Chatbot

A chatbot is a component that responds to questions sent by users in chat format with answers in chat format.
By linking with resources built using external services, you can use chatbot functions and chatbot functions that utilize RAG.

Note

Regarding integration with external services, we assume that the user has basic knowledge of or experience with the following technologies.

• Generative AI and RAG
• AWS Bedrock
• Amazon Bedrock Knowledge Bases
• Azure OpenAI
• Azure AI Search
• GCP Vertex AI
• Vertex AI Vector Search
• Pinecone

Dialog Property

ID	String		Auto setting
Width	Number String		640
Height	Number String		480
X	Number String
Y	Number String
Background color	String Color picker
Line	Boolean	・On ・Off	On
Width	Number		1
Roundness size	Number		0
Color	String Color picker		#DDDDDD
Agent name colour	String Color picker
Agent message color	String Color picker
Agent message background color	String Color picker
User name colour	String Color picker
User message color	String Color picker
User message background color	String Color picker		#BAD6F2
Submit button color	String Color picker
Message form input text color	String Color picker
Message form background color	String Color picker
Disclaimer Color	String Color picker
Read only	True/False	・On ・Off	Off
Visible	True/False	・On ・Off	On
Agent name	String		Agent
User name	String		You
Agent avatar	String		https://template-ui-image.webperformer.jp/image/Chatbot-Agent_alpha.svg
User avatar	String		https://template-ui-image.webperformer.jp/image/Chatbot-User_alpha.svg
RAG	True/False	・On ・Off	On
Service name	Select	・AWS Bedrock ・Azure Open AI ・GCP Vertex AI	AWS Bedrock
AI settings	JSON
AI credentials	String selection dialog
System prompt	String
Message form value	String
Message form placeholder	String		・Type your message… (default) ・Type your message… (English) ・メッセージを入力してください… (Japanese)
Number of message form rows	Number		1
Message	Tabular format		・[{"role":"agent","content":"Hello! How can I help you today?"}]（Default） ・[{"role":"agent","content":"Hello! How can I help you today?"}]（English） ・[{"role":"agent","content":"こんにちは！今日はどんなことをお手伝いしましょうか？"}]（Japanese）
Disclaimer	String		• AI-generated. May contain errors. (Default) •AI-generated. May contain errors. (English) • AI 応答のため、誤りを含む可能性があります。 (Japanese)

Action Property

Item	Property	Data Type	Option	Default Value
ID	id	string		Automatic setting
Width	width	number string		640
Height	height	number string		480
X	x	number string
Y	y	number string
Background Color	backgroundColor	string
Line	border	boolean	・true ・false	true
Width	borderWidth	number		1
Rounded Size	roundedSize	number		0
Color	borderColor	string		#DDDDDD
Agent name colour	agentNameColor	string
Agent Message Color	agentMessageColor	string
Agent Message Background Color	agentMessageBackgroundColor	string
User name color	userNameColor	string
User message color	userMessageColor	string
User message background color	userMessageBackgroundColor	string		#BAD6F2
Submit button color	submitButtonColor	string
Message form input text color	messageFormInputColor	string
Message form background color	messageFormBackgroundColor	string
Disclaimer Color	disclaimerColor	string
Read only	readOnly	boolean	・true ・false	false
Visibility	boolean	・true ・false	true
Agent name	agentName	string		Agent
User name	userName	string		You
Agent avatar	agentAvatar	string		https://template-ui-image.webperformer.jp/image/Chatbot-Agent_alpha.svg
User avatar	userAvatar	string		https://template-ui-image.webperformer.jp/image/Chatbot-User_alpha.svg
RAG	rag	boolean	・true ・false	true
Service name	serviceName	string	・awsBedrock ・azureOpenAI ・gcpVertexAI	awsBedrock
Message form value	messageFormValue	string
Message form placeholder	messageFormPlaceholder	string		・Type your message… (default) ・Type your message… (English) ・メッセージを入力してください… (Japanese)
Number of message form rows	messageFormRows	number		1
Message	messages	array of object		・[{"role":"agent","content":"Hello! How can I help you today?"}]（Default） ・[{"role":"agent","content":"Hello! How can I help you today?"}]（English） ・[{"role":"agent","content":"こんにちは！今日はどんなことをお手伝いしましょうか？"}]（Japanese）
Disclaimer	disclaimer	string		• AI-generated. May contain errors. (Default) •AI-generated. May contain errors. (English) • AI 応答のため、誤りを含む可能性があります。 (Japanese)

