Summarization with custom sections

Summarization with custom sections uses an LLM to summarize customer service conversations. The model takes in the conversations and custom sections, which are parts of text prompts defining the task the model performs, to then generate the summaries.

You can also write your own summarization task definitions in custom sections. Predefined and custom sections are described in a generator.

Predefined sections

This feature provides six predefined sections, from which you can select:

Situation. This is what the customer needs help with or has questions about.
Action. Defined as what the agent does to help the customer.
Resolution. There are four options.
- Y: Yes. All the customer issues and queries are resolved.
- P: Partial. Only some of the multiple customer issues and queries are resolved.
- N: No. None of the customer issues and queries are resolved.
- N/A: There are no specific issues or queries raised by the customer in the conversation.
Customer satisfaction. The options are dissatisfied or not dissatisfied.
- D: The customer is dissatisfied or has negative feelings by the end of the conversation.
- N: The customer is neutral or has positive feelings by the end of the conversation.
Reason for cancellation. Reason for cancellation if the customer requests a cancellation. N/A otherwise.
Entities. The key-value pairs of important entities extracted from the conversation.

Create a summarization generator

Console

To create a summarization generator using the Agent Assist Console, do the following:

In the console, go to the Summarization page and select Generator.

Go to Summarization
Enter the generator's name.
Select the generator's version.
Select output language.
Select the sections under Predefined sections.
Add sections under Custom sections: Define new summarization tasks as needed, optionally adding few-shot examples for enhancement.
Click Save.

REST

To create a summarization generator using the API, do the following:

Call the CreateGenerator method from the Generator resource.
Enter the generator's name in the description field.
Enter the generator's output language code in the output_language_code field.
Enter the trigger event MANUAL_CALL.
Select predefined sections and add custom sections with few-shot examples in SummarizationContext.

Before using any of the request data, make the following replacements:

PROJECT_ID: your Cloud project ID
LOCATION_ID: your location ID

HTTP method and URL:

POST https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/generators

Request JSON body:

{
  "parent": "projects/PROJECT_ID/locations/LOCATION_ID",
  "description": "test",
  "triggerEvent": "MANUAL_CALL",
  "summarizationContext": {
    "summarizationSections": [{
      "type": "SITUATION"
    }, {
      "type": "ENTITIES"
    }, {
      "key": "custom_section",
      "definition": "Describe what the customer needs help with or has questions about. If there is an order number mentioned in the conversation, INCLUDE the order number.",
      "type": "CUSTOMER_DEFINED"
    }],
    "version": "2.1",
    "outputLanguageCode": "en-US"
  }
}

To send your request, expand one of these options:

curl (Linux, macOS, or Cloud Shell)

Note: The following command assumes that you have logged in to the gcloud CLI with your user account by running gcloud init or gcloud auth login , or by using Cloud Shell, which automatically logs you into the gcloud CLI . You can check the currently active account by running gcloud auth list.

Save the request body in a file named request.json, and execute the following command:

curl -X POST \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     -H "x-goog-user-project: PROJECT_ID" \
     -H "Content-Type: application/json; charset=utf-8" \
     -d @request.json \
     "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/generators"

PowerShell (Windows)

Note: The following command assumes that you have logged in to the gcloud CLI with your user account by running gcloud init or gcloud auth login . You can check the currently active account by running gcloud auth list.

Save the request body in a file named request.json, and execute the following command:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred"; "x-goog-user-project" = "PROJECT_ID" }

Invoke-WebRequest `
    -Method POST `
    -Headers $headers `
    -ContentType: "application/json; charset=utf-8" `
    -InFile request.json `
    -Uri "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/generators" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "name": "projects/PROJECT_ID/locations/LOCATION_ID/generators/GENERATOR_ID",
  "description": "test",
  "inferenceParameter": {
    "maxOutputTokens": 1024,
    "temperature": 0.0,
    "topK": 40,
    "topP": 0.8
  },
  "triggerEvent": "MANUAL_CALL",
  "createTime": "2024-05-11T00:28:14.790937126Z",
  "updateTime": "2024-05-11T01:59:15.751724150Z",
  "summarizationContext": {
    "summarizationSections": [{
      "type": "SITUATION"
    }, {
      "type": "ENTITIES"
    }, {
      "key": "custom_section",
      "definition": "Describe what the customer needs help with or has questions about. If there is an order number mentioned in the conversation, INCLUDE the order number.",
      "type": "CUSTOMER_DEFINED"
    }],
    "version": "2.1",
    "outputLanguageCode": "en-US"
  }
}

Test a summarization generator

Within the console, you can test the generator in the Test generator section, as shown in the following image.

