Quick Start Guide for Frosty AI

Frosty AI allows you to seamlessly manage, optimize, and scale your AI models across multiple providers. Follow this guide to get started quickly!

To get deployed in minutes, use the Quickstart Wizard found at the bottom of the left-hand menu inside the Frosty AI platform. The wizard will guide you step-by-step through setting up your first workspace, connecting providers, and configuring routers.

1. Create a Workspace

Set up a collaborative environment to organize Routers and Providers. In the left-hand menu, select Workspaces and click Create Workspace.

2. Connect a Provider

Link an LLM provider like OpenAI or Anthropic to access models.
Learn how to get an API key from your provider: OpenAI, Anthropic, Mistral AI.

3. Set Up Your Router

Configure Routers to select the best LLM for your tasks. Frosty AI intelligently routes requests based on cost, performance, or custom-defined rules, allowing you to optimize workflows seamlessly. When you configure your route, you can set models for task-specific needs like cost, performance, and future-proofing by setting a failover provider.

4. Integrate Into Your Project

Copy the generated code into your project. You can find the generated code snippet inside the Frosty AI platform, under your configured Router settings.

Install the Frosty AI SDK

Install the SDK using pip:

pip install frosty-ai

Example Usage

# Import Frosty SDK
from frosty_ai import Frosty

def main():
    router_id = "[YOUR_ROUTER_ID]"
    router_key = "[YOUR_ROUTER_KEY]"

    try:
        # Create an instance of the Frosty class
        frosty_sdk = Frosty(router_id, router_key)

        # Make a text generation request
        chat_result = frosty_sdk.chat([{
            "role": "user",
            "content": "Tell me a 10-word joke about the weather."
        }])

        # Use a custom routing rule (e.g., "cost", "performance")
        chat_result = frosty_sdk.chat([{
            "role": "user",
            "content": "Tell me a 10-word joke about the weather."
        }], "cost")

        # Make an embeddings generation request
        embeddings_result = frosty_sdk.embeddings([
            "Embed this sentence.",
            "As well as this one."
        ])

        print(f"Text generation result: {chat_result}")
        print(f"Embeddings generation result: {embeddings_result}")

    except Exception as e:
        print(f"An error occurred: {e}")

if __name__ == "__main__":
    main()

5. Optimize & Validate

Test, compare, refine, and optimize your Router's performance in the Frosty AI Platform! Use built-in performance metrics, cost analysis, and logging to monitor and fine-tune your routing strategies.

---

You're all set! 🚀 Now that your router is up and running, you can explore more advanced configurations to fine-tune your setup.

Next, enhance your router by adding more providers, configuring models for specific tasks based on cost and performance, and ensuring reliability with a failover provider.

See the Magic of Frosty AI in Action

AI is evolving fast. Don’t get left behind. Whether you're just getting started or scaling AI across your organization, Frosty AI gives you the tools to stay ahead—effortlessly.

🚀 Ready to simplify AI adoption and future-proof your workflows?

Sign up today and see how Frosty AI can turn your AI ambitions into reality.

Explore our documentation for the latest tutorials and everything you need to know about Frosty AI.