Note

• Default icons with opaque backgrounds can be used by setting the following paths for the [Agent Avatar] and [User Avatar] properties.
  • Agent Avatar: https://template-ui-image.webperformer.jp/image/Chatbot-Agent_opaque.svg
  • User Avatar: https://template-ui-image.webperformer.jp/image/Chatbot-User_opaque.svg
• For the value of the [AI Settings] property, refer to the ‘AI Settings Specifications’ listed below and configure the connection information according to the service you are using and whether or not RAG is available.
• For the value of the [AI Credentials] property, prepare a JSON file containing your credentials with any file name in File Manager, and set the file path according to the service you are using and whether or not you are using RAG.
• It is not possible to reset the chat content.
• If there is a difference between the value of the [System Prompt] property and the language of the question, there may be a difference in the language of the response.
• The [System Prompt] property allows up to four lines of text.
• The [Message Form Value] property allows you to set the initial value of the message input field.
• The [Message Form Line Count] property can be set to a value between 1 and 10.
• The following multi-language properties are executed with the default values without language specification when obtaining and setting values in actions.
  • [Agent Name]
  • [User Name]
  • [Message Form Value]
  • [Message Form Placeholder]
  • [Message]
  • [Caution text]
• When setting a value for the [Message] property from an action, the value must have the following structure.

[
　{ "role": "user", "content": "How?"},
　{ "role": "agent", "content": "Just try"}
]

• If “user” is specified for “role”, the value of “content” will be drawn in the transmission message area, and if “agent” is specified, it will be drawn in the response message area.

External Services

The following external services can be linked.

• AWS Bedrock
• Azure OpenAI
• GCP VertexAI

The combinations of retrievers and embedding models supported for each service are shown in the table below.

Service	Retriever	Embedding model
AWS Bedrock	Amazon Bedrock Knowledge Bases	No setup required
	Pinecone	Amazon Bedrock
Azure Open AI	Azure AI Search	Azure Open AI
	Pinecone	Azure Open AI
GCP Vertex AI	Vertex AI Vector Search	Vertex AI
	Pinecone	Vertex AI

AI Settings Specifications

AI settings describe settings for using resources built with external services.
In addition, settings for controlling response content can be described as desired.
Please describe settings according to the specifications of the external services you use.

AWS Bedrock：
https://docs.aws.amazon.com/ja_jp/bedrock/latest/userguide/model-parameters.html

Azure OpenAI：
https://learn.microsoft.com/ja-jp/azure/ai-foundry/openai/reference

GCP Vertex AI：
https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/inference?hl=ja#request
https://ai.google.dev/gemini-api/docs/safety-settings?hl=ja#javascript
https://cloud.google.com/vertex-ai/docs/general/googleapi-access-methods?hl=ja

Note

• Even if there is only one string to specify, please describe it in string array format in the list of tokens or strings that stop output (stop, stopSequences).
• The following values can be specified for the Azure OpenAI search type (search):
  • similarity (default): vector search only
  • similarity_hybrid: hybrid of vector search and keyword search
  • semantic_hybrid: similarity_hybrid + semantic ranking processing
• For the retriever name (source), specify the following values depending on the service you are using:
  • Amazon Bedrock Knowledge Bases: ‘amazon-bedrock-knowledge-bases’
  • Azure AI Search: ‘azure-ai-search’
  • Vertex AI Vector Search: ‘vertex-ai-vector-search’
  • Pinecone: ‘pinecone’

