Anthropic Claude-1 and Claude-Instant 1 are advanced language models designed to assist developers in generating high-quality text. Developed by AI researchers, these models provide a strong foundation for natural language understanding and generation, enabling innovative applications and advancements in AI language technology.

The XO Platform lets you easily connect with the Anthropic Claude-1 and Claude-Instant 1 models to leverage their capabilities and safety across various LLM features supported on the XO Platform. Learn more about Anthropic models.

This article explains how to enable the models for the following LLM features currently supported on the XO Platform:

• Automatic Dialog Generation
• Conversation Test Cases Suggestions
• GenAI Prompt
• NLP Batch Test Cases Suggestions
• Rephrase Dialog Responses
• Training Utterance Suggestions
• Repeat Responses

To make the integration process smoother for developers and customers, the Kore.ai XO Platform supports authorization via API Key for Anthropic models. Learn more.

Anthropic Integration Setup

Prerequisites:

Before setting up the Anthropic Model, please complete the following prerequisites:

1. If you already have an Anthropic developer account, log in by providing your credentials. Otherwise, sign up by following the steps below:

• Send a developer account sign-up request by filling out a form here.
• Once you receive an email confirmation from Anthropic, click the activation link to access your account. Please note down your credentials once your account is created.

2. To generate an API Key, visit this link. Copy the API Key of your Anthropic account and keep it for future use to enable the integration. To generate the API key, follow the steps below:

• Log in to the web console using your email and password.
• Navigate to the Account Settings section and click Generate API Keys.
• Click Create Key.
• Click the Copy icon. The API key will be your access credential to utilize the Claude API in your applications.

3. Configure the Anthropic connector under Integration Setup.

Once you have created an Anthropic AI service and copied the required credentials, you can enable the integration.

Steps to Configure the Anthropic Connector

1. Go to Build > Natural Language > Advanced NLU Settings.
2. Click Configure Now for the Anthropic connector under Integration Setup.
   
3. In the resulting window, click the Authorization tab.
4. Provide the API Key (The secret API key of your Anthropic AI service, which you copied in Step 2 under Prerequisites) required to connect to the Virtual Assistant and access the external services securely.

5. Provide your consent to enable the Integration after reading the policy guidelines.
6. Click Save.

The integration status changes from Configure Now to X/Y models configured, where X is the number of models configured, and Y is the number of models available.

The following LLM features are available by default for Anthropic Claude-1 and Anthropic Claude-Instant-1. They automatically change to Active state once the model is configured. The features not available remain in the Inactive state.

Supported Features

The following features are currently supported on the XO Platform for Anthropic-Claude 1 LLM Model:

Automatic Dialog Generation

Please refer here for feature details and its usage.

Once the Anthropic connector is set up, you can configure Anthropic-Claude-1 or Anthropic-Clause-Instant-1 for this feature along with the Advanced Settings. Learn more.

Anthropic in Action

To see the Anthropic-Claude-1/Anthropic-Clause-Instant-1 model in action, follow the steps below:

1. Navigate to Build > Conversational Skills > Dialog Tasks.
2. On the Dialog Tasks panel, click Create Dialog to create a new task.
   
3. In the Create Dialog window, provide the Intent Name, Description, and utterance for intent training.
4. Enable Conversation Driven Dialog Builder.
5. Click Proceed.
   
   The system redirects to the Build section of the Dialog Builder.
6. In the Conversation Preview window, click and expand the Advanced Settings section.
7. Select Anthropic-Claude-1 or Anthropic-Claude-Instant-1 from the Model dropdown.
8. (Optional) Add System Context text, and change the Temperature (default is set to 0.5) and Max Tokens (default is set to 2500)  limits, if required.
9. Click Generate.
   
10. Wait a few moments until the Platform generates dialog flows for your specified intents using the selected Anthropic Claude model. When done, you see a mock conversation based on which dialog flows will be generated afterward. You can preview the conversation flow, view the Bot Actions taken, improvise the intent description, and Regenerate the conversation if needed.
11. Click Generate Dialog Task when the mock conversation looks good.
    
12. Once you generate the Dialog Task, the canvas is populated with nodes corresponding to the mock conversation. You can customize the flow per your requirements by editing the generated nodes and adding your own. You can configure the generated Dialog Tasks just like any other dialog task.

Conversation Test Cases Suggestions

Please refer here for feature details and its usage.

Once the Anthropic connector is set up, you can select the model for this feature and tweak the model’s Advanced Settings if required. Learn more.

Advanced Settings

Please refer to this link for more information on the Advanced Settings available for this feature.

Anthropic in Action

To see the Anthropic-Claude-1/Anthropic Claude-Instant-1 model in action, follow the steps below:

1. Navigate to Build > Testing > Conversation Testing.
2. Click on an existing test suite listed on the page or create a new test suite.
3. In the chat window, click the Settings icon for Suggestions: User Input.
4. Change the required settings in the Advanced Settings window where Anthropic-Claude-1/Anthropic Claude-Instant-1 is your default selection.
5. Click Save.
6. Select an intent listed on the chat window.

