Supports both models and Assistants. Choose one to continue.

Note: Only Assistants configured through the SalesIQ–OpenAI integration can be used.

Input 

Define what input (questions) is sent to the card from the visitor/user. Based on this, OpenAI will generate answers. 

Available options

Training

Define the data source used to generate responses.

Knowledge sources (select one):

1. SalesIQ resources:
   
1. The card uses SalesIQ's Articles, FAQs, Webpages, and Small Talks
   
2. Select specific resources or combinations based on your use case.
   
3. Training Status: Click on "Resource training status" to view training status for each resource. If a resource is not trained, you can retry training
   

1. Zoho Desk resources
   
1. The card uses the Desk's knowledge base.
   
2. Zia should be enabled in the Zoho Desk (Setup > General > Zia > and enable Ticket Reply Assistant).
   
3. A minimum of 30 articles is required in the Desk Knowledge base for this card to work. 
4. Zoho Desk integration is mandatory to select the Zoho Desk resource/knowledge base for the bot. 
5. The card uses the articles from the desk department that is mapped to the bot's department under integration configuration.  
6. So, ensure to map the essential desk department with SalesIQ in the SalesIQ-Desk Integration page. 
   

Global data 

Enable global data to handle broader, non-business-specific queries.

1. Define keywords (tags) to guide the AI toward better contextual search.
2. Example - "Trending shoes"
   
3. Use case - You own a shoe shop, when a visitor asks a general query like “What are the trending shoes for GenZ?”, the AI uses these tags to understand the context and generate a relevant response, even if the query is not strictly business-specific.
   

Supports up to 20 tags, each with 20 characters.

This allows the bot to respond to general queries while still prioritizing business context.

Response

Define how responses are handled and presented to the visitor/user.

Waiting state

1. While OpenAI generates a response, you can display interim messages. Example: “Searching…”, “Fetching best match…”, "Putting it together".
   
2. Supports up to 3 messages, displayed sequentially in a loop. 
   

Answer found

Triggered when OpenAI generates a response (answer) using a knowledge-backed context. 

Configurations

1. Allow answer streaming - Streams responses in real time as they are generated, reducing perceived wait time.
2. List reference sources - Displays the resources used to generate the answer in list.
1. Title for source list - Title to display before the reference. Example: “Know more:”
3. Follow-up action - Add buttons as per your preference for this case. Initially, the answers will be displayed, followed by these buttons. Each button will have a separate leg to handle the flow. 
   

Answer not found

Triggered when the OpenAI couldn't generate a response (answer). This can be due to the following reason.

1. No relevant knowledge is identified
   
2. OpenAI cannot generate a usable response
   
3. Detection conditions are met (for Assistants)
   

1. Add a text/message to be displayed when the "Answer not found case is triggered.
   
2. Follow-up action - Add buttons as per your preference for this case. Initially, the answers will be displayed, followed by these buttons. Each button will have a separate leg to handle the flow. 
   

Special Case: OpenAI Assistant

Example instruction to Assistant - When you don’t know or are unsure about an answer, respond with: “Cannot generate answer” (and nothing else)

Ratings

Enable feedback collection for each response. This will collect the rating for a response. 

1. Displays thumbs up / thumbs down options
   
2. Allows users to rate each answer
   

Note: Ratings are supported only for the Web and Mobile SDKs. Social channels don't support this.

1. Occurrence Count - Number of ratings required to trigger an action
2. Mode (one at a time): 
1. Consecutively - Back-to-back ratings
2. Entire chat - Across the full conversation

Unrated messages are not considered positive or negative.

Plug Configuration

1. Select a plug
2. Map required parameters
3. Trigger actions such as notifications, escalations, or workflows
   

Finally, click Save to apply the configuration.

To use Anthropic models, SalesIQ-Anthropic integration is mandatory

Input 

1. Use visitor’s last message - Uses the most recent user input, including responses from buttons, text inputs, or selection cards. Ensure the previous step captures meaningful input to avoid irrelevant answers.
2. Choose context variable - Use data stored from earlier steps in the flow.
3. Ask the visitor for input - Prompt the user with a question and use their response as input.
   

1. SalesIQ resources:
   
1. The card uses SalesIQ's Articles, FAQs, Webpages, and Small Talks
   
2. Select specific resources or combinations based on your use case.
   
3. Training Status: Click on "Resource training status" to view training status for each resource. If a resource is not trained, you can retry training.
   

1. Zoho Desk resources
   
1. The card uses the Desk's knowledge base.
   
2. Zia should be enabled in the Zoho Desk (Setup > General > Zia > and enable Ticket Reply Assistant).
   
3. A minimum of 30 articles is required in the Desk Knowledge base for this card to work. 
4. Zoho Desk integration is mandatory to select the Zoho Desk resource/knowledge base for the bot. 
5. The card uses the articles from the desk department that is mapped to the bot's department under integration configuration.  
6. So, ensure to map the essential desk department with SalesIQ in the SalesIQ-Desk Integration page. 
   

