Search

Flux has four search surfaces:

• Live filter - the search bar at the top of the window. A fuzzy, case-insensitive match on the file names of the current directory, updated as you type. No disk walk.
• Recursive search - starts from the search bar when the query contains * or ? and is longer than one character, or from the advanced panel. Walks the current directory tree.
• Content search - :term or :.ext:term in the search bar, started with Enter, or the Inside files field of the advanced panel. Reads file contents.
• Tag filter - triggered by #name, :tag:name, or :t:name. Looks up files by extended-attribute tag.

Size filters

Prepend > or < to a size token, or use a .. range, to filter by file size. A unit is mandatory - a bare number is rejected, because >10 matches everything and would lock the UI on large trees.

The ib and b suffixes are stripped before the unit is looked up, so 10MB, 10MiB, and 10M all mean 10 MiB.

In the live filter, the rest of the query is a case-insensitive substring match on the name, and folders always pass the size filter. A query containing * or ? is read as one glob, so it cannot be combined with a size token.

Examples:

>10MB
<500KB
10MB..50MB
report >10MB

Tag filters

Prefix a comma-separated tag list with one of the following. All three are equivalent.

Multiple tags are AND-ed. A free-text query after the tag list is also AND-ed:

#work                  - files tagged "work"
#work,urgent           - files tagged both "work" and "urgent"
#work report           - files tagged "work" whose name contains "report"

Content search

Prefix the query with : to search inside file contents, then press Enter. Content search never starts while you type. It walks the current directory recursively and returns the first match of each file.

The term must be at least 3 bytes long, not 3 characters. Multiple extensions can be comma-separated:

:hello
:.rs:hello
:.rs,py:hello

Details:

• Matching ignores case for ASCII letters only.
• Files are read in blocks of 256 KiB.
• A file is skipped when a NUL byte appears in its first 1024 bytes.
• A file larger than ui.content_search_max_file_mb is skipped. 0 removes the limit.
• The walk honors .gitignore, .ignore, the git global and exclude files and the ignore files of parent directories. It follows symlinks and skips hidden files unless the session shows them.

Recursive walk

The search bar starts a recursive walk when the query contains * or ? and is longer than one character. The whole query is then one glob pattern. A bare .ext does not start a walk, it is a fuzzy name filter.

The advanced panel starts a walk when Search inside subfolders, Only search folders or Regular expression is on, or when File name contains * and Exact match is off.

The walk matches files only, never directories, unless Only search folders is on. It honors the same ignore files and symlink rules as content search, and it stops at ui.max_search_results results.

Results are shown in list mode, one row per file, with a path relative to the current directory. They reach the window in batches: a batch is flushed when it holds 500 results or after 16 ms, whichever comes first, so the GTK main loop is never saturated. The batching applies to this walk, not to content search.

Advanced panel

Ctrl+F (shortcuts.search) toggles the Search panel on the right and switches the header to the search bar. The panel also opens from the Search entry in the sidebar, a place with the path search:// that the first-run config adds.