium_api_key
REVENIUM_METERING_BASE_URL=https://api.revenium.ai

# Optional: Logging
REVENIUM_LOG_LEVEL=INFO

What Gets Tracked

The middleware automatically captures comprehensive usage data:

Usage Metrics

• Token Counts - Input tokens, output tokens, total tokens
• Model Information - Model name, provider (Perplexity)
• Request Timing - Request duration, response time
• Cost Calculation - Estimated costs based on current pricing

Business Context (Optional)

• User Tracking - Subscriber ID, email, credentials
• Organization Data - Organization ID, subscription ID, product ID
• Task Classification - Task type, agent identifier, trace ID

Technical Details

• API Endpoints - Chat completions
• Request Types - Streaming vs non-streaming
• Error Tracking - Failed requests, error types
• Environment Info - Development vs production usage

Advanced Usage

Custom Metadata

Add business context to track usage by customer, product, or feature:

response = client.chat.completions.create(
    model="sonar-pro",
    messages=[{"role": "user", "content": "Hello!"}],
    usage_metadata={
        "organization_id": "org-123",
        "product_id": "prod-456",
        "subscriber": {
            "id": "user-789",
            "email": "[e-mail]"
        },
        "task_type": "chat",
        "trace_id": "trace-abc-123"
    }
)

Streaming Responses

The middleware automatically handles streaming:

stream = client.chat.completions.create(
    model="sonar-pro",
    messages=[{"role": "user", "content": "Write a poem"}],
    stream=True,
    usage_metadata={"task_type": "creative_writing"}
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

# Usage data automatically sent after stream completes!

Decorator Support

The middleware provides powerful decorators for automatic metadata injection, eliminating the need to pass usage_metadata to every API call.

@revenium_metadata

Automatically injects metadata into all Perplexity API calls within a function:

from revenium_middleware_perplexity import revenium_metadata
from openai import OpenAI

client = OpenAI(
    api_key="pplx_your_api_key",
    base_url="https://api.perplexity.ai"
)

@revenium_metadata(
    organization_id="acme-corp",
    task_type="customer-support",
    trace_id="session-12345"
)
def handle_customer_query(question: str) -> str:
    """
    All Perplexity calls within this function automatically include:
    - organization_id: "acme-corp"
    - task_type: "customer-support"
    - trace_id: "session-12345"
    """
    response = client.chat.completions.create(
        model="sonar",
        messages=[{"role": "user", "content": question}]
    )
    return response.choices[0].message.content

# No need to pass usage_metadata to each call!
answer = handle_customer_query("How do I reset my password?")

Benefits:

• ✅ Cleaner Code - No repetitive usage_metadata parameters
• ✅ Automatic Injection - Metadata applied to all calls in function scope
• ✅ API-Level Override - API-level metadata takes precedence when needed
• ✅ Async Support - Works with both sync and async functions
• ✅ Thread-Safe - Uses contextvars for proper isolation

@revenium_meter (Selective Metering)

Control which functions are metered when selective metering is enabled:

from revenium_middleware_perplexity import revenium_meter, revenium_metadata

# Set in .env file:
# REVENIUM_SELECTIVE_METERING=true

@revenium_meter()
@revenium_metadata(task_type="premium-feature")
def premium_feature(prompt: str):
    # ✅ This WILL be metered (decorated with @revenium_meter)
    response = client.chat.completions.create(
        model="sonar",
        messages=[{"role": "user", "content": prompt}]
    )
    return response.choices[0].message.content

def free_feature(prompt: str):
    # ❌ This will NOT be metered (no @revenium_meter decorator)
    response = client.chat.completions.create(
        model="sonar",
        messages=[{"role": "user", "content": prompt}]
    )
    return response.choices[0].message.content

Accepted values for REVENIUM_SELECTIVE_METERING:

• "true", "1", "yes", "on" (case-insensitive) → Selective metering enabled
• "false", "0", "no", "off", or unset → All calls metered (default)

Distributed Tracing

The middleware supports comprehensive distributed tracing through environment variables, allowing you to track requests across services and understand the context of API calls.