1. Allow answer streaming - Streams responses in real time as they are generated, reducing perceived wait time.
2. List reference sources - Displays the resources used to generate the answer in list.
1. Title for source list - Title to display before the reference. Example: “Know more:”
3. Follow-up action - Add buttons as per your preference for this case. Initially, the answers will be displayed, followed by these buttons. Each button will have a separate leg to handle the flow. 
   

Answer not found

1. Occurrence Count - Number of ratings required to trigger an action
2. Mode (one at a time): 
1. Consecutively - Back-to-back ratings
2. Entire chat - Across the full conversation

Unrated messages are not considered positive or negative.

1. Select a plug
2. Map required parameters
3. Trigger actions such as notifications, escalations, or workflows
   

To use Google AI models, SalesIQ-Google AI integration is mandatory

Input 

1. Use visitor’s last message - Uses the most recent user input, including responses from buttons, text inputs, or selection cards. Ensure the previous step captures meaningful input to avoid irrelevant answers.
2. Choose context variable - Use data stored from earlier steps in the flow.
3. Ask the visitor for input - Prompt the user with a question and use their response as input.
   

1. SalesIQ resources:
   
1. The card uses SalesIQ's Articles, FAQs, Webpages, and Small Talks
   
2. Select specific resources or combinations based on your use case.
   
3. Training Status: Click on "Resource training status" to view training status for each resource. If a resource is not trained, you can retry training
   

1. Zoho Desk resources
   
1. The bot uses the Desk's knowledge base.
   
2. Zia should be enabled in the Zoho Desk (Setup > General > Zia > and enable Ticket Reply Assistant).
   
3. A minimum of 30 articles is required in the Desk Knowledge base for this card to work. 
4. Zoho Desk integration is mandatory to select the Zoho Desk resource/knowledge base for the bot. 
5. The card uses the articles from the desk department that is mapped to the bot's department under integration configuration.  
6. So, ensure to map the essential desk department with SalesIQ in the SalesIQ-Desk Integration page. 
   

1. Allow answer streaming - Streams responses in real time as they are generated, reducing perceived wait time.
2. List reference sources - Displays the resources used to generate the answer in list.
1. Title for source list - Title to display before the reference. Example: “Know more:”
3. Follow-up action - Add buttons as per your preference for this case. Initially, the answers will be displayed, followed by these buttons. Each button will have a separate leg to handle the flow. 
   

Answer not found

1. Occurrence Count - Number of ratings required to trigger an action
2. Mode (one at a time): 
1. Consecutively - Back-to-back ratings
2. Entire chat - Across the full conversation

Unrated messages are not considered positive or negative.

1. Select a plug
2. Map required parameters
3. Trigger actions such as notifications, escalations, or workflows
   

To use DeekSeek models, SalesIQ-DeekSeek integration is mandatory

Input 

1. Use visitor’s last message - Uses the most recent user input, including responses from buttons, text inputs, or selection cards. Ensure the previous step captures meaningful input to avoid irrelevant answers.
2. Choose context variable - Use data stored from earlier steps in the flow.
3. Ask the visitor for input - Prompt the user with a question and use their response as input.
   

1. SalesIQ resources:
   
1. The card uses SalesIQ's Articles, FAQs, Webpages, and Small Talks
   
2. Select specific sources or combinations based on your use case.
   
3. Training status: Click on "Resource training status" to view training status for each resource. If a resource is not trained, you can retry training
   

1. Zoho Desk resources
   
1. The card uses the Desk's knowledge base.
   
2. Zia should be enabled in the Zoho Desk (Setup > General > Zia > and enable Ticket Reply Assistant).
   
3. A minimum of 30 articles is required in the Desk Knowledge base for this card to work. 
4. Zoho Desk integration is mandatory to select the Zoho Desk resource/knowledge base for the bot. 
5. The card uses the articles from the desk department that is mapped to the bot's department under integration configuration.  
6. So, ensure to map the essential desk department with SalesIQ in the SalesIQ-Desk Integration page. 
   

1. Allow answer streaming - Streams responses in real time as they are generated, reducing perceived wait time.
2. List reference sources - Displays the resources used to generate the answer in list.
1. Title for source list - Title to display before the reference. Example: “Know more:”
3. Follow-up action - Add buttons as per your preference for this case. Initially, the answers will be displayed, followed by these buttons. Each button will have a separate leg to handle the flow. 
   

Answer Not Found

1. Occurrence Count - Number of ratings required to trigger an action
2. Mode (one at a time): 
1. Consecutively - Back-to-back ratings
2. Entire chat - Across the full conversation

Unrated messages are not considered positive or negative.

