In windows opened by netssh multi, prefix S toggles broadcast for all
panes and prefix e excludes the current pane (a pane-level
synchronize-panes off, shown as [SOLO] in the border) or includes it
again. netssh sync --pane does the same from a shell.
S and e are unbound in a default tmux (E would be "spread evenly").
netssh writes the bindings to a small tmux config and sources it, binds
a key only if the user's tmux.conf does not use it (netssh's own earlier
binding is refreshed), and the bindings act only in windows marked
@netssh. multi.sync_key and multi.solo_key change the keys or turn them
off (none); also in the configurator. zellij syncs whole tabs only, so
excluding a pane is tmux-only.
testenv: the render test presses the keys through the xterm client
(send-keys -K) and checks [SOLO], that broadcast skips the excluded
pane, that S toggles, and that other windows are unaffected. The client
gets GNU grep: BusyBox grep -oF counted only one match per line.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
tests/16-render.sh runs tmux, zellij, and the TUI in xterm on Xvfb,
takes screenshots, and reads them back with tesseract: remote prompts
in all panes, borders and titles with the hosts, [SYNC] appearing and
disappearing, panes side by side, broadcast output in every pane and
input in one pane only without it, windows split by panes_per_window,
the status bar, and TUI marks. Screenshots stay in testenv/.artifacts.
OCR reads three preprocessing variants (plain, stretched for dim gray,
brightest channel for colored text) and normalizes characters that
DejaVu Sans Mono makes ambiguous (1/l, C/().
The screenshots showed tmux cutting the session name in the status bar
to "[netssh-pv"; netssh multi now widens status-left-length for its
own session.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
netssh multi opens one pane per host in a tmux window or zellij tab.
Hosts come from names, tag:TAG (NetBox and static hosts), cluster:NAME
(NetBox virtualization clusters; devices by default, --kind vm|all),
and groups from the config (@name, nested, with cycle detection).
Every pane runs netssh HOST, so lookup, rules, and jump hosts apply.
- tmux: a new window inside tmux, else a new session that is attached;
tiled layout, pane borders with the host and a [SYNC] marker.
- zellij: a generated grid layout; a new tab inside zellij, else a new
session.
- --sync broadcasts typing to all panes; netssh sync toggles it in the
current tmux window or zellij tab.
- TUI: Space marks hosts, Ctrl+A marks all, Enter opens them together.
- Configurator: multiplexer, panes per window, groups.
NetBox SearchAll gains a cluster filter (name to cluster IDs per
instance), and cache entries keep the host's cluster for completion
and an offline fallback. Panes skip the 24-hour background refresh
(checked once by multi) and get HOME/XDG paths but no NETSSH_* values,
which would expose tokens in the process list.
testenv/tests/15-multi.sh checks real tmux and zellij sessions,
including a broadcast command arriving on all three cluster nodes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
testenv/ starts two real NetBox 4.7 instances (v2 and v1 token), SSH
targets on public, mgmt, and prod networks, a jump host, an IPv6-only
host, a static host, and a client with ssh, sftp, scp, tmux, zellij,
bash, zsh, and fish. A seed script fills NetBox through the REST API.
141 end-to-end checks cover cache and refresh filters, connecting,
all resolver strategies, multiple instances, shortcuts and rewrites,
SSH options and rules, static hosts, sftp/scp incl. IPv6 and jump
hosts, tunnels, local notes and NetBox comments, completion and shell
hooks, the TUI (driven through tmux), the configurator, and uninstall.
Local only: testenv/up.sh, test.sh, shell.sh, down.sh. Not part of
CI/CD.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Found by the new Docker test environment against a real NetBox 4.7:
- The interface_name strategy filtered IP addresses with interface_name
and vminterface_name, which NetBox ignores, so it returned the first
address of the host instead of the named interface's. Use the real
filters interface and vminterface; a missing interface (HTTP 400)
falls through to the next strategy.
- sftp and scp took instance-qualified hosts (prod/web01:/path) for
local paths, because a "/" comes before the ":". Such arguments are
remote when the prefix is a configured NetBox instance.
