Box Developer Documentation

Ask questions to Box AI

Ask questions to Box AI

Box AI API is available to all Enterprise Plus customers.

Box AI API allows you to ask a question about a supplied file or a set of files, and get a response based on the content. For example, while viewing a document in Box, you can ask Box AI to summarize the content.

Before you start

Make sure you followed the steps listed in getting started with Box AI to create a custom app and authenticate.

Send a request

To send a request containing your question, use the POST /2.0/ai/ask endpoint and provide the mandatory parameters.

cURL
curl -i -L POST "https://api.box.com/2.0/ai/ask" \
     -H "content-type: application/json" \
     -H "authorization: Bearer <ACCESS_TOKEN>" \
     -d '{
         "mode": "single_item_qa",
         "prompt": "What is the value provided by public APIs based on this document?",
         "items": [
            {
            "type": "file",
            "id": "9842787262"
            }
         ],
         "dialogue_history": [
              {
              "prompt": "Make my email about public APIs sound more professional",
              "answer": "Here is the first draft of your professional email about public APIs",
              "created_at": "2013-12-12T10:53:43-08:00"
              }
          ],
          "include_citations": true,
          "ai_agent": {
            "type": "ai_agent_ask",
            "long_text": {
              "model": "azure__openai__gpt_4o_mini",
              "prompt_template": "It is `{current_date}`, and I have $8000 and want to spend a week in the Azores. What should I see?",
            },
            "basic_text": {
              "model": "azure__openai__gpt_4o_mini",
           }
         }
      }'
TypeScript Gen
await client.ai.createAiAsk({
  mode: 'multiple_item_qa' as AiAskModeField,
  prompt: 'Which direction sun rises?',
  items: [
    new AiItemBase({
      id: fileToAsk1.id,
      type: 'file' as AiItemBaseTypeField,
      content: 'Earth goes around the sun',
    }),
    new AiItemBase({
      id: fileToAsk2.id,
      type: 'file' as AiItemBaseTypeField,
      content: 'Sun rises in the East in the morning',
    }),
  ],
} satisfies AiAsk);
Python Gen
client.ai.create_ai_ask(
    CreateAiAskMode.MULTIPLE_ITEM_QA,
    "Which direction sun rises?",
    [
        AiItemBase(
            id=file_to_ask_1.id,
            type=AiItemBaseTypeField.FILE,
            content="Earth goes around the sun",
        ),
        AiItemBase(
            id=file_to_ask_2.id,
            type=AiItemBaseTypeField.FILE,
            content="Sun rises in the East in the morning",
        ),
    ],
)
.NET Gen
await client.Ai.CreateAiAskAsync(requestBody: new AiAsk(mode: AiAskModeField.MultipleItemQa, prompt: "Which direction sun rises?", items: Array.AsReadOnly(new [] {new AiItemBase(id: fileToAsk1.Id, type: AiItemBaseTypeField.File) { Content = "Earth goes around the sun" },new AiItemBase(id: fileToAsk2.Id, type: AiItemBaseTypeField.File) { Content = "Sun rises in the East in the morning" }})));
Swift Gen (Beta)
try await client.ai.createAiAsk(requestBody: AiAsk(mode: AiAskModeField.multipleItemQa, prompt: "Which direction sun rises?", items: [AiItemBase(id: fileToAsk1.id, type: AiItemBaseTypeField.file, content: "Earth goes around the sun"), AiItemBase(id: fileToAsk2.id, type: AiItemBaseTypeField.file, content: "Sun rises in the East in the morning")]))
Java
BoxAIResponse response = BoxAI.sendAIRequest(
    api,
    "What is the content of the file?",
    Collections.singletonList("123456", BoxAIItem.Type.FILE),
    BoxAI.Mode.SINGLE_ITEM_QA
);
Python
items = [{
    "id": "1582915952443",
    "type": "file",
    "content": "More information about public APIs"
}]
ai_agent = {
    'type': 'ai_agent_ask',
    'basic_text_multi': {
        'model': 'openai__gpt_3_5_turbo'
    }
}
answer = client.send_ai_question(
    items=items, 
    prompt="What is this file?",
    mode="single_item_qa",
    ai_agent=ai_agent
)
print(answer)
.NET
BoxAIResponse response = await client.BoxAIManager.SendAIQuestionAsync(
    new BoxAIAskRequest
    {
        Prompt = "What is the name of the file?",
        Items = new List<BoxAIAskItem>() { new BoxAIAskItem() { Id = "12345" } },
        Mode = AiAskMode.single_item_qa
    };
);
Node
client.ai.ask(
    {
        prompt: 'What is the capital of France?',
        items: [
            {
                type: 'file',
                id: '12345'
            }
        ],
        mode: 'single_item_qa'
    })
    .then(response => {
        /* response -> {
            "answer": "Paris",
            "created_at": "2021-10-01T00:00:00Z",
            "completion_reason": "done"
        } */
    });

Parameters

To make a call, you need to pass the following parameters. Mandatory parameters are in bold.

