Details Panel

The details panel is a sidebar that opens when you select a file or folder. It gives you a full picture of the item — its core properties, optional annotations, media-specific metadata, and the status of every intelligence pipeline stage that makes the file searchable and AI-readable.

Opening and closing the panel

Open the details panel by pressing ⌘i / Ctrl+i, or by clicking the Tools pill in the bottom-right corner and choosing Details. On narrow windows the panel may be hidden by default.

Click the same shortcut or pill button again to close it, or switch to a different panel view.

Preview area

The top of the panel shows a visual preview of the selection.

• Single file — a large thumbnail fills the preview area
• 2–4 files selected — a 2×2 grid of thumbnails
• 5–9 files selected — a 3×3 grid of thumbnails
• 10 or more files selected — a 4×4 grid showing up to 16 thumbnails

Adding and removing optional properties

Below the preview, the Add / Remove button opens a menu listing every optional property that can be attached to a file. A checkmark appears next to any property that already exists on the file. Click an item to toggle it on or off.

Optional properties you can add or remove:

Removing a property deletes its stored value from the file. There is no undo — if you remove Notes, the text is gone. Re-adding the property creates a blank field.

Properties

The Properties section shows fixed, system-managed file attributes:

Tags

Tags are short, freeform keyword labels you attach to a file. They serve two purposes:

Search — Tags are indexed and fully searchable. You can filter search results to files that have a specific tag using faceted search, or include a tag name in a text query to surface matching files.

Agent context — When the AI agent reads or works with a file, it sees the file's tags. This makes tags useful for annotating files in ways that steer how the agent reasons about them — for example, tagging a contract as client-a and draft gives the agent immediate categorical context without it needing to read the full document.

Click Add Tags to create the first tag on a file. Once tags exist, click directly on the tag area to edit. Tags are stored as comma-separated values, and you can add as many as you like.

Notes

Notes are a freeform, multi-line text field for anything you want to record about a file. Like tags, they serve two purposes:

Search — Notes are indexed and included in full-text search. If you write detailed notes on a file, searching for phrases from those notes will surface the file even if the content itself doesn't contain those words.

Agent context — The AI agent reads a file's Notes when it accesses the file. This is a direct way to leave instructions, summaries, or background information that the agent will pick up and use when answering questions or performing tasks involving that file. For example, you could add a note to a folder of vendor contracts saying "These are active contracts — do not modify" and the agent will respect that context.

Click Add Notes to open the notes editor. The editor supports multi-line input. Click outside or press Escape to save and close.

Rating

A 1–5 star rating you can assign to any file for personal sorting or prioritization. Ratings are stored as a file property and can be used as a filter in faceted search. They are not currently used by the AI agent.

Media-specific metadata

For recognized file types, the panel shows an additional metadata section with format-specific attributes extracted automatically:

• Images — pixel dimensions, color space, camera model, shutter speed, aperture, ISO, and GPS coordinates (if present in EXIF data)
• PDFs — page count, embedded author, and document title
• Audio — duration, bitrate, sample rate, artist, album, and track title from ID3 tags
• Video — duration, resolution, frame rate, and codec
• Links — archived page title and source domain

No configuration is needed — these sections appear when Poly detects the file's type.

Generation metadata

If a file was produced by an AI image generator (such as Stable Diffusion or Midjourney), Poly can store and display the original generation parameters under a Generation Metadata section. These fields are the same ones listed under Adding and removing optional properties above. They are added automatically if the generator embeds them in the file, or you can add them manually using the Add / Remove button to annotate any AI-generated file.

Intelligence Features

Every file in Poly goes through a processing pipeline that extracts information used by search, similarity matching, and the AI agent. The Intelligence Features section shows the live status of each pipeline stage.

The section only appears when the selected item is not a top-level folder, and — for multi-file selections — only when all selected files share the same pipeline status.

Status indicators

Metadata

Applies to: all file types

Metadata shows whether Poly has successfully extracted and indexed the file's core properties — its name, type, size, dates, tags, notes, and other structured attributes. This stage runs for every file.

A green checkmark means the file is findable by name and by property-based filters. If Metadata fails (red ✗), the file will not appear in any search results at all, regardless of the status of other stages. Refresh the page to prompt a re-check.

Content

Applies to: all file types once a supported format is detected

Content shows whether Poly successfully loaded the file's raw bytes from storage. This is a prerequisite for every downstream stage — Poly cannot extract text, generate embeddings, or transcode media until the bytes are accessible.

