Beyond Text: Introducing Artifacts for Vectara Agents

Vectara Agents have evolved from text-only assistants to multimodal, file-aware systems capable of handling various types of content through the introduction of artifacts, which are persistent pieces of content stored within an agent session. These artifacts enable users to upload and manage files like PDFs, spreadsheets, and images, allowing agents to analyze, convert, and create new content such as summaries and reports that remain accessible throughout the session. Artifacts are session-specific and have size and storage limitations, as well as configurable time-to-live settings, ensuring they are automatically deleted after a certain period or when a session expires. Critical tool configurations are necessary for agents to interact with these artifacts, which include reading text files, viewing images, converting documents, and creating new content. This development facilitates complex, file-heavy AI workflows, empowering users to conduct financial analyses, perform visual debugging, and convert and summarize documents, thereby unlocking potential applications in contract analysis, chart interpretation, and report automation.