documentation
03 dashboard

Files

Every file that flows through Exolvra — agent screenshots, generated PDFs, uploaded attachments, images from tools — lives in the file store. This page is where you search, preview, and download them.

What this page is for

The Files page is the unified view of every file Exolvra is holding on to. That includes:

  • Agent-generated files — screenshots, generated PDFs, tool outputs, saved artifacts
  • User uploads — attachments you added to chat messages, issues, or comments
  • Import paths — files imported via the upload_path flow from agent workspaces
  • Binary artifacts — images, audio, anything that isn’t text

It doesn’t show files stored inside the data store (those are structured text entries) or inside document collections (those are managed separately). The Files page is for standalone binary and text files that don’t belong in either of those.

Layout

The page is a list or grid view, switchable at the top. Each file shows:

  • Thumbnail or icon — images render a preview, other files show a format icon
  • Filename
  • Size
  • Kind — image, PDF, text, audio, video, other
  • Source — which agent, user, or session produced it
  • Project — if scoped to one
  • Uploaded / created — timestamp

A filter bar lets you scope by kind, source, project, and time range. A search box runs across filenames and (for text files) content.

Common tasks

Preview a file

Click any card. For common formats the preview opens inline:

  • Images — full-size viewer with zoom
  • PDFs — embedded PDF viewer
  • Text — syntax-highlighted viewer
  • Audio — inline audio player
  • Video — inline video player
  • Other — generic download prompt

Click outside the preview to close, or press Esc.

Download a file

From the preview or from the card, click Download. The file is served from the gateway and saved to your browser’s default download location.

Upload a file

Click Upload in the page header, pick a file from your disk. You can optionally:

  • Tag it with a project — so it’s scoped and searchable by project
  • Add a description — helpful if you’ll search later
  • Choose public within the instance or private to you

Uploaded files immediately appear in the list and can be referenced by agents or users.

Delete a file

From the file detail view, click Delete. Confirm. The file is removed from storage permanently. If any messages, data entries, or issue attachments referenced it, those references break — you’ll see a “file not found” placeholder instead.

Find files a specific agent produced

Filter the source dropdown to a specific agent. The list narrows to just files that agent generated through tool calls. Useful when you’re reviewing what an agent built.

Storage backend

By default Exolvra stores files on the local disk under the data directory. For production deployments you can configure an external backend:

  • S3-compatible — AWS S3, MinIO, DigitalOcean Spaces
  • Azure Blob Storage

Switching backends is done in Config → Storage. Existing files stay where they were; new files land in the new backend. You can migrate in bulk from the backend settings page if you want full consistency.

All backends share the same content-addressable deduplication: uploading the same file twice stores it once. That matters when agents save screenshots — repeated screenshots of the same page don’t take double space.

Common pitfalls

Treating the files page as a long-term filesystem. Files here are associated with work, not with your user account. If you upload something and then delete the associated project, the files go too. For long-term personal storage, keep files somewhere outside Exolvra.

Granting file_store access to every agent. Most agents don’t need to write binary files. Granting it broadly leads to clutter in the files list. Only specialists that actually produce artifacts (image generators, PDF writers, screenshot takers) need it.

Forgetting about cleanup. The files list grows forever by default. Periodically review and delete stale artifacts, or set retention policies on individual projects.

Where to go next

  • Data browser — the structured-text counterpart
  • Documents — for knowledge collections that feed chatbots
  • Projects — to see files scoped to a specific initiative