sgcFSearch

1Introduction

sgcFSearch is a whole-disk file search tool for Windows. It keeps an index of the names of every file and folder on your fixed drives, answers queries in milliseconds while you type, and keeps the index current while files change. A search bar opens with a global hotkey, the sgcfs command line client gives the same results in a terminal, and a local named pipe API lets other programs ask questions too.

Besides file names, sgcFSearch can search the text inside files located in your content folders (your user folder by default), including Office documents and, through Windows filters, PDF files.

sgcFSearch is modelled on scry and inspired by fsearch. It is an independent implementation written in Delphi.

Main features

• Name index of every fixed drive, built by a multi-threaded crawler on the NT directory API and saved to names.idx for fast restarts.
• Live updates while the app runs, plus a catch-up through the NTFS change journal (USN journal) for changes made while it was closed. No administrator rights are needed.
• Search bar on a global hotkey with ranked results, file icons, highlighted matches, filter chips, search history, open, show in folder and copy.
• Typo tolerant matching and a rich query syntax: name modifiers, path tokens, extension, type, kind, folder scope, path text, size, modified date, regular expression, limit and sort filters.
• Ranking that learns from the files you open or copy (frecency).
• Content index of the files under your content folders, searched with grep:, regex: and sym:.
• sgcfs command line client with plain, long and JSON output, a JSON lines bridge, status, stop, show, install, uninstall and bench commands.
• Named pipe API (one JSON object per line) restricted to the current user.
• Tray icon with the index status, a Folders dialog, Start at login and the daemon log.
• Two downloads, both digitally signed: a per user installer (recommended) and a zip for a manual installation. No administrator rights are needed.

System requirements

The parts

2Installation

Downloads

sgcFSearch has two downloads. Both are per user and need no administrator rights. The installer and the executables inside the zip are digitally signed by eSeGeCe.

Use one of them, not both. See Uninstall.

Install with the setup

Download and run sgcFSearch_setup.exe. The setup is per user: it does not ask for administrator rights and installs only for the current account.

Both tasks are on the "Additional options" page of the wizard. When you install over an existing version, the setup first asks the running app to stop (sgcfs stop), waits until it has saved its index, and then replaces the files, so you can run the setup again to update.

Manual installation from the zip

Download sgcFSearch.zip and unzip it to any folder you own, for example C:\Users\you\Tools\sgcFSearch. Then run sgcFSearch.exe. Nothing is written to the system until you ask for it.

Keep sgcFSearch.exe and sgcfs.exe in the same folder. sgcfs starts sgcFSearch.exe from its own folder when the app is not running. Settings and index files go to the data folder (%LOCALAPPDATA%\sgcFSearch, or the folder in SGCFSEARCH_HOME, see Environment variables), not to the program folder.

To update, quit the app from the tray menu, then unzip the new files over the old ones.

Command line and start at login

With the manual installation, two optional commands of sgcfs set this up for the current user. The setup does the same through its two tasks.

You can also switch Start at login on or off from the tray menu item Start at login.

Uninstall

Installed with the setup. Uninstall sgcFSearch from the Windows Settings app, under Apps. The uninstaller:

• stops the running app with sgcfs stop,
• removes the program files and the Start menu shortcut,
• removes the install folder from the user PATH,
• removes the sgcFSearch login entry from the Run key, also when the tray menu created it.

Installed manually from the zip.

1. Run sgcfs uninstall. It removes the login entry and removes the folder of sgcfs.exe from the user PATH.
2. Quit sgcFSearch from the tray menu.
3. Delete the folder you unzipped.

In both cases the data folder %LOCALAPPDATA%\sgcFSearch (index, settings, history, usage, log) is kept. Optionally delete it by hand to remove every trace.

3Getting started

First run

Start sgcFSearch from the Start menu entry created by the setup, or run sgcFSearch.exe from the folder where you unzipped it. The sgcFSearch icon appears in the notification area of the taskbar. When you start the app yourself, without --background, the search bar also opens right away.

On the first run there is no index yet, so sgcFSearch crawls every fixed drive. While it crawls:

• the tray tooltip reads sgcFSearch: Indexing N entries,
• the status line of the search bar reads Indexing... N files,
• you can already search. Results come from the part of the disk crawled so far and refresh as the index grows.

When the crawl ends, the index is saved to names.idx and the state becomes Ready. Then the content indexer starts and reads the files of your content folders in the background, at low priority. While it works, the tray tooltip adds N files being read.

