Connecting to OpenAI models in LangChain - Step-by-Step Execution

Concept Flow - Connecting to OpenAI models

Import LangChain OpenAI class

↓

Create OpenAI instance with API key

↓

Call model with prompt

↓

Receive response from OpenAI

↓

Use or display the response

This flow shows how to import, create, call, and get a response from an OpenAI model using LangChain.

Execution Sample

LangChain

import os
from langchain.chat_models import ChatOpenAI
from langchain.schema import HumanMessage

llm = ChatOpenAI(openai_api_key=os.getenv("OPENAI_API_KEY"), model_name="gpt-4", temperature=0.7)
response = llm([HumanMessage(content="Hello!")])
print(response.content)

This code connects to the OpenAI GPT-4 model, sends a greeting prompt, and prints the model's reply.

Execution Table

Step	Action	Input/State	Output/Result
1	Import ChatOpenAI class	None	ChatOpenAI class ready to use
2	Create OpenAI instance	openai_api_key=os.getenv('OPENAI_API_KEY'), model_name='gpt-4', temperature=0.7	llm object created with settings
3	Call model with prompt	[HumanMessage(content="Hello!")]	Request sent to OpenAI API
4	Receive response	OpenAI processes prompt	Response object with content text
5	Print response content	response.content	Printed text from model reply

💡 Execution stops after printing the model's response.

Variable Tracker

Variable	Start	After Step 2	After Step 3	After Step 4	Final
llm	None	ChatOpenAI instance	Same instance	Same instance	Same instance
response	None	None	Response object	Response object	Response object with content

Key Moments - 3 Insights

Why do we need to create an instance of ChatOpenAI before calling the model?

Creating the instance sets up the model name and parameters like temperature, which configures how the model responds. This is shown in step 2 of the execution_table.

What does the 'call' method expect as input?

The 'call' method expects a list of BaseMessage objects, like [HumanMessage(content="Hello!")]. This is shown in step 3.

Where does the actual AI response come from?

The response comes from the OpenAI API after processing the prompt, shown in step 4 where the response object is received.

Visual Quiz - 3 Questions

Look at the execution_table, what is the state of 'llm' after step 2?

A It is None

B It is a ChatOpenAI instance configured with model and temperature

C It contains the response from the model

D It is a string with the prompt

💡 Hint

Check the variable_tracker row for 'llm' after step 2.

At which step does the program send the prompt to the OpenAI API?

A Step 3

B Step 4

C Step 1

D Step 5

💡 Hint

Look at the 'Action' and 'Output/Result' columns in the execution_table.

If you change the temperature parameter to 0.0 when creating 'llm', what changes in the execution?

A The model name changes

B The prompt sent changes

C The model's response becomes more deterministic

D The response is printed differently

💡 Hint

Temperature controls randomness in the model's output, set in step 2.

Concept Snapshot

Connecting to OpenAI models with LangChain:
- Import ChatOpenAI from langchain.chat_models and HumanMessage from langchain.schema
- Create instance: llm = ChatOpenAI(openai_api_key=os.getenv("OPENAI_API_KEY"), model_name="gpt-4", temperature=0.7)
- Call model with messages: llm([HumanMessage(content="Hello!")])
- Receive and use response.content
- Temperature controls response creativity

Full Transcript

To connect to OpenAI models using LangChain, first import the ChatOpenAI class and HumanMessage. Then create an instance with your API key, chosen model name and parameters like temperature. Next, call the model with a list of messages like [HumanMessage(content='Hello!')]. The model sends the prompt to OpenAI's API and returns a response object. Finally, access the response content to see the model's reply. This process sets up the connection, sends input, and receives output step-by-step.