How to Integrate OpenAI with IBM Watson

Prerequisites

• Create accounts on both OpenAI and IBM Watson platforms if you haven't already.

 

• Obtain API keys from both OpenAI and IBM Watson for authentication purposes.

 

• Install the necessary SDKs for both OpenAI and IBM Watson using your preferred package manager.

pip install openai ibm-watson

Set Up Your Environment

• Securely store your API keys and access them via your application. You can use environment variables or a configuration file for this purpose.

 

• Ensure that your environment is equipped with Python 3.6+ to support the libraries.

import os
from openai import OpenAIAPI
from ibm_watson import AssistantV2

openai_api_key = os.getenv("OPENAI_API_KEY")
ibm_watson_api_key = os.getenv("IBM_WATSON_API_KEY")

Initialize Clients

• Initialize the OpenAI and IBM Watson clients using the stored API keys. Set up any necessary configurations like service URLs for IBM Watson.

# OpenAI Client Initialization
openai_client = OpenAIAPI(api_key=openai_api_key)

# IBM Watson Client Initialization
ibm_assistant = AssistantV2(
    version='2021-06-14',
    iam_apikey=ibm_watson_api_key,
    url='https://api.us-south.assistant.watson.cloud.ibm.com'
)

Create a Function to Call OpenAI API

• Design a function that can send text prompts to OpenAI and handle the response as needed. Customize parameters for your specific use case.

def query_openai(prompt):
    response = openai_client.Completion.create(
        engine="text-davinci-002",
        prompt=prompt,
        max_tokens=150
    )
    return response.choices[0].text.strip()

Create a Function to Call IBM Watson API

• Create another function to interact with IBM Watson, sending requests and managing the responses appropriately.

def query_ibm_watson(session_id, message):
    response = ibm_assistant.message(
        assistant_id='your-assistant-id',
        session_id=session_id,
        input={
            'message_type': 'text',
            'text': message
        }
    ).get_result()
    return response['output']['generic'][0]['text']

Integrate Both APIs

• Implement a workflow that integrates both APIs. Use OpenAI's response as input for IBM Watson or vice versa, depending on your integration needs.

def integrated_workflow(prompt):
    openai_response = query_openai(prompt)
    session_id = ibm_assistant.create_session(assistant_id='your-assistant-id').get_result()['session_id']
    ibm_response = query_ibm_watson(session_id, openai_response)
    return ibm_response

Test the Integration

• Test the workflow end-to-end to ensure successful integration. Print the final response or handle it according to your application logic.

if __name__ == "__main__":
    prompt = "How does OpenAI and IBM Watson integration work?"
    result = integrated_workflow(prompt)
    print("Final Output from IBM Watson:", result)

Refine and Optimize

• Analyze performance and cost metrics. Adjust the number of tokens, response handling, and other parameters as needed.

 

• Implement error handling for API downtimes or rate limits.

try:
    result = integrated_workflow(prompt)
    print("Final Output from IBM Watson:", result)
except Exception as e:
    print("An error occurred:", e)