To test a generator in the console, do the following:

Console

A transcript of the conversation. You can either manually input utterances or upload a transcript in JSON format, as shown in the following image.
Click Generate to display the summary.

Configure a conversation profile

A conversation profile configures a set of parameters that control the suggestions made to an agent during a conversation. The following steps create a ConversationProfile with a HumanAgentAssistantConfig object.

The following samples show how to create a conversation profile.

Console

In the Agent Assist Console, go to the Conversation profiles page.
Enter the display name.
Select Conversation summarization (generator) as the suggestion type.
Select the generator's name from the list.

REST

Call the create method from the conversation profile resource.
Provide a name for the new conversation profile.
Enter the language code.
Enter the generator name in human_agent_suggestion_config.

Before using any of the request data, make the following replacements:

PROJECT_ID: your Cloud project ID
LOCATION_ID: your location ID
CONVERSATION_PROFILE_ID: your conversation profile ID
GENERATOR_ID: your generator ID

HTTP method and URL:

POST https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/conversationProfiles

Request JSON body:

{
  "name": "projects/PROJECT_ID/locations/LOCATION_ID/conversationProfiles/CONVERSATION_PROFILE_ID",
  "displayName": "conversation-profile-with-generator",
  "humanAgentAssistantConfig": {
    "humanAgentSuggestionConfig": {
      "generators": "projects/PROJECT_ID/locations/LOCATION_ID/generators/GENERATOR_ID"
    }
  },
  "languageCode": "en-US"
}

To send your request, expand one of these options:

curl (Linux, macOS, or Cloud Shell)

Save the request body in a file named request.json, and execute the following command:

curl -X POST \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     -H "x-goog-user-project: PROJECT_ID" \
     -H "Content-Type: application/json; charset=utf-8" \
     -d @request.json \
     "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/conversationProfiles"

PowerShell (Windows)

Save the request body in a file named request.json, and execute the following command:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred"; "x-goog-user-project" = "PROJECT_ID" }

Invoke-WebRequest `
    -Method POST `
    -Headers $headers `
    -ContentType: "application/json; charset=utf-8" `
    -InFile request.json `
    -Uri "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/conversationProfiles" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "name": "projects/PROJECT_ID/LOCATION_ID/conversationProfiles/CONVERSATION_PROFILE_ID",
  "displayName": "conversation-profile-with-generator",
  "humanAgentAssistantConfig": {
    "humanAgentSuggestionConfig": {
      "generators": "projects/PROJECT_ID/locations/LOCATION_ID/generators/GENERATOR_ID"
    }
  },
  "languageCode": "en-US"
}

Verify with the simulator

You can verify the conversation profile in the Agent Assist simulator.

Conversations at runtime

At runtime, when a dialog begins between a user and an agent, you create a conversation, which is an interaction between an agent (human or virtual) and a support customer or user. To see suggestions, create both a user-participant and an agent-participant, and add them to the conversation.

Create a conversation

To create a conversation, call the create method on the Conversation resource.

REST

Before using any of the request data, make the following replacements:

PROJECT_ID: your Cloud project ID
LOCATION_ID: your location ID
CONVERSATION_PROFILE_ID: the ID you received when creating the conversation profile

HTTP method and URL:

POST https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/conversations

Request JSON body:

{
  "conversationProfile": "projects/PROJECT_ID/locations/LOCATION_ID/conversationProfiles/CONVERSATION_PROFILE_ID",
}

To send your request, expand one of these options:

curl (Linux, macOS, or Cloud Shell)

Save the request body in a file named request.json, and execute the following command:

curl -X POST \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     -H "x-goog-user-project: PROJECT_ID" \
     -H "Content-Type: application/json; charset=utf-8" \
     -d @request.json \
     "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/conversations"

PowerShell (Windows)

Save the request body in a file named request.json, and execute the following command:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred"; "x-goog-user-project" = "PROJECT_ID" }

Invoke-WebRequest `
    -Method POST `
    -Headers $headers `
    -ContentType: "application/json; charset=utf-8" `
    -InFile request.json `
    -Uri "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/conversations" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "name": "projects/PROJECT_ID/locations/LOCATION_ID/conversations/CONVERSATION_ID",
  "lifecycleState": "IN_PROGRESS",
  "conversationProfile": "projects/PROJECT_ID/locations/LOCATION_ID/conversationProfiles/CONVERSATION_PROFILE_ID",
  "startTime": "2018-11-05T21:05:45.622Z"
}

The path segment after conversations contains your new conversation ID.