On later starts, sgcFSearch loads names.idx, catches up with the changes made while it was closed, and is ready in about a second.

Index states

Searches work in every state.

The hotkey

Press Alt+Shift+Space to show the search bar. Press it again, or Esc, to hide it. If another program already owns that hotkey, sgcFSearch tries Win+Shift+F and then Ctrl+Alt+F. The hotkey in use is shown next to Search in the tray menu and written to the log. You can choose another hotkey in the Folders dialog.

A left click on the tray icon, the tray menu item Search and sgcfs show also open the bar.

A first search

1. Press Alt+Shift+Space.
2. Type part of a file name, for example invoice. Results appear as you type, best matches first.
3. Narrow it down: invoice ext:pdf mtime:month shows PDF files with "invoice" in the name modified in the last 30 days.
4. Use Up and Down to select a result and press Enter to open it, or Ctrl+Enter to show it in its folder.

The same search from a terminal:

sgcfs invoice ext:pdf mtime:month

4The search bar

Layout

The bar opens centered on the monitor that holds the mouse pointer, in the upper part of the screen, and stays on top of other windows. It grows with the number of results and hides itself when it loses the focus (for example when you click another window). It follows the Windows light or dark app mode.

Result rows

The bar shows results as two-line rows, not as columns:

The status line shows, from left to right:

• No file name matches, showing matches inside files, when the content fallback produced the results,
• the number of results and the search time, for example 1,234 results 12.5 ms. For name searches this is the total number of matches, even when only the first 300 are loaded,
• a warning about the query, for example invalid size: 10q, or an error in red, for example for an invalid regular expression,
• the index state while it is not ready, for example Indexing... 1.2M files.

Keyboard and mouse

Open, Show in folder and Copy file feed the frecency ranking. Copy path does not.

Filter chips

The chips are a mouse friendly way to write filters. Each chip reads and writes one filter in the search box, so a filter you type by hand also lights up its chip. Choosing an option replaces the current value of that filter. Remove filter, at the bottom of the menu of an active chip, deletes it.

An active chip shows its value: the Type chip lists the categories (for example Images, Code), the In folder chip shows In and the folder name, and the Sort chip shows Sort: Name.

History

The bar remembers your last 50 searches in history.json in the data folder. A search is added when the bar closes with text in the box, and when you copy a file or a path. Repeating a search moves it to the top instead of adding a duplicate.

With an empty search box, press Up to recall the most recent search, and keep pressing Up for older ones. Down goes back toward the newest one and finally to an empty box. As soon as you edit the recalled text, the arrow keys move the selection again.

Folder picker

Two ways lead to the standard Windows folder dialog:

• the In folder... chip, menu item Choose folder..., which sets the in: filter,
• typing in: or path: and pressing Tab while the cursor is on that token. The chosen folder is inserted after the colon, in double quotes when it contains spaces.

5Query syntax

A query is a list of words and filters separated by spaces. Every word must match the name of the file or folder (not its path), case-insensitive. Filters have the form key:value. Only the keys listed in the filter reference are filters: any other text:text token is searched as a plain word. The order of words and filters does not matter.

Words and modifiers

Matching is on the name only. To match folder names along the path, use path tokens, in: or path:.

Typo tolerance

A plain word of 5 or more letters also matches a name that contains it with one error: one letter missing, one letter extra, one letter different, or two neighbour letters swapped. confgi finds config.json, raport finds report.pdf. Typo matches get the lowest match tier, so they normally rank below exact matches. Words shorter than 5 letters, and words with ', ^ or $, must match exactly.

Path tokens

A word that contains \ or / is a path token. Its last part matches the name, and the earlier parts must match the names of ancestor folders, in that order, as substrings.

Filter reference

Filter keys are case-insensitive. Put the value in double quotes when it contains spaces, for example in:"C:\My Projects".

When a filter is given twice, ext: and type: add to the list, in: widens the scope, path:, size: and mtime: must all hold, and for kind:, re:, limit:, sort:, grep:, regex: and sym: the last one wins. If a query has more than one content filter, grep: is used first, then regex:, then sym:.

An invalid value never stops the search. It is ignored and reported in the status line, for example invalid size: 10q, unknown type: pics, invalid mtime: lastweek or empty value: ext:.

The home folder ~

In in: values, and in the content folders and skip list of the settings, a leading ~ stands for your user folder (%USERPROFILE%, for example C:\Users\you). Use ~ alone or followed by a backslash: in:~, in:~\Documents.

