LangChain
Interacting with LLMs
Getting Started with LLMs
Learn how to seamlessly integrate and switch between large language models (LLMs) using LangChain. In this tutorial, you will:
- Install and import the necessary modules
- Configure API keys via environment variables
- Initialize and invoke the OpenAI LLM
- Swap to Google’s Generative AI (Gemini Pro) on the fly
Prerequisites
- Python 3.7+
- An OpenAI API key (Get yours here)
- A Google Cloud API key with access to Gemini Pro
Note
Store your API keys securely. Avoid committing them to version control.
1. Imports and Environment Setup
Begin by installing and importing LangChain, the os
module, and (optionally) the Google GenAI client.
pip install langchain langchain_google_genai
import os
from langchain import OpenAI
# from langchain_google_genai import GoogleGenerativeAI
To configure your API keys:
# export OPENAI_API_KEY="your_openai_api_key"
# export GOOGLE_API_KEY="your_google_api_key"
2. Initializing the OpenAI LLM
LangChain’s OpenAI
class only requires the OPENAI_API_KEY
environment variable. No additional parameters are needed.
# Initialize the OpenAI client
llm = OpenAI()
3. Defining a Prompt and Generating Text
With the client initialized, craft your prompt and call the model:
prompt = "What would be a good company name for a startup that makes educational toys for kids?"
response = llm.invoke(prompt)
print(response)
Sample Output
Playful Pals Toys
4. Switching to Google Generative AI (Gemini Pro)
LangChain makes it effortless to swap LLM backends. Simply import and initialize the GoogleGenerativeAI
class:
from langchain_google_genai import GoogleGenerativeAI
# Initialize the Gemini Pro client
llm = GoogleGenerativeAI(model="gemini-pro")
# Reuse the same prompt
response = llm.invoke(prompt)
print(response)
Sample Output Suggestions
- Joyful Creations
- Imagination Unbound
- Wonder & Play
Comparing LLM Clients
Model | Initialization Code | Environment Variable |
---|---|---|
OpenAI | llm = OpenAI() | OPENAI_API_KEY |
Google Gemini Pro | llm = GoogleGenerativeAI(model="gemini-pro") | GOOGLE_API_KEY |
Next Steps
- Explore LangChain Documentation for advanced features
- Integrate embedding models and vector databases
- Automate prompt engineering with chains and agents
Warning
Always monitor your API usage to avoid unexpected charges.
References
Watch Video
Watch video content