Service	Pattern	Setting items (* indicates required fields)
AWS Bedrock	chat	{ "chat": { "region": <Region (string) *>, "model": <AI model identifier (string) *>, "topP": <Probability value (numerical value) that controls the diversity of generation>, "temperature": <Random value (numerical value) of generation>, "maxTokens": <Maximum number of tokens to output in response (numerical value)>, "stop": <List of tokens or strings that stop output (string array)> } }
	chat + rag Retriever：Amazon Bedrock Knowledge Bases	{ "chat": { "region": <Region（string）*>, "model": <AI model identifier（string）* >, "topP": <Probability value (numerical value) that controls the diversity of generation>, "temperature": <Random values (numbers) generated>, "maxTokens": <Maximum number of tokens to output in response (numerical value)>, "stop": <List of tokens or strings (string array) to stop outputting> }, "rag": { "source": <Retriever name (string) *>, "param": { "topK": <Maximum number of documents to be retrieved (numerical value) * >, "knowledgeBaseId": <Knowledge Base ID（string）*>, "region": <Region（string）*> } } }
	chat + rag Retriever：Pinecone	{ "chat": { "region": <Region（string）*>, "model": <AI model identifier（string）*>, "topP": <Probability value (numerical value) that controls the diversity of generation>, "temperature": <Random values (numbers) generated>, "maxTokens": <Maximum number of tokens to output in response (numerical value)>, "stop": <List of tokens or strings (string array) to stop outputting> }, "rag": { "source": <Retriever Name（string）*>, "param": { "indexName": <Index name (string) *>, "namespace": <Namespace (string)>, }, "embedding": { "region": <Region（string）*>, "model": <Embedding model identifier (string) *> }, "retriever": { "k": <Maximum number of documents to be retrieved (numerical value) *> } } }
Azure OpenAI	chat	{ "chat":{ "azureOpenAIApiDeploymentName": <Deployment name (string) *>, "azureOpenAIApiInstanceName": <Resource name (string) *>, "azureOpenAIApiVersion": <API version (string) *>, "topP": <Probability value (number) that controls the diversity of generation>, "temperature": <Random value (numerical value) for generation>, "maxTokens": <Maximum number of tokens to output in response (number)>, "stop": <List of tokens or strings to stop outputting (string array)>, "frequencyPenalty": <Penalty value (number) to suppress repetition of the same word>, "presencePenalty": <Penalty value (number) to encourage the introduction of new topics> } }
	chat+rag Retriever：Azure AI Search [Notes] The index structure should be designed as follows ・ id (String) => Retrievable, searchable ・ content (String) => Retrievable, searchable ・ content_vector (SingleCollection) => Searchable ・ metadata (ComplexType)   〇 source (String) => Retrievable, searchable   〇 chunk (String) => Searchable   〇 attributes (ComplexTypeCollection)     ・ key (String) => Filterable     ・ value (String) => Searchable	{ "chat": { "azureOpenAIApiDeploymentName": <Deployment name (string) *>, "azureOpenAIApiInstanceName": <Resource name (string) *>, "azureOpenAIApiVersion": <API version (string) *>, "topP": <Probability value (numerical value) that controls the diversity of generation>, "temperature": <Random values (numbers) generated>, "maxTokens": <Maximum number of tokens to output in response (numerical value)>, "stop": <List of tokens or strings (string array) to stop outputting>, "frequencyPenalty": <Penalty value (numerical value) for suppressing repetition of the same word>, "presencePenalty": <Penalty value (numerical value) to encourage the introduction of new topics> }, "rag": { "source": <Retriever Name(string)*>, "param": { "endpoint": <End point（string）*>, "indexName": <Index name (string)*>, "search": { "type": <Search type（string）> } }, "embedding": { "azureOpenAIApiDeploymentName": <Deployment Name（string）*>, "azureOpenAIApiInstanceName": <Resource name（string）*>, "azureOpenAIApiVersion": <API Version（string）*>, "dimensions": <Vectorisation dimension (numerical value)> }, "retriever": { "k": <Maximum number of documents to be retrieved (numerical value) *> } } }
	chat+rag Retriever：Pinecone	{ "chat": { "azureOpenAIApiDeploymentName": <Deployment name (string) *>, "azureOpenAIApiInstanceName": <Resource Name（string）*>, "azureOpenAIApiVersion": <API version（string）*>, "topP": <Probability value (numerical value) that controls the diversity of generation>, "temperature": <Random value (numerical value) of generation>, "maxTokens": <Maximum number of tokens to output in response (numerical value)>, "stop": <List of tokens or strings that stop output (string array)>, "frequencyPenalty": <Penalty value (numerical value) to suppress repetition of the same word>, "presencePenalty": <Penalty value (numerical value) to encourage the introduction of new topics> }, "rag": { "source": <Retriever Name（string）*>, "param": { "indexName": <Index Name（string）*>, "namespace": <Namespace (string)> }, "embedding": { "azureOpenAIApiDeploymentName": <Deployment Name（string）*>, "azureOpenAIApiInstanceName": <Resource Name（string）*>, "azureOpenAIApiVersion": <API Version（string）*>, "dimensions": <Vectorisation dimension (numerical value)> }, "retriever": { "k": <Maximum number of documents to be retrieved (numerical value) *> } } } [Supplement] Instead of azureOpenAIApiInstanceName, it is also possible to specify azureOpenAIEndpoint. { "azureOpenAIEndpoint": <Endpoint　URL（string）*>, }
GCP Vertex AI	chat	{ "chat": { "model": <AI model identifier (string) *>, "topP": <Parameters for controlling output diversity (numerical values)>, "topK": <Select from the top K tokens in terms of probability (numerical value)>, "temperature": <Random value of output (the lower the value, the more deterministic) (numerical value)>, "maxOutputTokens": <Maximum number of tokens output in response (numerical value)>, "safetySettings": <Safety settings (array)>, "stopSequences": <List of strings that stop response output (string array)> } }
	chat+rag Retriever：Vertex AI Vector Search	{ "chat": { "model": <AI Model Identifier (string) *>, "topP": <Parameters for controlling output diversity (numerical values)>, "topK": <Select from the top K tokens in terms of probability (numerical value)>, "temperature": <Random value of output (the lower the value, the more deterministic) (numerical value)>, "maxOutputTokens": <Maximum number of tokens output in response (numerical value)>, "safetySettings": <Safety settings (array)>, "stopSequences": <List of strings that stop response output (string array)> }, "rag": { "source": <Retriever Name（String）*>, "param": { "index": <Index ID (string) *>, "indexEndpoint": <Index endpoint ID (string) *>, "location": <Region (string) *>, "endpoint": <API endpoint (string) *>, "deployedIndexId": <Deployed index ID (string) *> }, "docstore": { "bucket": <Cloud storage bucket name (string) *>, "prefix": <Cloud storage prefix name (string) *> }, "embedding": { "model": <Embedding model identifier (string) *> }, "retriever": { "k": <Maximum number of documents to retrieve (numeric) *> } } }
	chat+rag Retriever：Pinecone	{ "chat": { "model": <AI model identifier (string) *>, "topP": <Parameter to control the diversity of output (numerical value)>, "topK": <Select from the top K tokens in terms of probability (numerical value)>, "temperature": <Random value of output (the lower the value, the more deterministic) (numerical value)>, "maxOutputTokens": <Maximum number of tokens to output in response (numeric value)>, "safetySettings": <Security settings (array)>, "stopSequences": <List of strings to stop response output (string array)> }, "rag": { "source": <Retriever name (string) *>, "param": { "indexName": <Index name (string) *>, "namespace": <Namespace (string)> }, "embedding": { "model": <Embedding model identifier (string) *> }, "retriever": { "k": <Maximum number of documents to be retrieved (numerical value) *> } } }