Size values

Units are b (default), k, kb, kib, m, mb, mib, g, gb, gib, t, tb, tib, all 1024 based. Decimals are allowed: size:>1.5g.

Date values

Dates use your local time zone. Dates are written YYYY-MM-DD. Relative ages use the units h (hours), d (days, the default), w (weeks) and y (365 days).

Sort values

Items with the same name, size or date keep their rank order.

Type categories

type: accepts the category names below, and the alternative spellings in brackets. Several categories are combined with commas.

Worked examples

6Content search

What is indexed

Content search looks inside files that are located below the content folders. The default content folder is ~, your user folder. Change the list in the Folders dialog or in the roots key of config.json.

When the name index is ready, a background thread at low priority reads the files below the content folders and builds a trigram index of their text in content.idx. After a restart, only files whose size or modified time changed are read again. While the app runs, changed files are queued and read again automatically.

File formats

Limits and exclusions

• Size cap. Files larger than content_max_size_mb (default 16 MB) are not content indexed.
• Cloud files. Online-only files from OneDrive and other cloud providers (offline or recall-on-access placeholders) are never opened, because opening them would download them.
• Skip list. Nothing below a folder in the skip list is indexed, neither names nor content.
• Links and denied folders. Junctions, symbolic links and mount points are not followed. Folders that cannot be opened (access denied) are left out.
• Excluded folder names. Files inside folders with these names are never read for content (their names are still searchable):

node_modules  .git  .svn  .hg  .bzr  obj  .vs  .idea  __pycache__  .gradle
bower_components  $Recycle.Bin  AppData  .nuget  .npm  .yarn  .cache  .tox
.venv  venv  .mypy_cache  .pytest_cache  __history  __recovery  .terraform
System Volume Information  .next  .nuxt  .angular  .parcel-cache  .dart_tool
.pub-cache  .m2  .cargo  .rustup

Search modes

Smart-case. Content search is case-insensitive, unless the pattern contains an uppercase letter: then it is case-sensitive. grep:todo finds "TODO" and "todo", grep:TODO finds only "TODO". In regex: and sym:, escaped sequences such as \W do not count as uppercase.

Results. Each hit line is one result, with the file, the line number and the line text. Up to 5 lines are reported per file, and the limit counts lines, not files. Every candidate file is read again at search time to confirm the match, so results never come from stale index data. In the bar, content rows show line N: and the text.

Symbol search

sym:Name finds lines that define Name. It recognises these patterns, which cover Pascal and Delphi, C, C++, C#, Java, JavaScript, TypeScript, Python, Go, Rust, Swift, Kotlin and similar languages:

Calls such as return Name( or if Name( are not reported as definitions.

Combining with name filters

Name filters also apply to content results: grep:TODO ext:pas in:~\Projects searches only Pascal files below that folder. Plain words in a query with a content filter must also appear in the file name: grep:TODO engine searches for "TODO" only in files whose name contains "engine".

Automatic fallback

When a query has plain words and no content filter, and no file or folder name matches, sgcFSearch searches the content instead: it looks for files below the content folders that contain every word, and reports the lines that contain any of them. Name filters still apply. The bar then shows No file name matches, showing matches inside files, and the pipe response has "content":true.

7Ranking

Match quality

Each word gets a quality tier for each name. With several words, the weakest word decides the tier of the result.

The score

With the default sort:rank, results are ordered by a score:

score = 100 x tier
      - 2 x folder depth (depth counted up to 40)
      + usage bonus (0 to 60, see Frecency)
      + 1 when the first word matches with the exact letter case

The tier dominates: one tier is worth 100 points, while the depth penalty is at most 80 and the usage bonus at most 60. Items closer to the drive root rank above deeply nested ones, and files you use often move up. Ties go to the shorter name. The other sort orders use their own field first and the score to break ties.

Frecency

sgcFSearch remembers the files you open and copy, in usage.json in the data folder. Each use adds a weight that halves every 14 days:

The usage bonus is 10 x ln(1 + sum of the decayed weights), capped at 60. A file used often and recently gets a large bonus, one used long ago fades out. Up to 4,000 paths are remembered, the lowest scores are dropped first. Copy path does not count as a use.

See Sort values for the other orders.

8Tray menu and settings

Tray icon and menu

The tray icon tooltip shows the state, for example sgcFSearch: Ready, 3,404,349 files, 447,577 folders, with , N files being read added while the content indexer works. A left click opens the search bar. A right click opens the menu:

Folders dialog

The dialog sgcFSearch Folders edits the settings in config.json.

