v0.5.0

eidos v0.5.0

The first release of eidos: a private, self-hosted search for the files on
your Windows computers and network shares. It watches your drives, keeps a
catalog of every file and folder, indexes the text inside them, and answers
from one web page - by name, by path, by what a file contains, by what a
folder holds, by size, by date - without ever changing a file.

Download and install

Download eidos-v0.5.0-setup.exe from this release and run it. 64-bit
Windows 10 or later (Windows Server 2019 or later).

The setup asks two things: whether eidos is just for you (no administrator
rights; it runs while you are signed in) or for all users on this computer
(a Windows service that starts with the machine and keeps indexing when
nobody is signed in), and where to keep its data. Then open
http://127.0.0.1:7700/ (the Start-menu entry does), add a drive or a
share as a source, and start searching while it indexes.

Everything is signed. eidos-v0.5.0.msi is the bare package for
administrators and scripted installs; *.sha256 are checksums. The
eidos-collector-* files are a development profiler used to measure eidos
itself - you do not need them and can ignore them.

See installing.md
for unattended installs, service accounts, and network shares.

What you get

Your files, found fast. Local NTFS drives are enumerated natively and
kept current from the change journal, so a saved file shows up in search
within seconds. Network shares are crawled read-only and refreshed on a
schedule, and every result says how fresh its source is.

Search the way you remember. Type a few words and get ranked matches on
names and paths. Narrow with fields: ext:cs, size:>1M, mtime:7d,
path:G:\Tools, kind:dir, attr:hidden, source:G; use globs
(*.log), regular expressions (name:/^v\d+/), phrases, OR, grouping,
and - to exclude. The full grammar is in
query-syntax.md.

Search inside files. Text files - source code, logs, notes, JSON, HTML,
configuration, anything eidos recognises as text in any common encoding -
are indexed as you add sources. content:zephyr finds the word,
content:"build 4.2" the phrase, content:=ErrCode an exact
case-sensitive identifier, content:/re/ a regular expression. Hits come
with line-aware snippets and a preview of the file. Indexing runs in the
background with bounded memory and never blocks name search; you can watch,
pause, and resume it from the Activity page.

Folders are results too. Ask for directories, not just files:
has:idb has:cs finds a tree that contains both kinds anywhere beneath it;
subtree:>1G, files:>1000 and subtree_mtime:7d rank folders by what
they hold. Browse any tree, see apparent and allocated sizes, and open a
treemap of where the space went.

Archives, listed. ZIP files are inventoried, so the members of an
archive are searchable by name, path and size without unpacking it.

Honest results. Every answer carries the completeness of the sources
behind it: whether a source is fully indexed, still catching up, or
offline. An unavailable share keeps its last known contents instead of
silently returning nothing.

Take it with you. Export any result set as CSV, JSON or NDJSON; save
searches; use the same queries from the command line (eidos search) or
the HTTP API the web page itself uses.

Nothing is written to your files. eidos only reads. Its catalog and
indexes live in its own data folder.

Also in this release

• macOS (Apple silicon and Intel): no installer package yet, but the
  agent builds from source in one script and runs as a LaunchAgent with
  live updates from FSEvents. See
  installing-macos.md.
• Private fleet (experimental). Several installations can replicate
  their catalogs into one central and search the union across hosts
  (host:laptop ext:dmp). It is off until you explicitly enrol a node from
  the command line, uses mutually authenticated TLS with pinned per-node
  certificates, and replicates metadata only - content search stays on the
  machine that holds the file. See
  fleet.md.

Good to know

• Keep it on loopback. There is no login yet. eidos listens on
  127.0.0.1 by default; on any other address, everyone who can reach the
  computer can search and read what it indexed.
• Content search covers plain text. Office documents, PDFs and
  scanned images are catalogued by name, path and size but their contents
  are not extracted in this release. ZIP members are listed, not indexed;
  other archive formats are plain files for now.
• The fleet is a first version: one central, no discovery or relay,
  command-line enrolment, and nodes must be able to reach each other on
  the private network.
• Linux is not supported yet.

Upgrading later

Run the newer setup: it upgrades in place and keeps your settings, data
and indexes. Removing eidos keeps the data folder unless you ask for it to
be deleted; the files it indexed are never touched.

Feedback

Bugs and ideas: https://github.com/josiah-nelson/eidos/issues.