AI Credential File Specification

AI Credential files contain authentication information for using resources built by external services.

Note

• For AWS, refer to the access key information of the IAM user who has permission to use the built resources.
• For Azure, refer to the API key information of the built resources.
• For GCP, refer to the project ID where the constructed resources are located and the key information for the service account with permissions to use the constructed resources. For scope, refer to the following.
  https://cloud.google.com/docs/authentication?hl=ja#authorization-gcp

Services	Patterns	Setting items (* indicates required items)
AWS Bedrock	chat	{ "chat": { "credentials": { "accessKeyId": <Access Key ID (string) *>, "secretAccessKey": <Secret Access Key (string) *> } } }
	chat + rag Retriever：Amazon Bedrock Knowledge Bases	{ "chat": { "credentials": { "accessKeyId": <Access Key ID (string) *>, "secretAccessKey": <Secret access key (string) *> } }, "rag": { "param": { "credentials": { "accessKeyId": <Access Key ID (string) *>, "secretAccessKey": <Secret access key (string) *> } } } }
	chat + rag Retriever：Pinecone	{ "chat": { "credentials": { "accessKeyId": <Access Key ID（string）*>, "secretAccessKey": <Secret Access Key（string）*> } }, "rag": { "param": { "config": { "apiKey": <API Key（string）*> } }, "embedding": { "credentials": { "accessKeyId": <Access Key ID（string）*>, "secretAccessKey": <Secret Access Key（string）*> } } } }
Azure OpenAI	chat	{ "chat":{ "azureOpenAIApiKey": <API Key（string）*> } }
	chat+rag Retriever：Azure AI Search	{ "chat": { "azureOpenAIApiKey": <API Key（string）*> }, "rag": { "param": { "key": <API Key（string）*> }, "embedding": { "azureOpenAIApiKey": <API Key（string）*> } } }
	chat+rag Retriever：Pinecone	{ "chat": { "azureOpenAIApiKey": <API Key（string）*> }, "rag": { "param": { "config": { "apiKey": <API Key（string）*> } }, "embedding": { "azureOpenAIApiKey": <API Key（string）*> } } }
GCP Vertex AI	chat	{ "chat": { "authOptions": { "credentials": <Service account key file json (object) *>, "projectId": <Project ID (string) *> } } }
	chat+rag Retriever：Vertex AI Vector Search	{ "chat": { "authOptions": { "credentials": <Service account key file json (object) *>, "projectId": <Project ID（string）*> } }, "rag": { "param": { "authOptions": { "credentials": <Service account key file json (object) *>, "projectId": <Project ID（string）*>, "scopes": <Scope (string array) *> } }, "docstore": { "authOptions": { "credentials": <Service account key file json (object) *>, "projectId": <Project ID（string）*> } }, "embedding": { "authOptions": { "credentials": <Service account key file json (object) *>, "projectId": <Project ID（string）*> } } } }
	chat+rag Retriever：Pinecone	{ "chat": { "authOptions": { "credentials": <Service account key file json (object) *>, "projectId": <Project ID (string) *> } }, "rag": { "param": { "config": { "apiKey": <API Key（string）*> } }, "embedding": { "authOptions": { "credentials": <Service account key file json (object) *>, "projectId": <Project ID (string) *> } } } }