Anthropic-Claude-1/Anthropic Claude-Instant-1 model generates the best alternative suggestions for the intent, as shown below:

GenAI Prompt

Please refer here for feature details and its usage.

Once the Anthropic connector is set up, you can select the model for this feature and tweak the model’s Advanced Settings if required. Learn more.

Anthropic in Action

To see the Anthropic-Claude-1 model in action, follow the steps below:

1. Navigate to Build > Conversational Skills > Dialog Tasks.
2. On the Dialog Tasks panel, click the Dialog Task you want to configure.
3. On the Dialog Flow Builder window, in the Build section, drag and drop the GenAI Prompt node from the left menu onto the canvas. If a GenAI Prompt node already exists, click it.
   
4. On the Component Properties panel, General Settings section, type the prompt text you want to include in the dialog flow.
   
5. Scroll down to the Advanced Settings section, expand the window and select Anthropic-Claude-1 or Anthropic-Claude-Instant-1 in the Model dropdown.
6. (Optional) Add System Context text, change the Temperature (default is set to 0.5) and Max Tokens (default is set to 2500)  limits, if required.
7. Click the Preview Response link for the selected model to generate the prompt preview.
   
   The selected Anthropic model generates an enhanced and context-specific prompt for the dialog flow.

NLP Batch Test Cases Suggestions

Please refer here for feature details and its usage.

Once the Anthropic connector is set up, you can select the model for this feature and tweak the model’s Advanced Settings if required. Learn more.

Anthropic in Action

To see the Anthropic-Claude-1/Anthropic-Clause-Instant-1 model in action, follow the steps below:

1. Navigate to Build > Testing > Batch Testing.
2. Click Manage for the desired Utterance on the list.
3. Click Generate Test Case on the Details page.
4. Type or select the intent from the List.
5. Under Settings, select the scenarios.
6. Under Advanced Settings, select Anthropic-Claude-1/Anthropic-Clause-Instant-1 for Model.
7. Set the Temperature and Max Tokens values.
8. Click Generate.

The system generates test cases based on the selected Anthropic model. You can delete test cases if required or generate more test cases.

9. Click Add Test Cases to add the suggestions to the test suite.

Rephrase Dialog Responses

Please refer here for feature details and its usage.

Once the Anthropic connector is set up, you can select the model for this feature and tweak the model’s Advanced Settings if required. Learn more.

Anthropic in Action

To see the Anthropic-Claude-1/Anthropic Claude-Instant-1 model in action, follow the steps below:

1. Navigate to Build > Conversational Skills > Dialog Tasks.
2. Click the desired dialog task on the list.

3. Click the desired Entity or Message node.
4. Scroll down to Bot Responses and add your response.
5. Enable Rephrase Responses.

6. Select Anthropic-Claude-1/Anthropic Claude-Instant-1 for Model under Advanced Settings.
7. Configure the values for No. of previous user inputs, Temperature, and Max Tokens in the Advanced Settings section.
8. In the chat window, the Bot Response you’ve added in the previous step is rephrased by the Bot, as shown below.

To learn about other LLM and Generative AI features, click here.

Training Utterance Suggestions

Please refer here for feature details and its usage.

Once the Anthropic connector is set up, you can select the model for this feature and tweak the model’s Advanced Settings if required. Learn more.

Anthropic in Action

To see the Anthropic-Claude-1/Anthropic Claude-Instant-1 model in action, follow the steps below:

1. Navigate to Build > Natural Language > Training.
2. Click the desired Intent.

3. In the Utterances section, add a new utterance if required. Otherwise, click the Settings icon in the Suggestions window.

4. Change the required settings in the Advanced Settings window where Anthropic-Claude-1/Anthropic Claude-Instant-1 is the default selection.
5. Click Save.
6. Click Generate in the Suggestions window.

The Anthropic-Claude-1 model generates the best alternative suggestions for your utterance, as shown below:

You can either add all the suggestions, generate more suggestions, accept individual suggestions, or delete a suggestion as required. All the added suggestions are listed under Utterances.

Repeat Responses

Please refer here for feature details and its usage.

Once the Anthropic connector is set up, you can select the model for this feature and tweak the model’s Advanced Settings if required. Learn more.

Anthropic in Action

To see the Anthropic-Claude-1/Anthropic Claude-Instant-1 model in action, follow the steps below:

1. Follow the steps under Enabling the Repeat Bot Response Event on this page to configure the event.

2. To enable an Anthropic model, select Auto-generate Response in the Event Configuration section of the Repeat Bot Response Event window.
3. Select Anthropic-Claude-1 or Anthropic-Claude-Instant-1 in the Model dropdown.
4. (Optional) Change the Temperature (default is set to 0.9) and Max Tokens (default is set to 2068) limits, if required.
5. Click Save.
   
   The selected Anthropic LLM Model enhances the repeat response from the VA triggered by the event during a live conversation on the selected voice channel.