A red ✗ usually points to an upload interruption or storage issue. Re-uploading the file is the most reliable fix.

Similarity

Applies to: images, documents, PDFs, and other supported types

Similarity shows whether Poly has generated a vector embedding for the file. Embeddings power:

• Semantic search — finding files by meaning rather than exact keywords
• Visual similarity search — finding images that look like a reference image
• Related-file suggestions across your library

An amber — means the file type does not support embeddings — the file still appears in keyword and metadata searches, just not in semantic or visual results. A red ✗ means generation failed.

Transcoding

Applies to: videos, PDFs, documents, ebooks, spreadsheets, presentations, and links

Transcoding shows whether Poly has converted the file into an optimized intermediate form used for previews and downstream extraction.

• Video — transcoded to a streaming format so playback and audio extraction work
• Documents and PDFs — pages rendered into a normalized form for preview and text extraction
• Ebooks — converted to a consistent format for parsing
• Links — the linked page is fetched and archived

A red ✗ typically means an unsupported codec, an unexpected format variant, or a corrupted file.

Structure

Applies to: videos, audio, PDFs, documents, ebooks, spreadsheets, presentations, and links

Structure shows whether Poly has extracted the file's content into an indexed, readable form. This is what enables full-text search within a file and gives the AI agent something to read.

• Documents and PDFs — extracted text, headings, and paragraphs
• Spreadsheets — cell data and sheet layout
• Presentations — slide text and speaker notes
• Audio and video — a transcript from the audio track
• Ebooks — chapter text
• Links — parsed page content

A red ✗ means extraction failed; the file may still appear in search by name and metadata, but its content is not searchable and the agent cannot read it.

Hidden

The Hidden toggle (shown when you have permission to edit the file) excludes a file from all indexing and search.

When Hidden is on:

• All pipeline stages are skipped and every row shows a gray —
• The file does not appear in any search results — text, semantic, visual, or color search
• The AI agent cannot read or reference the file

Turning Hidden on for a folder cascades to every file and sub-folder inside it. Each descendant inherits the setting and displays gray — indicators. To un-hide a specific file inside a hidden folder, open its details panel and toggle Hidden off — this sets an explicit override that takes precedence over the parent.

If a file stops appearing in search results, check whether it or any of its parent folders has Hidden turned on.

Search impact at a glance

Multi-selection

When more than one file is selected:

• The Selection section shows how many items are selected
• The Properties section is hidden (each file has its own values)
• The Details section (tags, notes, rating) is hidden
• Intelligence Features appears only when all selected files share the same pipeline status

Frequently asked questions

A stage shows a red ✗. What should I do?

Start by refreshing the page — Poly verifies file integrity on load and re-queues failed stages automatically. On the desktop app, try opening the file's containing folder in the web app to trigger a fresh sync. Wait a few minutes and check again.

If the error persists:

• The file's media type may be mislabeled. Re-upload it with the correct file extension.
• The file may be corrupted. Open it in its native application to verify.
• The format may not be supported. Check supported file types.
• The format version may be too new. Ask the AI agent why the file might not be processing — it can often diagnose the issue.
• Reach out via Help or the Discord community if nothing resolves it.

Why doesn't a stage appear for my file?

Not every stage applies to every type:

• Transcoding only appears for video, PDF, document, ebook, spreadsheet, presentation, and link files
• Structure only appears for video, audio, PDF, document, ebook, spreadsheet, presentation, and link files
• Similarity may show an amber — for file types that don't support embedding generation

If the entire Intelligence Features section is missing, the item is likely a top-level folder, or the multi-file selection contains files with different processing states.

My file doesn't appear in search. What should I check?

1. Check the Hidden toggle — look under Intelligence Features. Also check any parent folders.
2. Check Metadata — a red ✗ here means the file won't appear in search at all.
3. Check Content — a red ✗ blocks all downstream processing.
4. Wait a few minutes — newly uploaded files take time to process, especially large ones.

How long does processing take?

Most files complete all stages within a minute or two. Large videos, long audio files, and multi-page documents may take several minutes. If a spinner has not resolved after ten minutes, try refreshing the page.

Can I re-trigger processing?

There is no manual re-process button. Refreshing the page or re-opening the containing folder typically re-queues failed stages. For persistent failures, re-uploading the file is the most reliable solution.