OK saves the settings and restarts sgcFSearch in the background to apply them. Cancel closes the dialog without changes. The content size cap is not in the dialog, edit content_max_size_mb in config.json instead.

Start at login

Start at login writes the value sgcFSearch = "<path>\sgcFSearch.exe" --background under HKCU\Software\Microsoft\Windows\CurrentVersion\Run. At sign-in the app starts in the tray without showing the bar. Three places control the same entry: the tray menu item, the start at login task of the setup, and sgcfs install --login / sgcfs uninstall for the manual installation.

9Configuration reference

Data folder and files

All data lives in %LOCALAPPDATA%\sgcFSearch (for example C:\Users\you\AppData\Local\sgcFSearch), unless SGCFSEARCH_HOME points elsewhere. The folder is created when needed.

config.json

A file with every key at its default value:

{
  "roots": ["~"],
  "skip": [],
  "hotkey": "Alt+Shift+Space",
  "drives": [],
  "content_max_size_mb": 16
}

Environment variables

10Command line (sgcfs)

sgcfs talks to the running app through the pipe. A search, show, stdio and bench start sgcFSearch.exe --background from the folder of sgcfs.exe when the app is not running, print sgcfs: starting sgcFSearch... and wait up to 15 seconds for it to answer. status and stop never start it.

Commands and options

A query given as one argument, such as sgcfs "notepad ext:exe", is used exactly as typed, while with several arguments an argument that contains spaces stays one phrase, as in sgcfs "annual report" in:~\Documents. sgcfs show follows the same rule.

Output and exit codes

• Results go to stdout, one path per line. Folders end with \. Content hits are printed as path:line: text.
• The summary goes to stderr: N results (x ms), or N results, M shown (x ms) when the limit cut the list. Messages and errors also go to stderr, so piping stdout gives clean paths.

Examples

sgcfs report ext:pdf mtime:month
sgcfs -l -n 20 type:videos size:>1gb sort:size
sgcfs "annual report" in:~\Documents
sgcfs grep:TODO ext:pas in:~\Projects
sgcfs sym:ParseQuery -n 5
sgcfs show ext:log mtime:today
sgcfs status
sgcfs stop

Use the output in other commands, for example in PowerShell:

sgcfs ext:tmp in:~\AppData\Local\Temp mtime:>30d | Measure-Object
sgcfs 'package.json !node_modules\ | ForEach-Object { Split-Path $_ }

sgcfs status prints one value per line (example values):

state            ready
version          1.0.0
pid              12345
files            3,404,349
folders          447,577
crawled          3,851,926
content pending  0
content indexed  36,000
roots            C:\, D:\
log              C:\Users\you\AppData\Local\sgcFSearch\daemon.log

sgcfs bench prints a table with the columns query, server ms, round trip and results for the queries one word (config), typo (confgi), two words (system config), filtered (config ext:json,xml) and content grep (TODO in pas, txt and md files).

JSON and stdio mode

--json prints the response of a search exactly as the pipe returns it (see Requests). The summary still goes to stderr.

sgcfs --json -n 2 report ext:pdf

{"ok":true,"total":57,"ms":3.1,"content":false,"results":[{"path":"C:\\Users\\you\\Documents\\report.pdf","name":"report.pdf","dir":false,"size":48211,"mtime":"2026-10-28T08:14:00Z","score":503,"match":[0,6]}, ...]}

sgcfs stdio keeps one pipe connection open, sends every non-empty line of stdin as a request and writes every response line to stdout. It is the easiest way to use the pipe API from a script that cannot open named pipes:

echo {"id":1,"status":true} | sgcfs stdio

11Pipe API (JSON lines)

Connection and security

Requests

Search

q is a query with the full syntax. limit is optional, default 100. A limit: filter inside the query wins. The content fallback applies too.

{"id":1,"q":"report ext:pdf","limit":50}

{"id":1,"ok":true,"total":1234,"ms":0.8,"content":false,"results":[
 {"path":"C:\\Docs\\report.pdf","name":"report.pdf","dir":false,"size":12345,
  "mtime":"2026-11-03T10:15:00Z","score":512.0,"match":[0,6]}]}

Grep

A content search with the pattern apart from the query. q provides the name filters. mode is literal (default), words, regex or sym. The response has the search shape with "content":true.

