SUTRA with Portkey

This guide walks you through integrating SUTRA models (V2 or R0) with Portkey. Portkey acts as an AI gateway, providing enhanced observability, reliability, and seamless integration with various AI models, including SUTRA.

📦 Step 1: Install Dependencies

pip install -qU portkey-ai

🔐 Step 2: Set Up API Keys

Ensure you have the following:

• Portkey API Key: Obtain this from your Portkey Dashboard.
• SUTRA API Key: Acquire this from the SUTRA Platform.

Set them as environment variables:

export PORTKEY_API_KEY="your_portkey_api_key"
export SUTRA_API_KEY="your_sutra_api_key"

⚙️ Step 3: Initialize Portkey Client with SUTRA

from portkey_ai import Portkey

client = Portkey(
    api_key="your_portkey_api_key",
    virtual_key="your_virtual_key",  # Optional: If using Portkey's virtual keys
    base_url="https://api.two.ai/v2"  # SUTRA's API endpoint
)

💬 Step 4: Make a Chat Completion Request

response = client.chat.completions.create(
    messages=[{"role": "user", "content": "Translate 'Hello, how are you?' to Hindi."}],
    model="sutra-v2"  # Use "sutra-r0" for reasoning tasks
)

print(response.choices[0].message.content)

🧠 Advanced: Utilize SUTRA-R0 for Reasoning

response = client.chat.completions.create(
    messages=[{"role": "user", "content": "If all humans are mortal and Socrates is a human, is Socrates mortal?"}],
    model="sutra-r0"
)

print(response.choices[0].message.content)

📎 Tips

• Model Selection:

  • Use sutra-v2 for tasks like translation, summarization, and multilingual Q&A.
  • Use sutra-r0 for logical reasoning and analytical tasks.

• Portkey Features:

  • Observability: Monitor requests and responses seamlessly.
  • Retries & Fallbacks: Enhance reliability with automatic retries.
  • Semantic Caching: Reduce latency and costs with intelligent caching.

🔗 Resources

• Portkey Documentation
• SUTRA API
• SUTRA GitHub Repository

Harness the power of SUTRA models with Portkey to build robust, multilingual, and reasoning-capable AI applications.