Trace Fields

Set these environment variables to enable distributed tracing:

# Deployment Context
REVENIUM_ENVIRONMENT=production          # Environment (dev, staging, prod)
REVENIUM_REGION=us-east-1               # Cloud region

# Trace Identification
REVENIUM_TRACE_TYPE=customer-workflow   # Categorical trace identifier
REVENIUM_TRACE_NAME=Customer Onboarding # Human-readable trace label

# Distributed Tracing
REVENIUM_PARENT_TRANSACTION_ID=trace-123 # Parent transaction ID
REVENIUM_TRANSACTION_NAME=analyze-doc    # Operation name

# Retry Tracking
REVENIUM_RETRY_NUMBER=0                  # Retry attempt (0 = first attempt)

# Credential Identification
REVENIUM_CREDENTIAL_ALIAS=Perplexity Prod Key  # Human-readable credential name

Distributed Tracing Example

Track a request flowing through multiple services:

import os
import uuid
from openai import OpenAI

client = OpenAI(
    api_key="pplx_your_api_key",
    base_url="https://api.perplexity.ai"
)

# Generate workflow trace ID
workflow_id = str(uuid.uuid4())

# Set trace context
os.environ["REVENIUM_ENVIRONMENT"] = "production"
os.environ["REVENIUM_TRACE_TYPE"] = "document-pipeline"
os.environ["REVENIUM_TRACE_NAME"] = "Document Processing"
os.environ["REVENIUM_PARENT_TRANSACTION_ID"] = workflow_id

# Step 1: Analysis Service
os.environ["REVENIUM_TRANSACTION_NAME"] = "analyze-document"
response1 = client.chat.completions.create(
    model="sonar",
    messages=[{"role": "user", "content": "Analyze document"}],
    extra_body={"usage_metadata": {"service": "analyzer", "step": 1}}
)

# Step 2: Extraction Service
os.environ["REVENIUM_TRANSACTION_NAME"] = "extract-content"
response2 = client.chat.completions.create(
    model="sonar",
    messages=[{"role": "user", "content": "Extract content"}],
    extra_body={"usage_metadata": {"service": "extractor", "step": 2}}
)

# All calls are linked by workflow_id in Revenium dashboard!

Trace Field Reference

Field	Environment Variable	Description	Example
Environment	REVENIUM_ENVIRONMENT	Deployment environment	production, staging, dev
Region	REVENIUM_REGION	Cloud region	us-east-1, eu-west-1
Trace Type	REVENIUM_TRACE_TYPE	Categorical identifier	customer-workflow, batch-job
Trace Name	REVENIUM_TRACE_NAME	Human-readable label	Customer Onboarding, Daily Report
Parent Transaction ID	REVENIUM_PARENT_TRANSACTION_ID	Parent transaction for distributed tracing	uuid-string
Transaction Name	REVENIUM_TRANSACTION_NAME	Operation name	analyze-document, generate-summary
Retry Number	REVENIUM_RETRY_NUMBER	Retry attempt number	0 (first), 1 (first retry)
Credential Alias	REVENIUM_CREDENTIAL_ALIAS	Human-readable credential name	Perplexity Production Key

These fields are automatically included in all metering data and help you:

• 🔍 Track requests across multiple services
• 📊 Analyze patterns by environment, region, or workflow type
• 🔄 Monitor retries and failure rates
• 🎯 Identify bottlenecks in distributed systems

Examples

The package includes comprehensive examples in the examples/ directory:

Basic Examples

• getting_started.py - Simplest usage example
• basic.py - Custom metadata example
• streaming.py - Streaming responses

Advanced Examples

• example_decorator.py - Using @revenium_metadata and @revenium_meter decorators
• example_tracing.py - Distributed tracing with trace fields
• trace_visualization.py - Trace visualization setup

Run any example:

python examples/getting_started.py
python examples/example_decorator.py
python examples/example_tracing.py

See examples/README.md for detailed documentation.

How It Works

1. Import: Import revenium_middleware_perplexity to automatically patch OpenAI
2. Use Normally: Use the OpenAI client with Perplexity's base URL
3. Automatic Tracking: All requests are automatically tracked
4. Async Metering: Usage data is sent to Revenium in the background
5. Transparent: Original responses are returned unchanged

The middleware never blocks your application - if Revenium tracking fails, your Perplexity requests continue normally.

Supported APIs

• Chat Completions API (client.chat.completions.create())
• Streaming API (client.chat.completions.create(stream=True))

Requirements

• Python 3.8+
• Revenium API key
• Perplexity API key

Documentation

For detailed documentation, visit docs.revenium.io

License

This project is licensed under the MIT License - see the LICENSE file for details.

Support

For issues, feature requests, or contributions:

• GitHub Repository: revenium/revenium-middleware-perplexity-python
• Issues: Report bugs or request features
• Documentation: docs.revenium.io

---

Built by Revenium