{"id":2,"grep":"TODO","mode":"literal","q":"ext:pas in:~","limit":50}

{"id":2,"ok":true,"total":3,"ms":4.2,"content":true,"results":[
 {"path":"C:\\Users\\you\\Projects\\Main.pas","name":"Main.pas","dir":false,"size":8120,
  "mtime":"2026-10-30T15:02:11Z","line":42,"text":"  // TODO: check the result","tmatch":[5,4]}]}

Status

{"id":3,"status":true}

{"id":3,"ok":true,"state":"ready","files":5080000,"folders":610000,"crawled":123456,
 "content_pending":42,"content_indexed":36000,"pid":1234,"version":"1.0.0",
 "roots":["C:\\","D:\\"],"log":"C:\\Users\\you\\AppData\\Local\\sgcFSearch\\daemon.log"}

state is one of starting, loading, crawling, catchup, ready, stopping, stopped (see Index states). roots lists the indexed drives.

Use

Reports that the user opened or copied a file, to feed the frecency ranking. kind is open or copy.

{"id":4,"use":"C:\\Docs\\report.pdf","kind":"open"}

{"id":4,"ok":true}

Show

Opens the search bar, optionally with text.

{"show":true,"q":"optional text"}

{"ok":true}

Stop

The app replies, then saves its index and exits.

{"stop":true}

{"ok":true}

Errors

Errors come back as one line with "ok":false and a message:

{"id":5,"ok":false,"error":"text"}

A request with no known operation gets {"ok":false,"error":"unknown request"}. An invalid regular expression in re: is reported as a warning and ignored, while an invalid content pattern returns an error.

Client sample

An illustrative PowerShell client that sends one search and prints the paths. It is a sample, not part of the product:

# Illustrative sample: one search over the sgcFSearch pipe
$sid  = [Security.Principal.WindowsIdentity]::GetCurrent().User.Value
$pipe = New-Object IO.Pipes.NamedPipeClientStream('.', "sgcFSearch-$sid",
          [IO.Pipes.PipeDirection]::InOut)
$pipe.Connect(5000)
$utf8 = New-Object Text.UTF8Encoding($false)
$w = New-Object IO.StreamWriter($pipe, $utf8)
$w.NewLine = "`n"
$w.AutoFlush = $true
$r = New-Object IO.StreamReader($pipe, $utf8)
$w.WriteLine('{"id":1,"q":"report ext:pdf","limit":10}')
$resp = $r.ReadLine() | ConvertFrom-Json
if ($resp.ok) { $resp.results | ForEach-Object { $_.path } } else { $resp.error }
$pipe.Dispose()

The same connection can send more requests before it is closed. Without a named pipe library, run sgcfs stdio as a child process and exchange lines through its stdin and stdout.

12How it works

Architecture

The crawl

The crawler lists folders with the native NT API: each folder is opened relative to its parent handle (NtOpenFile) and listed in bulk, many entries per call, instead of one FindNextFile call per file. Several threads crawl in parallel. Junctions, symbolic links and mount points are recorded but not entered, which avoids loops and duplicates. Folders that cannot be opened are marked as denied, folders in the skip list are never opened, and cloud placeholder folders are crawled and marked as offline.

Before a drive is crawled, its change journal position is stored. After the crawl, the changes made during the crawl are applied from the journal, so nothing is missed.

The name index

The index lives in memory: one compact record per file or folder (parent, first child, next sibling, name, size, modified time, flags), with the names stored once in a shared pool. Searches scan it on several threads and keep only the best results, which is why a query over millions of names takes milliseconds. The index is saved to names.idx together with the journal position of each drive, and loaded in about a second at the next start.

Live updates

• While running. A watcher per drive uses ReadDirectoryChangesW on the whole drive. Changed folders are listed again and the differences applied. Renamed and moved folders keep their contents. A change is visible in searches within a fraction of a second.
• Buffer overflow. When too many changes arrive at once, the watcher catches up from the change journal, or rescans when that is not possible.
• After a restart. The NTFS change journal is read from the saved position, through the unprivileged journal interface, so no administrator rights are needed. Only the folders that changed are listed again. A drive without a usable journal is rescanned.
• Checkpoints. Every 5 minutes, when the index changed, the journal positions are refreshed and names.idx and content.idx are saved. They are saved again on exit. An index that was interrupted during a crawl or catch-up is never saved, so a later start does not trust a partial index.

The content index

The content indexer walks the content folders through the name index, so it does not crawl the disk again. It extracts the text of each file (see File formats) and records every trigram, a sequence of three case-folded characters, with the list of files that contain it. A content query first takes the trigrams of the pattern (for a regular expression, of its required literal parts) to find candidate files, applies the name filters, then reads each remaining candidate again on a few worker threads to find the exact lines. Files that turn out to be changed or deleted are queued for reindexing.