Create a user-participant

To create a user-participant, call the create method on the Participant resource. Provide your conversation ID and END_USER for the role field.

REST

Before using any of the request data, make the following replacements:

PROJECT_ID: your Cloud project ID
LOCATION_ID: your location ID
CONVERSATION_ID: your conversation ID

HTTP method and URL:

POST https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/conversations/CONVERSATION_ID/participants

Request JSON body:

{
  "role": "END_USER",
}

To send your request, expand one of these options:

curl (Linux, macOS, or Cloud Shell)

Save the request body in a file named request.json, and execute the following command:

curl -X POST \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     -H "x-goog-user-project: PROJECT_ID" \
     -H "Content-Type: application/json; charset=utf-8" \
     -d @request.json \
     "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/conversations/CONVERSATION_ID/participants"

PowerShell (Windows)

Save the request body in a file named request.json, and execute the following command:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred"; "x-goog-user-project" = "PROJECT_ID" }

Invoke-WebRequest `
    -Method POST `
    -Headers $headers `
    -ContentType: "application/json; charset=utf-8" `
    -InFile request.json `
    -Uri "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/conversations/CONVERSATION_ID/participants" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "name": "projects/PROJECT_ID/locations/LOCATION_ID/conversations/CONVERSATION_ID/participants/PARTICIPANT_ID",
  "role": "END_USER"
}

The path segment after participants contains your new user-participant ID.

Create an agent-participant

Call the create method on the Participant resource. Provide your conversation ID and HUMAN_AGENT or AUTOMATED_AGENT for the role field.

REST

Before using any of the request data, make the following replacements:

PROJECT_ID: your Cloud project ID
LOCATION_ID: your location ID
CONVERSATION_ID: your conversation ID

HTTP method and URL:

POST https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/conversations/CONVERSATION_ID/participants

Request JSON body:

{
  "role": "HUMAN_AGENT",
}

To send your request, expand one of these options:

curl (Linux, macOS, or Cloud Shell)

Save the request body in a file named request.json, and execute the following command:

curl -X POST \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     -H "x-goog-user-project: PROJECT_ID" \
     -H "Content-Type: application/json; charset=utf-8" \
     -d @request.json \
     "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/conversations/CONVERSATION_ID/participants"

PowerShell (Windows)

Save the request body in a file named request.json, and execute the following command:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred"; "x-goog-user-project" = "PROJECT_ID" }

Invoke-WebRequest `
    -Method POST `
    -Headers $headers `
    -ContentType: "application/json; charset=utf-8" `
    -InFile request.json `
    -Uri "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/conversations/CONVERSATION_ID/participants" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "name": "projects/PROJECT_ID/locations/LOCATION_ID/conversations/CONVERSATION_ID/participants/PARTICIPANT_ID",
  "role": "HUMAN_AGENT"
}

The path segment after participants contains your new agent-participant ID.

Analyze message

There are two ways to analyze messages:

During a conversation. To add and analyze an agent message for the conversation, call the analyzeContent method on the Participant resource. Provide the conversation ID and agent-participant ID. To add and analyze a user message for the conversation, call the analyzeContent method on the Participant resource. Provide the conversation ID and user-participant ID. Don't make duplicate calls to the analyzeContent method if it was called for other Dialogflow features.
After a conversation. You can use this option if you don't use the analyzeContent method during the conversation. Instead, you can use the batchCreate method on the messages resource to ingest historical messages of the conversation.

Generate a summary

You can get a suggestion at any time for the latest message from either participant.

Call the suggestConversationSummary method on the conversations.suggestions resource. Provide the conversation ID and the latest message ID from either participant.
Call the generateStatelessSuggestion method. Provide the ConversationContext, generator name, and trigger event, which should be MANUAL_CALL.

