Back to Blog

Universal Search on Windows: Index PDFs, Emails, and Archives Locally

Build a local full-text index for PDFs, Outlook messages, archives, and OCR text: boolean queries and privacy without cloud upload.

  • search
  • privacy
  • windows

Everything by Voidtools finds filenames instantly. It does not read inside PDFs, .msg files, or zip archives. Recoll and DocFetcher index content but bring their own setup curves and older UIs.

DupeZappa Universal Search keeps a local SQLite index on your machine: filenames, extracted document text, email bodies, optional archive contents, OCR from images, and metadata from smart tags and face recognition. All searchable without sending files to the cloud.

What gets indexed by default

Category Formats (examples)
Documents .pdf, .docx, .xlsx, .pptx, .odt, .epub
Email .eml, .msg, .mbox, .mht
Structured JSON, CSV, XML, HTML-style tokens
Video Sidecar subtitles (.srt, .vtt) + metadata

Optional paths (when local tools are installed):

  • Archives: .zip, .7z, .rar, .iso via 7-Zip (one level deep).
  • CHM help files: deep text extraction via 7-Zip.
  • Legacy Office: .doc, .xls, .ppt via LibreOffice soffice.
  • OCR: image text and scanned PDF fallback on Windows.

Scan exclusions from Settings apply to indexing jobs the same way they apply to duplicate scans.

Add roots and refresh

  1. Open Universal Search in DupeZappa.
  2. Add one or more folder roots, recursive or flat per root.
  3. Run index. Large libraries are background jobs; you can filter and remove roots later.
  4. Re-index after big imports (new email archive, document dump).

The index lives in your app config folder as search_index.db. It is yours, not a vendor cloud index.

Query power beyond plain text

Boolean operators work in search:

  • invoice AND pdf: both terms required.
  • budget OR forecast: either term.
  • draft NOT template: exclude noise.

Field and source filters narrow scope:

Filter Example
source:email Messages only
source:pdf PDF text layer
name:invoice Filename token
ext:docx Extension
size:>=1mb Large files
date:2025-01-01..2025-06-30 Date range
tag:"Project Alpha" Smart Tag membership
source:tag All tagged files

Preset query suggestions appear based on what is actually in your index; email presets only show when email sources exist.

How this differs from other tools

Full comparison: DupeZappa vs Everything (Voidtools) →

Tool Strength Gap
Everything Fastest filename search No content index
Recoll / DocFetcher Deep document search Separate app; setup overhead
Agent Ransack Content search without index Slower on huge trees
DupeZappa Content, tags, faces, and duplicates together Windows only; index build takes time

If you already use Everything for launchers, keep it. Universal Search complements it when you need to find the PDF that mentions the clause on page 12.

Tie search to cleanup

Search finds candidates; other modules act on them:

  • Duplicate Scanner on a dense folder from size:>=10mb results.
  • Smart Tags to build virtual collections from search result sets.
  • Face Recognition people synced to tags; search tag:"Sarah" after naming a person.

See Smart Tags Without Moving Files for the tagging bridge.

Privacy default

Indexing and search run locally. Optional AI features (AI Tidy, semantic search) use Ollama or your own endpoint, but are not required for Universal Search v1 text indexing.

Start with one document root and one email archive. Expand when query latency and index size feel predictable.

Related reading