In this section, we will go over the various options for integrating GPT with Salesforce. There are three ways to achieve Chat GPT with Salesforce:
- Build a custom integration using the Salesforce APEX to call GPT Model API
- Pre-built integration packages or solutions are available on the Salesforce AppExchange
- Create and Host a GPT model on any public cloud or local machine.
I don’t believe any ISV partner (https://appexchange.salesforce.com/) has already developed a package to integrate with the GPT Model. Also, creating and hosting a GPT model can be a complex task that requires a good understanding of machine learning and artificial intelligence.
GPT-3 is a pre-trained model that OpenAI provides; it is not publicly available to download and host on one’s local server or in the public cloud. So, options #2 and #3 are invalid at this moment.
Integrate GPT-3 Model API with Salesforce
We must use the API made available by the service that hosts the GPT model in order to call the predicted text from a Chat GPT model. Send an HTTP request from APEX to a specific endpoint. The body of the request is where you put any needed authentication information or parameters. High-level steps to Integrate Chat GPT with Salesforce
- Sign up for an API key from OpenAI
- Go to the OpenAI website (https://openai.com/) and click on the “Sign In” button in the top right corner of the page.
- After signing in, click on the “API” button at the top of the page.
- Click on the “Create an API Key” button on the API page.
- Provide a name for your API Key and select the permissions you want to grant to the key.
- Click on the “Create” button to create the key.
- Once you have an API key, you can use it to request the GPT-3 API.
- The endpoint for generating text from a GPT-3 model is:
https://api.openai.com/v1/engines/davinci-codex/completions - You can make a POST request to this endpoint with a JSON payload that includes your API key, the text you want to generate completions for, and any other parameters you want to customize the generated text.
import json
import requests
api_key = "YOUR_API_KEY"
model = "davinci-codex"
prompt = "Write something interesting about Salesforce integration"
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {api_key}"
}
data = {
"prompt": prompt,
"engine": model,
"max_tokens": 50,
"stop": ["Salesforce integration"]
}
response = requests.post(
"https://api.openai.com/v1/engines/davinci-codex/completions",
headers=headers,
json=data
)
print(response.json())
Make a GPT-3 API HTTP request in Apex
Http http = new Http();
HttpRequest request = new HttpRequest();
request.setEndpoint('https://api.openai.com/v1/engines/davinci-codex/completions');
request.setMethod('POST');
request.setHeader('Content-Type', 'application/json');
request.setHeader('Authorization', 'Bearer ' + apiKey);
request.setBody(jsonPayload);
HttpResponse response = http.send(request);
if (response.getStatusCode() == 200) {
// Success, parse the JSON response
Map<String, Object> jsonResponse = (Map<String, Object>) JSON.deserializeUntyped(response.getBody());
// process the jsonResponse
}
else {
// Handle the error
}This merely illustrates how to use the GPT-3 API that OpenAI provides; other providers might require various endpoints, authentication methods, and parameters.
Browse Open API documentation about the endpoints
https://beta.openai.com/docs/models/gpt-3
It’s worth noting that OpenAI only provides the GPT models and not a chatbot service; however, the GPT models can be used as the core of a chatbot. But you must build the chatbot architecture, prompt and context management yourself.
Browse more at https://twirltech.in/architect-blogs/
