portkey-ai

Control Panel for AI Apps

pip install portkey-ai

Features

The Portkey SDK is built on top of the OpenAI SDK, allowing you to seamlessly integrate Portkey's advanced features while retaining full compatibility with OpenAI methods. With Portkey, you can enhance your interactions with OpenAI or any other OpenAI-like provider by leveraging robust monitoring, reliability, prompt management, and more features - without modifying much of your existing code.

AI Gateway

Unified API Signature If you've used OpenAI, you already know how to use Portkey with any other provider.	Interoperability Write once, run with any provider. Switch between any model from_any provider seamlessly.
Automated Fallbacks & Retries Ensure your application remains functional even if a primary service fails.	Load Balancing Efficiently distribute incoming requests among multiple models.
Semantic Caching Reduce costs and latency by intelligently caching results.	Virtual Keys Secure your LLM API keys by storing them in Portkey vault and using disposable virtual keys.
Request Timeouts Manage unpredictable LLM latencies effectively by setting custom request timeouts on requests.

Observability

Logging Keep track of all requests for monitoring and debugging.	Requests Tracing Understand the journey of each request for optimization.
Custom Metadata Segment and categorize requests for better insights.	Feedbacks Collect and analyse weighted feedback on requests from users.
Analytics Track your app & LLM's performance with 40+ production-critical metrics in a single place.

Usage

Prerequisites

• Sign up on Portkey and grab your Portkey API Key
• Add your OpenAI key to Portkey's Virtual Keys page and keep it handy

# Installing the SDK

$ pip install portkey-ai
$ export PORTKEY_API_KEY=PORTKEY_API_KEY

Making a Request to OpenAI

• Portkey fully adheres to the OpenAI SDK signature. You can instantly switch to Portkey and start using our production features right out of the box.
  
• Just replace from openai import OpenAI with from portkey_ai import Portkey:

from portkey_ai import Portkey

portkey = Portkey(
    api_key="PORTKEY_API_KEY",
    virtual_key="VIRTUAL_KEY"
)

chat_completion = portkey.chat.completions.create(
    messages = [{ "role": 'user', "content": 'Say this is a test' }],
    model = 'gpt-4'
)

print(chat_completion)

Async Usage

• Use AsyncPortkey instead of Portkey with await:

import asyncio
from portkey_ai import AsyncPortkey

portkey = AsyncPortkey(
    api_key="PORTKEY_API_KEY",
    virtual_key="VIRTUAL_KEY"
)

async def main():
    chat_completion = await portkey.chat.completions.create(
        messages=[{'role': 'user', 'content': 'Say this is a test'}],
        model='gpt-4'
    )

    print(chat_completion)

asyncio.run(main())

Strands Agents Integration (optional)

Installation:

pip install 'portkey-ai[strands]'

Usage with Strands:

from strands.agent import Agent
from portkey_ai.integrations.strands import PortkeyStrands

model = PortkeyStrands(
    api_key="PORTKEY_API_KEY",
    model_id="@openai/gpt-4o-mini",
#   base_url="https://api.portkey.ai/v1",  ## Optional    
)

agent = Agent(model=model)

import asyncio

async def main():
    result = await agent.invoke_async("Tell me a short programming joke.")
    print(getattr(result, "text", result))

asyncio.run(main())

Google ADK Integration (optional)

Installation:

pip install 'portkey-ai[adk]'

Usage with ADK:

import asyncio
from google.adk.models.llm_request import LlmRequest
from google.genai import types
from portkey_ai.integrations.adk import PortkeyAdk

llm = PortkeyAdk(
    api_key="PORTKEY_API_KEY",
    model="@openai/gpt-4o-mini",
#   base_url="https://api.portkey.ai/v1",  ## Optional    
)

req = LlmRequest(
    model="@openai/gpt-4o-mini",
    contents=[
        types.Content(
            role="user",
            parts=[types.Part.from_text(text="Tell me a short programming joke.")],
        )
    ],
)

async def main():
    # Print only partial chunks to avoid duplicate final output
    async for resp in llm.generate_content_async(req, stream=True):
        if getattr(resp, "partial", False) and resp.content and resp.content.parts:
            for p in resp.content.parts:
                if getattr(p, "text", None):
                    print(p.text, end="")
    print()

asyncio.run(main())

Non-streaming example (single final response):

import asyncio
from google.adk.models.llm_request import LlmRequest
from google.genai import types
from portkey_ai.integrations.adk import PortkeyAdk

llm = PortkeyAdk(
    api_key="PORTKEY_API_KEY",
    model="@openai/gpt-4o-mini",
)

req = LlmRequest(
    model="@openai/gpt-4o-mini",
    contents=[
        types.Content(
            role="user",
            parts=[types.Part.from_text(text="Give me a one-line programming joke (final only).")],
        )
    ],
)

async def main():
    final_text = []
    async for resp in llm.generate_content_async(req, stream=False):
        if resp.content and resp.content.parts:
            for p in resp.content.parts:
                if getattr(p, "text", None):
                    final_text.append(p.text)
    print("".join(final_text))

asyncio.run(main())

Configuration notes:

• system_role: By default, the adapter sends the system instruction as a developer role message to align with ADK. If your provider expects a strict system role, pass system_role="system" when constructing PortkeyAdk.

  llm = PortkeyAdk(
      model="@openai/gpt-4o-mini",
      api_key="PORTKEY_API_KEY",
      system_role="system",  # switch from default "developer"
  )
  

• Tools: When tools are present in the ADK request, the adapter sets tool_choice="auto" to enable function calling by default (mirrors the Strands adapter behavior).

Compatibility with OpenAI SDK

Portkey currently supports all the OpenAI methods, including the legacy ones.

Methods	OpenAI V1.26.0	Portkey V1.3.1
Audio	✅	✅
Chat	✅	✅
Embeddings	✅	✅
Images	✅	✅
Fine-tuning	✅	✅
Batch	✅	✅
Files	✅	✅
Models	✅	✅
Moderations	✅	✅
Assistants	✅	✅
Threads	✅	✅
Thread - Messages	✅	✅
Thread - Runs	✅	✅
Thread - Run - Steps	✅	✅
Vector Store	✅	✅
Vector Store - Files	✅	✅
Vector Store - Files Batches	✅	✅
Generations	❌ (Deprecated)	✅
Completions	❌ (Deprecated)	✅

Portkey-Specific Methods

Methods	Portkey V1.3.1
Feedback	✅
Prompts	✅

Check out Portkey docs for the full list of supported providers

Contributing

Get started by checking out Github issues. Email us at [e-mail] or just ping on Discord to chat.