$rag.chat.retrieveChunksFromChat

The method returns chunks relevant to the user’s query from a knowledge base chat.

caution

The method is available only in the ECMAScript 6 runtime and is asynchronous.

Syntax

The method accepts the following arguments:

Argument	Type	Description	Required
secretName	String	Name of the knowledge base secret.	Yes
chatId	Integer	Chat identifier. Use the identifier returned by the $rag.chat.create method.	Yes
query	String	Text of the user query.	Yes
settings	Object	Chunk search settings. By default, the settings from the chat are used. The settings format matches the format of the identically named object in the POST /api/knowledge-hub/chat/{chatId}/retrieve request in the Tovie Data Agent API.	No
timeout	Number	Timeout in milliseconds for the method execution. If the timeout is exceeded, an error will occur. By default, the timeout is not set.	No

Positional arguments

Specify the arguments in order:

await $rag.chat.retrieveChunksFromChat("MyKnowledgeHub", 12345, "What does the Example service do?", undefined, 5000);

Via object

Pass an object whose fields match the names of the arguments:

await $rag.chat.retrieveChunksFromChat({
    secretName: "MyKnowledgeHub",
    chatId: 12345,
    query: "What does the Example service do?",
    timeout: 5000
});

Return value

The method returns an object with a list of chunks. They are sorted in descending order of relevance.

{
    "chunks": [
        {
            "score": 0.7486038,
            "content": "The Example service processes user requests …",
            "docId": "Documentation.pdf",
            "metadata": null
        },
        {
            "score": 0.7337575,
            "content": "The Example service can handle high loads …",
            "docId": "Services.pdf",
            "metadata": null
        }
    ]
}

The object has the same format as the response to the POST /api/knowledge-hub/chat/{chatId}/retrieve request in the Tovie Data Agent API.

How to use

state: Chunks
    intent!: /question
    scriptEs6:
        $client.chat = $client.chat || await $rag.chat.create("MyKnowledgeHub");

        const chunks = await $rag.chat.retrieveChunksFromChat("MyKnowledgeHub", $client.chat.id, $request.query);
        $reactions.answer(chunks.chunks[0].content);

In this state:

1. If there is no chat yet, the bot creates a chat.
2. The bot sends the most relevant chunk to the user.