Admins can add and manage custom LLM and embedding endpoints within QAnswer. This includes defining provider type, modality (text, image), sensitivity level, and connection parameters (endpoint URL, API key). This feature is restricted to organization admins to ensure secure configuration and governance.

QAnswer now integrates with Confluence to enable AI-driven retrieval across pages, spaces, and project documentation. The assistant processes Confluence content at scale, applies semantic search to identify the most relevant fragments, and delivers contextual answers directly in chat. With support for multilingual queries and heterogeneous datasets, this integration reduces time spent navigating Confluence manually and ensures consistent, transparent access to organizational knowledge.

General Chat can now automatically route questions to the right assistant. Each query is checked against all available assistants, and if one is a good match, the system forwards the query to it. The interface clearly shows which assistant is being used, so users can see how their request is handled. This makes it possible to work with multiple assistants in one place without manual switching.

We’ve made a series of refinements across QAnswer to improve stability, usability, and overall performance. From smarter chat interactions to enhanced search, new connectors, and UI polish, these updates make everyday use faster and more seamless. Click below to explore the full list of improvements.

QAnswer now offers more granular source-level traceability for answers pulled from large documents. It highlights precise sentence fragments, anchors them with numbered references, and scrolls directly to their location in the file. Whether working with contracts, specs, policy docs, or multi-section reports, this update reduces the overhead of manual verification and improves transparency across complex datasets.

Report generation has been a core use case for QAnswer, particularly in workflows where summaries or structured documents follow predefined formats. This release introduces three improvements to streamline that process. Users can now create reusable templates to standardize output across recurring tasks. External formats—such as Word, Markdown, or HTML—can be imported and applied directly. The system also enforces better control over structure and formatting, ensuring that generated content aligns more closely with the expected layout and reduces the need for manual adjustments.

Acronym definitions (e.g., CPNP: Cosmetic Products Notification Portal) were previously only used during retrieval, helping the system find relevant sources but not improving how the LLM interpreted the query itself. With this release, acronym data is now also injected into the LLM prompt. This allows the model to understand and respond more accurately to queries that reference acronyms—even when the full term isn’t present in the user’s input. No changes are required in how acronym data is added; the enhancement improves coverage and consistency across both search and generation.