Nscale is a high-performance compute platform that enables you to fine-tune, evaluate, deploy, and run AI models at scale.

You can now access all models available through Nscale from Portkey, bringing the power of Nscale’s model hub into your existing AI infrastructure.

With this integration, you get:

• Unified access to Nscale models alongside 250+ other LLMs

• Observability into every request and response

• Built-in retries, load balancing, and failover logic

• Centralized prompt management and versioning

• Secure API key handling with Portkey’s virtual key system

Read more about the integration here 

Portkey now integrates with OpenAI Agents SDK, enabling you to create production-ready agents!

With this integration, you get:

• Full observability into every tool call and agent step

• Built-in fallbacks, retries, and load balancing

• Cost tracking and support for 250+ LLMs

• Guardrails to ensure safe, compliant agent behavior

• Prompt versioning for consistent outputs

Build, deploy, and scale agents confidently. Get started with the integration guide 📘  

Portkey now integrates with Lepton AI, giving you fast, secure access to Lepton AI APIs.

With this integration, you get:

✅ Built-in observability, retries, and load balancing

✅ Secure key management with virtual API keys

✅ Prompt management and guardrails

 Explore Lepton AI integration 

We're excited to announce that Portkey now integrates with PydanticAI, bringing production-grade capabilities to your GenAI agents.

With this integration, you get:

✅ Observability into every agent step

✅ Fallbacks, retries, and load balancing

✅ Cost tracking and LLM routing across 200+ models

✅ Built-in guardrails and versioned prompts

 Take your agents to production with Pydantic x Portkey! 

We’ve rolled out a new optimization to the Portkey's Gateway that reduces latency - without changing a thing on your end.

No config changes. No breaking updates. Just faster responses.

More such optimizations are coming soon. 🚀🚀

Build production-grade browser automation with enterprise-level controls. Our latest cookbook shows you how to:

• Route and monitor Computer Use API calls

• Build a complete Playwright-based browser automation solution

• Add observability, logging, and cost controls with Portkey

 Explore Computer Use Tool 

1. Portkey now supports PDF processing for Anthropic models across Anthropic, Bedrock, and VertexAI providers. Token usage is calculated based on both text content and the visual representation of pages.

2. Translation, transcription, and speech APIs are now supported on Groq.

3. You can now make text-to-image requests to generate images via WorkersAI provider.

Need to try out or switch to the latest Llama models? There's an easier way to do it.

Llama Prompt Ops transforms prompts that work well with other LLMs into ones that are optimized specifically for Llama models. This helps you get better performance and more reliable results without having to rewrite everything yourself.

If you work in customer support, we've put together a helpful guide that will show you how to build a system that analyzes support messages for urgency and sentiment, and helps categorize them properly.

Check it out here 

Here are two updates on the platform that will strengthen granular observability and governance for your organization:

• Metadata with each API request: Portkey allows organisation owners to define mandatory metadata fields for API keys and workspaces.

When the same key appears at multiple levels, the precedence order is:

 Workspace metadata > API key metadata > Request metadata  

• Default configs with API Keys: With default configs, each API call made with the key will have the configs, without adding the Config ID to the code, with an option to override the config.

Note: Existing API keys and Workspaces will not be affected by the new metadata schema and configs

For more information and implementation steps, read our detailed docs.

 Enforcing metadata 

 Enforce default config 

Org admins can now define a default cache TTL for all API keys and workspaces within the organization.

This means:

✅ Consistent cache retention policies

✅ Flexibility to override TTL at the request level when needed

To implement this, check out the docs here