> ## Documentation Index
> Fetch the complete documentation index at: https://portkey-docs-chore-gateway-2-19-0.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.
# Arize AX
> Extend Portkeyβs AI Gateway with Arize AX for unified LLM observability, tracing, and evaluation.
[Arize AX](https://arize.com/products/ax/) is the full-featured observability and evaluation platform from [Arize AI](https://arize.com/?utm_source=portkey-docs\&utm_medium=partner\&utm_campaign=partner-docs\&utm_content=tracing-provider-arize) for production teams, AI-native companies, and enterprises. It is available as managed cloud or enterprise self-hosted deployment.
Portkey is a **production-grade AI Gateway and Observability platform** for AI applications. The [OpenInference Portkey instrumentation](https://github.com/Arize-ai/openinference/tree/main/python/instrumentation/openinference-instrumentation-portkey) lets Arize AX capture Portkey-routed calls, including gateway requests and the underlying LLM completions.
Use Arize AX for production AI observability and evaluation. If you want an open-source path for local development, experimentation, or single-container self-hosting, use the separate [Arize Phoenix integration](/integrations/tracing-providers/phoenix).
## Why Portkey + Arize AX?
Thanks to OpenInference instrumentation, Portkey can emit structured traces automatically. This gives you visibility into each LLM call routed through the gateway, making it easier to debug behavior, inspect token usage, and evaluate production traffic in Arize AX. For production evaluation patterns, see Arize's [agent evaluation guide](https://arize.com/guides/ai-agent-handbook/agent-evaluation/) and [LLM evaluation guide](https://arize.com/resources/llm-evaluation/).
* **1600+ LLM Providers**: Single API for OpenAI, Anthropic, AWS Bedrock, and more
* **Advanced Routing**: Fallbacks, load balancing, conditional routing
* **Cost Optimization**: Semantic caching and request controls
* **Security**: PII detection, content filtering, compliance controls
* **40+ Key Metrics**: Cost, latency, tokens, error rates
* **Detailed Logs & Traces**: Request/response bodies and custom tracing
* **Custom Metadata**: Attach custom metadata to your requests
* **Custom Alerts**: Real-time monitoring and notifications
With this integration, you can route LLM traffic through Portkey and gain deep observability in Arize AX, bringing together gateway orchestration and ML observability.
## Getting Started
### Prerequisites
* Python 3.10+
* An Arize AX account
* A Portkey API key
* An OpenAI API key, or another provider key routed through Portkey
### Step 1: Install dependencies
Install the required packages for Arize AX and Portkey:
```bash theme={null}
pip install arize-otel openinference-instrumentation-portkey portkey-ai openai
```
### Step 2: Configure credentials
Set the Arize AX and Portkey credentials in the same shell that runs your application:
```bash theme={null}
export ARIZE_SPACE_ID="your-space-id"
export ARIZE_API_KEY="your-arize-api-key"
export ARIZE_PROJECT_NAME="portkey-gateway"
export PORTKEY_API_KEY="your-portkey-api-key"
export OPENAI_API_KEY="your-openai-api-key"
```
### Step 3: Register Arize AX and instrument Portkey
First, set up the Arize AX OpenTelemetry configuration:
```python theme={null}
import os
from arize.otel import register
from openinference.instrumentation.portkey import PortkeyInstrumentor
tracer_provider = register(
space_id=os.environ["ARIZE_SPACE_ID"],
api_key=os.environ["ARIZE_API_KEY"],
project_name=os.environ["ARIZE_PROJECT_NAME"],
)
PortkeyInstrumentor().instrument(tracer_provider=tracer_provider)
```
Set up Portkey in the same process after registering the `PortkeyInstrumentor`:
```python theme={null}
import os
from portkey_ai import Portkey
portkey = Portkey(
api_key=os.environ["PORTKEY_API_KEY"],
provider="openai",
Authorization=f"Bearer {os.environ['OPENAI_API_KEY']}",
)
response = portkey.chat.completions.create(
model="gpt-4o-mini",
messages=[{"role": "user", "content": "Explain machine learning"}]
)
print(response.choices[0].message.content)
```
## Complete Integration Example
Here's a complete working example that connects Portkey's AI Gateway with Arize AX for centralized monitoring.
