# Amazon Q Business features **Note** Not all features in Amazon Q Business Pro are available in Amazon Q Business Lite. For information about what's included in Amazon Q Business Lite and what's included in Amazon Q Business Pro, see [Amazon Q Business tiers](https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/tiers.html#user-sub-features). You must use the Amazon Q Business console to assign subscription tiers to your users. In addition to [enhancements](https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/enhancements.html), Amazon Q Business offers the following features: + **Filtering using metadata** – Use document attributes to customize and control the end user chat experience. Currently supported for use with the Amazon Q Business API. + **Source attribution with citations** – Verify responses using Amazon Q Business source attributions. + **Upload files to a chat** – Let end users upload files directly to a chat and use uploaded file data to perform web experience tasks. + **Personalizing chat responses** – Allow Amazon Q Business to customize chat responses to your end users by using the metadata that's associated with them—specifically, address and job related information—in your IAM Identity Center instance. + **Quick prompts** – Feature sample prompts to inform end users about the capabilities of their Amazon Q Business web experience. + **Hallucination mitigation** – Reduce Amazon Q Business chat hallucinations and improve response accuracy. + **Agentic RAG** – Enhances Amazon Q Business standard retrieval augmented generation (RAG) capabilities with agentic workflows. **Topics** + [ # Filtering chat responses using document attributes and metadata ](metadata-filtering.md) + [ # Source attribution with citations in Amazon Q Business ](source-attribution-citations.md) + [ # Upload files to a chat in Amazon Q Business ](upload-chat-files.md) + [ # Personalizing Amazon Q Business chat responses ](personalizing-chat-responses.md) + [ # Quick prompts in Amazon Q Business ](quick-prompts.md) + [ # Hallucination mitigation in Amazon Q Business ](hallucination-reduction.md) + [ # Agentic Retrieval Augmented Generation (RAG) in Amazon Q Business ](agentic-rag.md) # Filtering chat responses using document attributes and metadata Filtering using document attributes and metadata **Note** This section assumes that you have an understanding of [document attributes and how they work](https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/doc-attributes.html) in Amazon Q. The Amazon Q Business API includes a feature that lets you filter by document attribute. By using this feature, you can customize and control chat responses for your end user using attributes—or metadata—attached to documents that are mapped to index fields. For example, if `data source type` is an attribute that's attached to your documents, you can specify that chat responses should be generated only from a specific data source. Or, you can allow end users to restrict the scope of chat responses using the attribute filters that you have selected. For example, an end user can choose to have their chat responses generated using only documents from data sources that they specify. Filtering chat responses using metadata has the following key benefits: + **Q&A on LLM knowledge** – Users can ask questions and get answers from the general knowledge that the LLM has. + **Ensure response relevance and accuracy** – You can indicate that responses should only be generated from specific authoritative sources within your data. + **Control response context** – You can specify the file type (for example, PDF) and library or collection of documents (for example, business requirements documents) that responses are generated from. + **Maintain response freshness** – You can restrict chat responses to come from only documents that were generated after a specific date. + **Scope chat responses** – You can help your end users to narrow the scope of their responses and get to the right answer more quickly. Amazon Q Business offers a set of reserved document attributes that you can use. You can also create custom document attributes that better represent your organization’s data and use cases for more finely-tuned chat response control. **Important** Filtering using document attributes in chat is only supported using the API. Boosting search results using document attributes is supported using the console or the API. # Source attribution with citations in Amazon Q Business Source attribution with citations The Amazon Q Business web experience chat response provides in-text source citations for responses that use the organization's data sources and knowledge base as a source. The chat response also provides a full list of the sources used to generate the response. Amazon Q Business supports source attribution with citations. If you specify the `_source_uri` metadata field when you add metadata to your Amazon S3 bucket, the source attribution links returned by Amazon Q Business in the chat results will direct users to the configured URL. If you don't specify a `_source_uri`, users can still access the source documents through clickable citation links that will download the file at query time. This allows users to verify information even when no source URI is configured. To learn how to add metadata for your Amazon S3 connector, see [Adding document metadata in Amazon S3](https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/s3-metadata.html). **In-text source citations** In-text citations are provided in the form of a numbered list at the end of a sentence. To view an in-text source citation, choose a citation number. Each citation provides the following attributes: + **Title** – The title of the source document. + **URL** – A linked URL that you can follow to view the source document. + **Snippet** – The snippet from the document from the source document that was used to generate each sentence in the response. **Source list** Sources used to generate the response are provided at the end of the response. Each source listed provides the following attributes: + **Citation number** – The number provided at the end of the sentences in the response. + **Title** – The title of the document that's the source for the generated response. + **Text segment** – A text extract from a source document that's used for source attribution. + **URL** – The URL of the document that's the source for the generated response. # Upload files to a chat in Amazon Q Business Upload files and chat Users using the Amazon Q Business web experience can upload documents and use the uploaded documents to ask questions and summarize or analyze data based on the content of the uploaded documents. Amazon Q provides a list of recent documents, enabling users to find and reuse recently attached files in new conversations without re-uploading. Users can upload new files from their computer, select from recent files, or drag and drop files directly into any conversation. Users can upload up to 20 files during a conversation. The maximum upload size for each file is 50 MB and 3.75 MB for images. The total parsed content for all files combined has to be under 665,000 characters. Documents uploaded through the chat interface are deleted with the associated conversation after 30 days of inactivity. Once files are attached to a conversation, they cannot be deleted individually. Users must delete the entire conversation to remove attached