Global search

Global search finds content in every document you may open in Documents: PDFs and Office files, images, videos and audio recordings – across all contacts. In recordings it also searches what is said and, when you open a hit, jumps straight to that moment.

Open the search

• Keyboard shortcut: Cmd + K (Mac) or Ctrl + K (Windows, Linux) – anywhere in Clapilot.
• Magnifier: the magnifier icon at the top left next to the menu.
• iPhone, iPad and Mac: the document search icon at the top of the side menu; on the Mac also Cmd + K.

Type at least two characters. Results appear while you type.

Search and filter

• Words, names and numbers: file references, invoice numbers and names are found as you type them, for example 12 O 345/24 or RE-2026-0042. Word forms are recognized: Rechnungen also finds Rechnung.
• By meaning: when an embedding model is set up in Clapilot, the search also finds hits phrased differently, such as spare parts warehouse for an "inventory of spare parts". Without an embedding model it searches the text only; a note below the search field says so.
• Type: All, Documents, Images, Videos and Audio narrow the results.
• Keyboard: ↑ ↓ select a hit, Enter opens it, Esc closes the search.

The results

Results are grouped by type. Each hit shows the title, an excerpt with the search words highlighted, the contact and the date.

• Images show a thumbnail.
• Videos show a still and – when the hit is in what is said or in a frame of the video – the timestamp, for example 1:30.
• Audio shows the matching part of the transcript and the timestamp.

A click opens the document in the preview. Videos and recordings start at the hit's timestamp.

Recordings and videos

Clapilot transcribes audio files (MP3, WAV, M4A, AAC, OGG, FLAC) and the sound of videos (MP4, MOV, M4V, WebM, MKV, AVI) when they are uploaded, in 30-second sections. For videos it also describes a few stills. Transcription uses the speech recognition model your administrator set up under Settings → ClapilotAICore → Audio. Long files are transcribed up to a fixed length (default: the first 30 minutes). Without speech recognition the search finds a recording only by its title and details.

Who finds what

• You find all shared documents of your workspace and your own private documents – exactly what you also see in Documents. Other people's private documents never appear.
• The search is just for you: your results are not passed to the agent. In chat, the agent keeps working only with the documents of the contact currently selected.