```python [expandable] theme={null}
import os
from arize.otel import register
from openinference.instrumentation.portkey import PortkeyInstrumentor
tracer_provider = register(
space_id=os.environ["ARIZE_SPACE_ID"],
api_key=os.environ["ARIZE_API_KEY"],
project_name=os.getenv("ARIZE_PROJECT_NAME", "portkey-production"),
)
PortkeyInstrumentor().instrument(tracer_provider=tracer_provider)
# Import Portkey after instrumentation is registered.
from portkey_ai import Portkey
portkey = Portkey(
api_key=os.environ["PORTKEY_API_KEY"],
provider="openai",
Authorization=f"Bearer {os.environ['OPENAI_API_KEY']}",
)
response = portkey.chat.completions.create(
model="gpt-4o-mini",
messages=[
{"role": "user", "content": "Explain quantum computing in simple terms."}
],
)
print(response.choices[0].message.content)
```
## Verify in Arize AX
Open your Arize AX space and select the project configured by `ARIZE_PROJECT_NAME`. You should see a trace for the Portkey request, including the LLM span, prompt, response, token usage, model, and provider metadata.
See the Arize-side Portkey tracing setup, verification steps, and troubleshooting guidance.
Learn how to use Portkey's Universal API to orchestrate multiple LLMs in a structured debate while tracking performance and evaluating outputs with Arize.
## Portkey AI Gateway Features
While Arize AX provides observability and evaluation, Portkey delivers a complete AI infrastructure platform. Here's everything you get with Portkey:
### π Core Gateway Capabilities
Access OpenAI, Anthropic, Google, Cohere, Mistral, Llama, and 1600+ models through a single unified API. No more managing different SDKs or endpoints.
Use the same code to call any LLM provider. Switch between models and providers without changing your application code.
Secure vault for API keys with budget limits, rate limiting, and access controls. Never expose raw API keys in your code.
Define routing strategies, model parameters, and reliability settings in reusable configurations. Version control your AI infrastructure.
### π‘οΈ Reliability & Performance
Automatically switch to backup providers when primary fails. Define fallback chains across multiple providers.
Distribute requests across multiple API keys or providers based on custom weights and strategies.
Configurable retry logic with exponential backoff for transient failures and rate limits.
Set custom timeouts to prevent hanging requests and improve application responsiveness.
Route requests to different models based on content, metadata, or custom conditions.
Gradually roll out new models or providers with percentage-based traffic splitting.
### π° Cost Optimization
Intelligent caching that understands semantic similarity. Reduce costs by up to 90% on repeated queries.
Set spending limits per provider, team, or project. Get alerts before hitting limits.
Set rate limits per provider, team, or project.
Real-time cost tracking across all providers with detailed breakdowns by model, user, and feature.
### π Built-in Observability
Track 40+ metrics including latency, tokens, costs, cache hits, error rates, and more in real-time.
Full request/response logging with advanced filtering, search, and export capabilities.
Trace requests across your entire AI pipeline with correlation IDs and custom metadata.
Set up alerts on any metric with webhook, email, or Slack notifications.
### π Security & Compliance
Automatically detect and redact sensitive information like SSN, credit cards, and personal data.
Block harmful, toxic, or inappropriate content in real-time based on custom policies.
Fine-grained RBAC with team management, user permissions, and audit logs.
Enterprise-grade security with SOC2 Type II certification and GDPR compliance.
Complete audit trail of all API usage, configuration changes, and user actions.
Zero data retention options and deployment in your own VPC for maximum privacy.
### π’ Enterprise Features
SAML 2.0 support for Okta, Azure AD, Google Workspace, and custom IdPs.
Multi-workspace support with hierarchical teams and department-level controls.
99.9% uptime SLA with dedicated support and custom deployment options.
Deploy Portkey in your own AWS, Azure, or GCP environment with full control.
## Next Steps
Discover all AI Gateway capabilities beyond observability
Secure your API keys and set budgets
Configure fallbacks, load balancing, and more
Use Portkey's native observability features
**Need help?** Join our [Discord community](https://portkey.sh/discord)
***
Contact Us