files. Using the Amazon Q Business web experience, you can upload documents and use them to ask questions and summarize or analyze data based on their content. When you start a new conversation, you can upload new files, choose from a saved list of recent documents, or drag and drop files directly into the conversation. To use this feature, enable **Allow end users to send queries directly to the LLM** in **Admin controls and guardrails**. For more information, see the [Response settings](https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/guardrails-global-controls.html#guardrails-global-response) topic in [Admin controls and guardrails](https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/guardrails-global-controls.html#guardrails-global-response). Amazon Q Business supports specific document types for upload. To learn more about the document types that can be uploaded, see [Supported document formats in Amazon Q Business](doc-types.md). If you plan to upload comma separated value (CSV) or Microsoft Excel (XLS, XLSX) files, we recommend using tables that have no more than 4 columns and 10 rows for best results. **Important** Audio formats such as MP3, MP4, FLAC, WAV and Video formats such as MP4 and MOV are not supported for chat upload. # Personalizing Amazon Q Business chat responses Personalizing chat responses Amazon Q Business uses location and job-related information from your IAM Identity Center instance to generate personalized responses that are relevant to your end user. For example, if your end user asks "What are the company holidays for this year?", Amazon Q Business might list region-specific holidays based on their country. To enable response personalization, add **Address** information and **Job related information** for users in the IAM Identity Center instance that connects to your Amazon Q Business application. For more information, see [Add users](https://docs.aws.amazon.com/singlesignon/latest/userguide/addusers.html) in the IAM Identity Center User Guide. User personalization data that's in your IAM Identity Center instance is connected to your Amazon Q Business application environment, and responses are personalized by default. You can deactivate response personalization at any time by using the [Admin controls and guardrails](https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/guardrails-global-controls.html#guardrails-global-controls-customizing) feature in Amazon Q Business. # Quick prompts in Amazon Q Business Quick prompts The Amazon Q Business web experience welcome page provides sample prompts to help your end users understand the types of questions and tasks that they can ask. Sample prompts are not enabled by default. If you're an AWS Management Console customer who needs to configure the web experience for your end users, you can enable the sample prompts feature when you preview the web experience. For more information, see [Customizing a web experience (IAM Identity Center)](https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/customizing-web-experience.html) or [Customizing a web experience (IAM)](https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/customizing-web-experience.html). **Important** Before you enable sample prompts, make sure that the **Only produce responses from retrieval augmented generation (RAG)** check box for **Application guardrails** is not selected. For more information, see [Customizing global controls](guardrails-global-controls.md#guardrails-global-controls-customizing). You can't create your own prompts or edit the provided sample prompts. # Hallucination mitigation in Amazon Q Business Hallucination mitigation A [hallucination](https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/concepts-terms.html#hallucination), in the machine learning context, is a confident response by an AI (artificial intelligence) application environment that isn't supported by its underlying data. Amazon Q Business includes a hallucination prevention system that works in real-time during chat conversations. The *hallucination mitigation* feature helps ensure more accurate retrieval augmented generation (RAG) responses from data connected to the application—either through connected data sources, or files uploaded during chat (up to 100,000 characters). During chat, Amazon Q Business evaluates a response for hallucinations. If a hallucination is detected with high confidence, it corrects the inconsistencies in its response real-time during chat and generates a new, edited message. To activate hallucination mitigation, use [Amazon Q Business admin controls and guardrails](https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/guardrails-global-controls.html). Hallucination mitigation isn't supported for the following use cases: + Applications where [chat orchestration](https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/guardrails-global-controls.html#guardrails-global-orchestration) is enabled. + [Plugin](https://docs.aws.amazon.com/amazonq/latest/qbusiness-ug/plugins.html) workflows. + Responses generated from tabular data, or from transcripts of images, audio and video. Hallucination mitigation applies only to responses generated from textual data. **Important** Activating hallucination mitigation will increase chat response latency. # Agentic Retrieval Augmented Generation (RAG) in Amazon Q Business Agentic RAG Agentic RAG enhances the standard RAG workflow of Amazon Q Business with agentic retrieval and response capabilities. Unlike standard RAG's document retrieval and simple response generation process, Agentic RAG uses multiple intelligent agents and specialized data retrieval tools to deliver more comprehensive and accurate responses while maintaining conversation context. With Agentic RAG system processes queries through a combination of the following coordinated steps: + Analyzes both the user's question and conversation history and determines which retrieval tools to use + Intelligently deconstructs complex queries into simpler ones + Intelligently triggers multiple data retrieval tools as needed + Provides disambiguating questions based on enterprise data to clarify user intent + Synthesizes information from various sources, and generates responses with its underlying large language model + Provides follow-up questions to intelligently continue the conversation Throughout this process, it continuously checks response quality and activates additional data retrieval tools when necessary, showing users real-time progress to the user as it processes queries. All responses maintain existing permissions and include clear citations to source material. Agentic RAG delivers several key improvements over standard RAG. It intelligently selects from available retrieval tools based on query requirements and performs multiple retrieval operations for complex queries. The system maintains conversation context awareness and adapts response generation through retries or disambiguation techniques based on the quality of retrieved information and the subsequent response generated. These capabilities result in higher accuracy, more comprehensive information gathering, and better handling of complex, multi-faceted queries. To use Agentic RAG, enable the feature using the **Advanced search** toggle in your web experience interface. **Note** While the Agentic RAG system provides more thorough responses, response times may be longer than standard RAG due to its multiple retrieval operations.