Tool Calling with SUTRA
This notebook demonstrates how to use SUTRA with LangChain’s tool-calling capabilities to build intelligent, tool-augmented AI applications. SUTRA-V2, developed by TWO AI, supports multilingual chat and integrates seamlessly with LangChain’s ecosystem for dynamic tool usage.
🔧 Prerequisites
- Obtain your SUTRA API key from https://developer.two.ai.
- Store the API key securely (e.g., in Google Colab secrets or environment variables).
- Install LangChain and OpenAI SDK.
📦 Step 1: Install Dependencies
# SUTRA models are OpenAI API compatible
!pip install -qU langchain langchain-openai openai🔐 Step 2: Initialize SUTRA-V2 with LangChain
from langchain_openai import ChatOpenAI
from langchain_core.tools import tool
from langchain_core.messages import HumanMessage, SystemMessage
import os
# Set SUTRA API key
api_key = os.getenv("SUTRA_API_KEY")
# Initialize SUTRA-V2 with ChatOpenAI
llm = ChatOpenAI(
model="sutra-v2",
api_key=api_key,
base_url="https://api.two.ai/v2"
)
# Define a simple tool
@tool
def get_greeting(language: str) -> str:
"""Returns a greeting in the specified language."""
greetings = {
"english": "Hello!",
"hindi": "नमस्ते!",
"french": "Bonjour!"
}
return greetings.get(language.lower(), "Hello!")
# Bind the tool to the LLM
llm_with_tools = llm.bind_tools([get_greeting])💬 Step 3: Tool Calling Example
# Create a prompt with a tool call
messages = [
SystemMessage(content="You are a helpful AI that uses tools to provide greetings."),
HumanMessage(content="Get a greeting in Hindi.")
]
# Invoke the LLM with the tool
response = llm_with_tools.invoke(messages)
print(response.content) # Displays the tool call result🌟 Why LangChain Tool Calling?
- Dynamic Tool Usage: Enables SUTRA-V2 to call custom or pre-built tools for specific tasks.
- Multilingual Support: Combines SUTRA-V2’s language capabilities with LangChain’s flexible tool ecosystem.
- Extensibility: Integrates with tools like search, RAG, and more for complex workflows.
🛠 Troubleshooting
- Invalid API Key: Verify your key at https://developer.two.ai.
- Model Not Found: Use
sutra-v2. SUTRA-V1 was deprecated on March 22, 2025. - Tool Errors: Ensure tools are correctly defined and bound to the LLM.