ParameterDescriptionAvailable valuesExample
modeThe type of request. It can be a question about a single file or a set of files. For a single file, Box AI API supports up to 1MB of text representation. If the file size exceeds 1MB, the first 1MB of text representation will be processed. If you want to list multiple files, the limit is 25 files. If you set mode to single_item_qa, the items array can list only one element.single_item_qa, multiple_item_qasingle_item_qa
promptThe question about your document or content. The prompt's length cannot exceed 10000 characters.What is this document about?
dialogue_history.promptThe prompt previously provided by the client and answered by the Large Language Model (LLM).Make my email about public APIs sound more professional
dialogue_history.answerThe answer previously provided by the LLM.Here is a draft of your professional email about public APIs.
dialogue_history.created_atThe ISO date formatted timestamp of when the previous answer to the prompt was created.2012-12-12T10:53:43-08:00
include_citationsSpecifies if the citations should be returned in the answer.true, falsetrue
items.idThe Box file ID you want to provide as input.112233445566
items.typeThe type of the provided input. Currently, it can be a single file or multiple files.filefile
items.contentThe content of the item. Usually it is the text representation.An application programming interface (API) is a way for two or more computer programs or components to communicate with each other. It is a type of software interface...
ai_agentThe AI agent used to override the default agent configuration. You can use this parameter replace the default LLM with a custom one using the model parameter for shorter and longer texts, tweak the base prompt to allow for a more customized user experience, or change an LLM parameter, such as temperature, to make the results more or less creative. Before you use the ai_agent parameter, you can get the default configuration using the GET 2.0/ai_agent_default request. For specific use cases, see the AI model overrides tutorial.

Use cases

Ask questions about an item

This example shows how to ask a question about one or more items using the POST ask/ai API. When using this endpoint, remember to specify the mode parameter depending on the number of items you want to supply.

curl -i -L POST "https://api.box.com/2.0/ai/ask" \
     -H "content-type: application/json" \
     -H "authorization: Bearer <ACCESS_TOKEN>" \
     -d '{
    "mode": "single_item_qa",
    "items": [
        {
            "id": "12345678",
            "type": "file"
        }
    ],
    "prompt": "List the guidelines on creating questions in Box AI for Documents"
}'

The response will be as follows:

{
    "answer": "The guidelines for working with questions in Box AI for Documents are as follows:\n\n1. Box AI pulls information only from the document loaded in preview.\n2. If questions fall outside the scope of the document, Box AI will inform you that it cannot answer.\n3. Be specific when asking questions; use parameters like numbered lists, brevity, tables, and central themes or key points.\n4. Aim to stay within the scope of the document.\n5. Focus on text-based responses only.",
    "created_at": "2024-11-04T02:30:09.557-08:00",
    "completion_reason": "done"
}

Ask questions with content parameter

If you use the content parameter as the source of input for Box AI, it will use it as the primary source.

curl -i -L POST "https://api.box.com/2.0/ai/ask" \
     -H "content-type: application/json" \
     -H "authorization: Bearer <ACCESS_TOKEN>" \
     -d '{
    "mode": "single_item_qa",
    "items": [
        {
            "id": "12345678",
            "type": "file",
            "content": "This is a document about Box AI For documents. It consists of the functionality summary and guidelines on how to work with Box AI. Additionally, it provides a set of best practices for creating questions."
        }
    ],
    "prompt": "List the guidelines on creating questions in Box AI for Documents"
}'

The response to this request is based on the content parameter instead of the file's content:

{
    "answer": "The document does not provide specific guidelines on working with questions in Box AI for Documents. It only mentions that it includes a set of best practices for creating questions, but the details of those guidelines are not included in the text provided. If you have more information or another document, I can help further!",
    "created_at": "2024-11-04T02:31:51.125-08:00",
    "completion_reason": "done"
}

Ask questions with citations parameter

Setting the citations parameter to true causes the response to include excerpts from source file or files Box AI used to compile the answer.

curl -i -L POST "https://api.box.com/2.0/ai/ask" \
     -H "content-type: application/json" \
     -H "authorization: Bearer <ACCESS_TOKEN>" \
     -d '{
    "mode": "multiple_item_qa",
    "include_citations": true,
    "items": [
        {
            "id": "12345678",
            "type": "file"
        }
    ],
    "prompt": "List the guidelines on working with responses in Box AI for Documents"
}'

The resulting answer includes the source file and direct content citations.

{
    "answer": "The guidelines for working with questions in Box AI for Documents are as follows:\n\n1. Box AI pulls information only from the document loaded in preview, and cannot answer questions outside its scope.\n2. Be specific when asking questions; use parameters like numbered lists, brevity, tables, and central themes or key points.\n3. Examples of better phrasing include asking for a numbered list of key points instead of just \"list key points,\" and requesting a succinct outline of important points rather than a general inquiry about the document's purpose.\n4. Stay within the scope of the document and focus on text-based responses only.",
    "created_at": "2024-11-04T02:35:00.578-08:00",
    "completion_reason": "done",
    "citations": [
        {
            "type": "file",
            "id": "12345678",
            "name": "Box AI for Documents.docx",
            "content": "Guidelines for Box AI questions\nBox AI pulls information only from the document you loaded in preview."
        },
        {
            "type": "file",
            "id": "12345678",
            "name": "Box AI for Documents.docx",
            "content": "If you ask any questions outside of the scope of the document, Box AI informs you that it cannot answer the question with the information provided."
        },
        {
            "type": "file",
            "id": "12345678",
            "name": "Box AI for Documents.docx",
            "content": "As you ask Box AI to analyze your document, consider these suggestions:\n· Be as specific as possible."
        },
        {
            "type": "file",
            "id": "12345678",
            "name": "Box AI for Documents.docx",
            "content": "Box AI for Documents\n\nWhen viewing a document in Box, you can ask Box AI to summarize document content, search key points, and write outline drafts based on your document files."
        }
    ]
}