REST

Before using any of the request data, make the following replacements:

PROJECT_ID: your Cloud project ID
LOCATION_ID: your location ID
GENERATOR_ID: your generator ID

HTTP method and URL:

POST https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/statelessSuggestion:generate

Request JSON body:

{
  "parent": "projects/PROJECT_ID/locations/LOCATION_ID",
  "generatorName": "projects/PROJECT_ID/locations/LOCATION_ID/generators/GENERATOR_ID",
  "conversationContext" {
    "messageEntries": [{
      "role": "HUMAN_AGENT",
      "text": "Hi, this is ABC messaging, how can I help you today?",
      "languageCode": "en-US"
    }, {
      "role": "END_USER",
      "text": "I want to return my order, it is broken",
      "languageCode": "en-US"
    }]
  },
  "triggerEvents": [
    "MANUAL_CALL"
  ]
}

To send your request, expand one of these options:

curl (Linux, macOS, or Cloud Shell)

Save the request body in a file named request.json, and execute the following command:

curl -X POST \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     -H "x-goog-user-project: PROJECT_ID" \
     -H "Content-Type: application/json; charset=utf-8" \
     -d @request.json \
     "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/statelessSuggestion:generate"

PowerShell (Windows)

Save the request body in a file named request.json, and execute the following command:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred"; "x-goog-user-project" = "PROJECT_ID" }

Invoke-WebRequest `
    -Method POST `
    -Headers $headers `
    -ContentType: "application/json; charset=utf-8" `
    -InFile request.json `
    -Uri "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/locations/LOCATION_ID/statelessSuggestion:generate" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "generatorSuggestion": {
    "summarySuggestion": {
      "summarySections": [
        {
          "section": "situation",
          "summary": "customer wants to return an order."
        },
        {
          "section": "resolution",
          "summary": "N"
        }
      ]
    }
  }
}

The response contains the summary.

Complete the conversation

To complete the conversation, call the complete method on the conversations resource. Provide the conversation ID.

REST

Before using any of the request data, make the following replacements:

PROJECT_ID: your GCP project ID
CONVERSATION_ID: the ID you received when creating the conversation

HTTP method and URL:

POST https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/conversations/CONVERSATION_ID:complete

To send your request, expand one of these options:

curl (Linux, macOS, or Cloud Shell)

Execute the following command:

curl -X POST \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     -H "x-goog-user-project: PROJECT_ID" \
     -H "Content-Type: application/json; charset=utf-8" \
     -d "" \
     "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/conversations/CONVERSATION_ID:complete"

PowerShell (Windows)

Execute the following command:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred"; "x-goog-user-project" = "PROJECT_ID" }

Invoke-WebRequest `
    -Method POST `
    -Headers $headers `
    -Uri "https://dialogflow.googleapis.com/v2/projects/PROJECT_ID/conversations/CONVERSATION_ID:complete" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "name": "projects/PROJECT_ID/conversations/CONVERSATION_ID",
  "lifecycleState": "COMPLETED",
  "conversationProfile": "projects/PROJECT_ID/conversationProfiles/CONVERSATION_PROFILE_ID",
  "startTime": "2018-11-05T21:05:45.622Z",
  "endTime": "2018-11-06T03:50:26.930Z"
}

Summarization with custom sections

Predefined sections

Create a summarization generator

Console

REST

curl (Linux, macOS, or Cloud Shell)

PowerShell (Windows)

Test a summarization generator

Console

Configure a conversation profile

Console

REST

curl (Linux, macOS, or Cloud Shell)

PowerShell (Windows)

Verify with the simulator

Conversations at runtime

Create a conversation

REST

curl (Linux, macOS, or Cloud Shell)

PowerShell (Windows)

Create a user-participant

REST

curl (Linux, macOS, or Cloud Shell)

PowerShell (Windows)

Create an agent-participant

REST

curl (Linux, macOS, or Cloud Shell)

PowerShell (Windows)

Analyze message

Generate a summary

REST

curl (Linux, macOS, or Cloud Shell)

PowerShell (Windows)

Complete the conversation

REST

curl (Linux, macOS, or Cloud Shell)

PowerShell (Windows)

Previous versions