Documents Chatbot Component

About the documents chatbot component

a!documentsChatField( label, labelPosition, instructions, showWhen, helpTooltip, documents, height )

Creates a chatbot for users to ask questions about a curated set of documents.

The documents chatbot component uses Appian AI Copilot to answer questions about a curated set of documents, enabling users to interact with conversational AI to get quick insights. This component requires no third-party integration to get started. All you need to do is configure the parameters to begin chatting with your documents.

Parameters

Using the documents chatbot component

The following sections cover important details about the documents chatbot component.

Using the documents parameter

The documents chatbot component requires documents to exist in Appian as a document object. The component doesn't support scanned PDFs or image-based text. Use text-based PDFs for the best experience.

External documents managed by record types are not supported in the documents chatbot component because they do not exist as document objects in Appian. Only documents uploaded and stored directly in Appian can be used with this component.

The documents chatbot component supports chatting with multiple documents if their total size is less than 25 MB. Only English documents and questions are supported at this time. We recommend testing non-English documents and questions to ensure the results are satisfactory.

Document retention and re-indexing

Documents that haven't been accessed in 30 days need to be re-indexed the next time you chat with them. This re-indexing may cause the first response to take a little longer to load. To keep a document ready for quick use, chat with it again within 30 days.

Reliability of responses

As with any generative AI feature, the documents chatbot component may produce invalid or undesirable answers. Generative AI models sometimes produce "hallucinations," or responses that contain inaccurate information. Users should verify information in chat responses for accuracy.

The documents chatbot component may not provide consistent answers. The response the component provides today might differ from an answer to the same question with the same data tomorrow.

Asking effective questions

The documents chatbot component is designed to answer questions based on the content in the documents. To get the best results, your questions must be clear, specific, and directly related to the information in the documents.

Recommended questions

The documents chatbot component works well when you ask specific questions that can be answered by the text in the document or documents you are chatting with. Clear and direct questions yield the most accurate and relevant answers:

• Contract questions: If you are chatting with a contract, you could ask, "What is the effective date of the agreement?"
• Meeting notes: When interacting with a set of meeting notes, a good question would be, "What were the key decisions made in the project kickoff meeting?"
• Large documents: For very large documents, ask about a specific section, term, or topic to get a better response. For example, instead of asking, "What are the requirements in this report?", ask, "What are the requirements listed in the Technical Specifications section?"
• IT documentation: If you are chatting with IT documentation, you could ask, "How do I connect to the company wifi?"
• Policy documents: For a policy document, you could ask, "What does the act say about small business?" or "What does section 16 cover?"

Unsupported use cases

The documents chatbot component is not intended for broad or vague questions, or for summarizing documents. These types of questions may return an answer, but the response is often incomplete or less useful:

• Broad questions: Avoid general prompts like "What is this document about?" Instead, ask about specific details.
• Summarization requests: The component cannot reliably summarize a document or provide main points.
• Out of scope: The component only answers based on the document's text. It cannot provide author, title, or other metadata.
• Images and charts: The component cannot read images, charts, or scanned text without selectable text.
• Comparison questions: For example, "Compare these documents and explain how they are similar or different."

Compliance

This component is powered by private AI, meaning that your inputs and outputs are protected and remain within the boundaries of your Appian Cloud site at all times.

You will receive the following error until the feature is enabled on your site: This function is currently unavailable. Please contact your system administrator if this problem persists.

This feature is available for environments in select regions.

If your Appian environment isn't in a supported region, you can elect to send your data to a supported region. This doesn't change your environment's region. Contact Appian Support to learn more.

Security

The documents chatbot component enforces object security, ensuring that users can only interact with data they are authorized to access. Users must have at least Viewer permission on a document to ask questions and receive responses regarding that document.

Appian Mobile and Appian for Windows considerations

This component is not supported in the Appian Mobile or Appian for Windows apps. To use records chat, view the interface in a browser.

Examples

The following examples demonstrate the documents chatbot component.

To experiment with examples, copy and paste the expression into an interface object.

These examples reference record types from another environment and will error if you try to use them in yours. Use them as a reference only.

Documents chatbot with a record type collection

1
2
3
4
5
6
7
8
{
  a!documentsChatField(
    label: "Documents Chat",
    labelPosition: "COLLAPSED",
    documents: ri!documentCollection[recordType!SMRT Document Collection.relationships.smrtDocument.fields.documentId],
    height: "AUTO"
  )
}

Feature compatibility

Feedback