What happened
DigUp 0.7.0-beta.3 is a pre-release for Windows and Linux, published on 22 October. The release carries fixes that had gone into the Mac 0.6.2 version over to those platforms, alongside several beta-specific changes. The Mac application is not included in this release: it remains on 0.6.2, with all of these fixes except the camera-photo change expected in a future 0.6.3 release.
The most visible new capability is support for office-style files. DigUp now reads text from XLSX, PPTX, ODS, and ODP files. For spreadsheets, it indexes sheet names and rows; for presentations, it indexes slide text and speaker notes. It also indexes Typst files with the .typ extension as documents. That makes a local search corpus less dependent on plain text, PDFs, images, or source files: a user can include planning sheets, decks, and notes in one index, subject to DigUp’s folder and file-selection rules.
Several changes make the indexing boundary more visible. Settings now shows a count of files that are not read for each folder, plus a “why” explanation covering reasons such as code projects, folders that resemble datasets, and unsupported types. A folder explicitly selected by the user is now read in full, even if it contains material that formerly caused DigUp to classify it as a code project. The release also says setup explains the folder step before it runs, labels its action “Start indexing,” and does not inspect folders until the initial setup page has been left. While folders are being walked before indexing, the interface shows “Checking folders…”. Removing the final folder from Settings now removes that folder’s files from results.
The Linux build changes idle-resource behavior: its search process exits after ten minutes without activity rather than remaining resident. The release reports that this reduces idle memory from 145 MiB to zero; the next search restarts the process and therefore takes slightly longer. It also fixes repeated tray-icon redraws and GTK warnings on desktops without a system tray. Other fixes include a progress bar that no longer restarts when files arrive during an indexing pass, removal of stale results after a folder is removed, skipping copy-protected M4P files rather than failing on each one, and a crash path involving a zero-sized image. Keyword search on a 29,000-vector index was measured by contributors as 18–20% faster after the database was memory-mapped.
Windows users can download a 64-bit Windows 10 or 11 installer that installs without administrator rights. The installer is not signed, and Windows may require the user to choose “More info” and then “Run anyway.” Linux users on Ubuntu 22.04, Debian 12, Mint 21, or newer can use a 64-bit DEB package; other supported distributions can use an AppImage. Existing beta.1 and beta.2 Windows and AppImage installations update themselves, while DEB users install the new package over the old one.
What the archive adds
This is a new fact on a known story. Earlier coverage described DigUp as a free Mac menu-bar application for semantic search across selected files. The beta changes that platform boundary by offering Windows and Linux builds, and it expands the searchable material to spreadsheets and presentations.
Trend signal
No supplied site pattern is a direct fit for this local file-search release.
Why it matters
Teams building with AI often have important working context distributed across decks, spreadsheets, project notes, screenshots, and repositories. A local search tool becomes more useful when it can retrieve text from the formats where planning, reporting, and decision records actually live. The release also makes exclusion rules inspectable instead of silently omitting material, which matters when a retrieved answer depends on whether a folder was indexed at all.
The operational trade-off is clear on Linux: DigUp gives up an always-resident search process to eliminate its stated 145 MiB of idle memory use. That is a reasonable default for an intermittent desktop tool, but users who expect instant search should account for the restart delay after ten idle minutes. The practical relevance is not a new agent interface or shared company-memory system; it is a broader local retrieval layer that may make existing files easier to find before they are manually supplied to an AI tool.
Limits and uncertainty
The release notes describe supported formats and fixes, but they do not establish retrieval quality, indexing speed for spreadsheets or decks, or how reliably complex formulas, embedded objects, charts, scanned content, and unusual presentation layouts are represented. The 18–20% keyword-search improvement is tied to a 29,000-vector index and contributor measurements, not a general benchmark across hardware or corpora. The source does not state the application’s privacy model, model behavior, or whether indexed content leaves a device. It is also explicitly a beta release. Windows installation has an additional trust concern because the installer is not yet signed.
Skill to build now
Audit one working folder before relying on local retrieval. Include a small representative set of a spreadsheet, a presentation, a note, and an image; then search for terms known to appear in each. Check the folder-level “not read” explanation and remove or relocate files that should not enter the index. On Linux, also test a search after more than ten minutes idle so that the restart delay is part of the team’s actual workflow expectation rather than a surprise.
Reader takeaway
DigUp’s beta is no longer confined to Mac and can now search the text inside common spreadsheets and presentations on Windows and Linux. Its value is practical local retrieval across more of a person’s working files, not a demonstrated replacement for a governed shared knowledge system.
Open question
How well do its extracted spreadsheet rows, slide text, and speaker notes rank against ordinary document results in real project folders?