Hava Docs

Attachments & multimodal chat

Hava chat accepts images, documents, structured data, source files, and project ZIPs directly in a conversation. Use the + button, drag files onto the composer, or paste an image from the clipboard.

How Hava prepares attachments

Attachments are uploaded to the signed-in user's file space and prepared according to type. Images use vision extraction when configured; text and code can be indexed; PDFs and documents use document extraction; ZIP projects can be unpacked and indexed for selective retrieval. Hava sends compact attachment context into the conversation rather than blindly placing entire large files in a model prompt.

Conversation scope

An attachment is user-provided conversation context. Uploading something in chat does not turn it into durable Memory or Business Knowledge. Business Knowledge remains an owner-approved surface.

Privacy and safety

Hava only exposes attachment metadata and prepared content to the runtime needed to answer the user's request. File access remains subject to authentication and workspace permissions. Avoid placing secrets in files unless the connected environment is intended to handle them.

Supported input

Images include PNG, JPEG, WebP, GIF, BMP and SVG. Documents and data include PDF, DOC/DOCX, TXT, Markdown, JSON, CSV, XML and YAML. Common web, mobile and backend source files are accepted, along with ZIP projects.