Skip to content

Using Document Search

Document Search in Generate Enterprise allows you to interact with your files, ask AI-powered questions, and retrieve relevant information. You can search without selecting documents or enhance your queries by uploading files.

General AI Search (No Documents Needed)

Before selecting documents, you can ask questions using the built-in AI model.

  • Type your query in the Ask me anything box.
  • The model will generate a response based on general knowledge.
  • Workflow cards are also available for specialized tasks.

You can also use the following interactive tools to enhance your experience:

Search Online

Click the search icon to retrieve the latest information from the web. The AI will provide sources for the retrieved data.

Click to Collapse

Web Search

Example:
If you ask, "What is the latest price of Bitcoin?", the system will search for the most recent market data.

Click to Collapse

Google Web Search

Voice Query

Click the microphone icon to speak your question instead of typing.

Click to Collapse

Voice Search

Selecting & Uploading Documents

To enable AI-powered document search, you first need to upload and select files.

Attach a Document

Click the attach icon to select a file from your saved documents. Once attached, the AI will reference the document to provide responses.

Click to Collapse

Attach Document

Alternative Ways to Upload Documents

Aside from directly attaching a document, you can also upload files in two other ways:

  1. If you have not uploaded any documents yet, you will see an option in the file panel to Pick Documents.

  2. Manage your uploaded documents in Settings:

    • Ensure Document Search is selected from the menu.
    • Navigate to My Documents in the left sidebar.
    • Manage your uploaded documents or upload new ones from this panel.
    • More details can be found in Configuring Settings.
Click to Collapse

Manage Documents

Document Processing

After selecting a document, it will be processed and embedded into the vector database.

  1. Processing Stage – After uploading, you will see a processing indicator. This may take a few seconds to a few minutes, depending on the file size.
  2. Completed Status – Once finished, the document’s status will update to Completed.
Click to Collapse

Document Processing Completed

  1. Accessing Uploaded Documents – Your document will now appear in Document Search. If it does not appear immediately, try switching to another tab and back. To filter your documents by integration, use the Integration Dropdown Menu to select the storage location.

For example, if you had just uploaded a file from Google Drive, switch the integration here to Google Drive.

Click to Collapse

Document Processing Integration

Chatting with Your Document

Now that your document is uploaded, you can start interacting with it.

  1. Select one or multiple documents by clicking on your uploaded files.

  2. Choose an action:

    • Summarize – Generates a summary of the document.
    • Ask a Question – Lets you chat with the document and retrieve specific details.

Summarization Example

Click to Collapse

Summarize Document

Ask a Question

Click to Collapse

Ask Question

Understanding the Chat Results Toolbar

At the bottom of the chat interface, there are various tools to enhance your interaction:

  • Voice Output – Reads the generated response aloud.
  • Copy to Clipboard – Copies the response text for easy sharing.
  • Feedback (Thumbs Up/Down) – Provides feedback on the AI-generated answer.
  • Export as PDF – Saves the response as a downloadable PDF.
  • IBM WatsonX Evaluation – Click to analyze AI performance metrics.
Click to Collapse

Chat Results

WatsonX Evaluation Metrics

When you click on IBM WatsonX Evaluation, the system will analyze your document query results using various retrieval performance metrics.

Metric Description Example Score
Answer Relevance Measures how well the AI-generated answer aligns with expected responses. 1
Context Relevance Determines the accuracy of retrieved context in answering a query. 0.8
Retrieval Precision Measures the fraction of retrieved documents that are relevant. 0.3333
Average Precision Computes mean precision for ranked retrieval results. 1
Reciprocal Rank Evaluates the rank of the first relevant result. 1
Faithfulness Checks if the generated response is factually correct. 1
Hit Rate Measures how often relevant results appear in retrieval. 1
NDCG Normalized Discounted Cumulative Gain for ranking quality. 1

Managing Your Chat History

To continue past conversations or review old queries:

  • Click a conversation to continue chatting or delete it if no longer needed.
  • Click New Chat to start a fresh conversation.
  • To view past interactions, select the History tab next to My Files.
Click to Collapse

Chat History

Now that you’ve learned how to use Document Search, explore Advanced Settings to customize your experience.