13Performance

QA measurements on one PC, with a Release build. They are not a guarantee: results depend on the disk, the CPU and the number of files.

The query times are server times as reported by sgcfs bench (median of the runs). The five bench queries are described in Examples. Run sgcfs bench to measure your own PC.

14Troubleshooting and FAQ

The hotkey does not open the bar

Another program may own Alt+Shift+Space. sgcFSearch then falls back to Win+Shift+F, then to Ctrl+Alt+F. To see which one is active, open the tray menu (the hotkey is shown next to Search) or the log:

app: hotkey Alt+Shift+Space not available, error 1409
app: hotkey Win+Shift+F active

Two other messages can appear: app: invalid hotkey "..." when the setting cannot be parsed, and app: no global hotkey available when none of the three could be registered.

Choose a free combination in the Folders dialog. Without any hotkey, open the bar from the tray icon or with sgcfs show.

Where is the log?

%LOCALAPPDATA%\sgcFSearch\daemon.log. The tray menu item Open log opens it and sgcfs status prints its path. It holds starts and stops, the hotkey, watcher and journal events, and errors. At 2 MB it moves to daemon.log.old.

How do I rebuild the index?

1. Quit sgcFSearch from the tray menu, or run sgcfs stop.
2. Delete names.idx from the data folder to rebuild the name index. Delete content.idx to rebuild the content index.
3. Start sgcFSearch again. It crawls the drives as on the first run.

What happens when I start sgcFSearch twice?

Only one instance runs per user. A second start does not build another index: it asks the running instance to show its search bar and exits. Started with --background, the second instance exits silently.

A file I know exists is not found

• Check the state. During the first crawl the index is still partial.
• Check the skip list and the Drives setting in the Folders dialog. Only fixed drives are indexed, network shares and removable drives are not part of the default set.
• Folders your account cannot open (access denied) are listed, but their contents are not.
• Junctions, symbolic links and mount points are not followed. Search for the file at its real location.
• Words match the name, not the path. Use a path token, in: or path: for folder names.
• ext:, type: (except folders) and size: (except size:empty) match files only. in: matches the contents of the folder, not the folder itself.
• Look at the status line for a warning about an ignored filter value.

A text is not found inside a file

• The file must be below a content folder (default: your user folder).
• It must not be larger than content_max_size_mb (default 16 MB).
• It must not be inside an excluded folder such as AppData, node_modules or .git (see Limits and exclusions).
• Online-only cloud files are not read. Make the file available offline to have it indexed.
• PDF and legacy Office files need a Windows IFilter for their extension.
• After a start, the content indexer may still be reading files: watch for N files being read in the tray tooltip, or content pending in sgcfs status.
• An uppercase letter makes the search case-sensitive (smart-case).
• Plain words next to a content filter must match the file name.

Why does the bar show lines from inside files?

No file or folder name matched your words, so the content fallback searched the content folders instead. The status line says so.

sgcfs says "sgcFSearch is not running"

status and stop do not start the app. Start sgcFSearch.exe, or run any search or sgcfs show, which start it in the background. If the start fails, check that sgcFSearch.exe is in the same folder as sgcfs.exe.

My manual changes to config.json have no effect

The settings are read at start. Run sgcfs stop and start sgcFSearch again, or use the Folders dialog, which restarts it for you.

15Version history

1.0.0, November 2026

First release.

• Name index of every fixed drive, built by a multi-threaded crawler on the NT directory API and saved to names.idx.
• NTFS USN journal catch-up at startup and live change watcher.
• Search bar with a global hotkey (Alt+Shift+Space, with fallbacks), ranked results, icons, filter chips and history.
• Query syntax: name modifiers, path tokens, ext:, type:, kind:, in:, path:, size:, mtime:, re:, limit: and sort:.
• Content index of text files under the content folders, with grep:, regex: and sym: content search.
• Frecency ranking from opened and copied results.
• Named pipe JSON lines API restricted to the current user.
• sgcfs command line client: search, JSON output, stdio bridge, status, stop, show, install, uninstall and bench.
• Tray icon with the index status, Folders dialog, Start at login and Open log.
• Single instance: a second start shows the search bar of the running instance.
• Distributed as a signed per user installer (recommended) and as a zip. The setup offers the PATH entry and Start at login as tasks; with the zip, sgcfs install adds them.

sgcFSearch 1.0.0. Copyright © 2026 eSeGeCe. https://www.esegece.com