1. Select a plug
2. Map required parameters
3. Trigger actions such as notifications, escalations, or workflows
   

1. Use visitor’s last message - Uses the most recent user input, including responses from buttons, text inputs, or selection cards. Ensure the previous step captures meaningful input to avoid irrelevant answers.
2. Choose context variable - Use data stored from earlier steps in the flow.
3. Ask the visitor for input - Prompt the user with a question and use their response as input.

1. SalesIQ resources:
   
1. The AI gets SalesIQ's Articles, FAQs, Webpages, and Small Talks
   
2. Select specific sources or combinations based on your use case.
   
3. Training Status: Click on the "Resource training status" to view training status for each resource. If a resource is not trained, you can retry training
   

1. Zoho Desk resources
   
1. The AI gets the Desk's knowledge base.
   
2. Zia should be enabled in the Zoho Desk (Setup > General > Zia > and enable Ticket Reply Assistant).
   
3. A minimum of 30 articles is required in the Desk Knowledge base for this card to work. 
4. Zoho Desk integration is mandatory to select the Zoho Desk resource/knowledge base for the bot. 
5. The AI uses the articles from the desk department that is mapped to the bot's department under integration configuration.  
   
6. So, ensure to map the essential desk department with SalesIQ in the SalesIQ-Desk Integration page. 
   

1. Connections - Connections handle authentication securely. Instead of adding credentials directly in the card, link your request to an existing connection created through the Data Resource Environment (DRE). If needed, click Add to create a new connection. All authentication-based connections are managed under DRE, ensuring a secure and centralized setup. 
   

Asynchronous request requirements

1. Include the conversation ID so the system can identify which chat the response belongs to. Use %active_conversation_id% variable to pass the current conversation ID. 
2. Include a verification key (key–value pair) for response authorization.
   
1. Example: "verification_key": "s9s89sdhfsljkgmow0823badokejorp"
   

1. Once your API generates a response, send it to: 
   

1. Replace <id> with the same conversation ID received in the request.
2. Include the same verification key in the response.
   

1. Request parameter - Select the parameter used to send the verification key in the request (query, header or body).
   

1. Response parameter - Specify the parameter in the response that carries the same value. In the below API response, the "verificationKey" holds the value sent from the request. 
   

1. {
2.   "status": "success",
3.   "data": {
4.     "startTime": 1778565755849,
5.     "endTime": 1778565757485,
6.     "request": {
7.       "message": "Hi, I need some help with my recently raised issue.",
8.     },
9.     "response": "Good morning! Could you please provide me with your email address and a brief description of the issue you're facing? This will help me find the correct ticket and provide you with the assistance you need.",
10.   },
11.   "timestamp": "2026-05-12T06:02:37.490758258Z",
12.   "verificationKey": "s9s89sdhfsljkgmow0823badokejorp"
13. }

1. Use dot notation to access nested values (e.g., data.<key-name>).
   

The system compares the conversation ID (request vs response) and the authorization value (request vs response). 

1. If both match, the response is accepted and displayed.
   
2. Upon a mismatch, the failure leg is triggered.
   

1. Response parameter - Specify the parameter containing the answer. Use dot notation to access nested values (e.g., data.response.text).
2. Follow-up action - Add buttons as per your preference for this case. Initially, the answers will be displayed, followed by these buttons.  
   
3. Each button will have a separate leg to handle the flow. 

1. Enter the response parameter (from the API response body). Use dot notation to access nested values (e.g., data.response.text).
2. Define conditions to detect invalid responses
   
1. Check for empty responses
2. Check for specific failure text using conditions (e.g., contains)
   

1. If the "data.response.text" contains “Cannot generate answer” - this will trigger the Answer not found case. 
   

1. Text - Add a text to handle the Answer not found case. 
   
2. Follow-up action - Add buttons as per your preference for this case. Initially, the answers will be displayed, followed by these buttons.  
3. Each button will have a separate leg to handle the flow. 
   

Ratings

Enable feedback collection for each response. This will collect the rating for a response. 

1. Displays thumbs up / thumbs down options
   
2. Allows users to rate each answer
   

Note: Ratings are supported only for the Web and Mobile SDKs. Social channels don't support this.

Trigger actions based on ratings:

Define automated actions via plugs based on rating patterns.

1. Occurrence Count - Number of ratings required to trigger an action
2. Mode (one at a time): 
1. Consecutively - Back-to-back ratings
2. Entire chat - Across the full conversation

Unrated messages are not considered positive or negative.

Plug Configuration

1. Select a plug
2. Map required parameters
3. Trigger actions such as notifications, escalations, or workflows
   

Finally, click Save to apply the configuration.