- Shell aliases of instance-qualified hosts ran the shortcut rules on
"instance/host", so anchored rewrite rules never matched.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Add the hostname resolver strategy. It connects by the NetBox name
instead of an IP, with resolver.hostname_domain appended to names
without a dot. Its position in the chain decides whether it is always
used or only as a fallback (e.g. [primary_ip, hostname]); names that
are not valid DNS hostnames fall through to the next strategy.
Add static hosts under hosts: for machines that are not in NetBox,
with address, user, port, tags, and their own SSH options. They are
checked before NetBox (by name or shortcut), never expire, survive
cache refresh and clear, and work without any NetBox configured. The
cache overlays them after every load and stores only their last use,
so they appear in completion, aliases, the TUI and its recent list.
SSH options merge global, static host, then rules; rules can match
kind: static. Manage them with netssh hosts list|add|remove and in
netssh configure, which can now also continue without NetBox.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Add a table of contents, a quick start, a full configuration reference
with defaults, the environment variables, a table of files and
directories, a command overview, details on rule matching and how
option layers combine (with real --dry-run output), recipes, a
troubleshooting table, and the project layout. Update How it works for
sftp, scp, tunnel, and forwards.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A new major version of a Go module is a separate module path. For an
indirect dependency nothing imports it, so Renovate only adds an unused
requirement, as in #24 (go.yaml.in/yaml/v3 to v4). Minor and patch
updates of indirect modules stay enabled.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Add netssh sftp and netssh scp. Both resolve the host like a connection
does (cache, shortcuts and rewrites, instance/host, default user and
port, SSH options and rules, --jump/--no-jump, --dry-run) and get shell
completion. The per-client flag tables handle the differences from ssh:
the port is -P, -l is a bandwidth limit (the default user goes in as
-o User=), and IPv6 addresses are bracketed in host:path. All remote
scp operands must name the same host.
Add named port forwards under ssh.forwards (global, per instance, per
rule; a more specific layer replaces a forward of the same name),
netssh tunnel HOST NAME|SPEC... (ssh -N with ExitOnForwardFailure) and
netssh --forward NAME. Specs such as 8080:localhost:80 work without
configuration. Forwards are editable in netssh configure and completed
in the shell; the TUI opens SFTP with ctrl+o and a tunnel with ctrl+t.
Host resolution moves out of runSSHWrapper into resolveTarget, shared
by ssh, sftp, scp, tunnel, and the TUI.
Options added by netssh are now inserted before the first operand or
"--": after "--", ssh read them as the destination, and scp stops
reading options at the first operand.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
shortcuts.rewrites applies regex replacements after the domain suffix
is stripped, e.g. ^([a-z]+)[0-9]+- -> $1 turns fsn1-bx1.hoster.de into
fsnbx1. The shortcut works for connecting (cached hosts), shell
aliases, and Tab completion, where it is offered when the full name
does not match the typed text. Rules are editable in netssh configure,
and netssh alias --test HOSTNAME shows a name's shortcut and the cached
hosts it matches.
The shortcut lookup only ran when domains or strip_hyphens were set;
it now also runs for rewrite-only configurations.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After the NetBox instances, netssh configure now opens a menu with all
sections: SSH defaults, global SSH options (jump host, host key
checking, target, address family, extra args, -o options), SSH rules,
resolver & cache (including the cache file), shortcuts & command alias,
and completion.max_results. SSH options can also be set per instance.
The wizard rebuilt the config from its own fields, so re-running it
dropped global SSH options, ssh.rules and completion.max_results from
the file. It now edits a copy of the loaded config, and save validates
SSH options and rules before writing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
bash (and zsh without a matcher) only accept candidates that start with
the typed text, so "fsn1-bx1<Tab>" found nothing when every candidate
was qualified as "prod/…". Hostnames that exist in only one instance
are now offered bare; names present in several instances stay
qualified. Connecting to a bare name still checks every instance for
ambiguity, so the connection behavior is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With several NetBox instances, netssh offers "prod/fsn1-bx1…" for the
typed text "fsn1". compadd discarded these matches because they do not
start with the typed word, so <Tab> found nothing. The matcher
'l:|=*' lets zsh accept text before the typed word; netssh has already
filtered the candidates.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Remove everything netssh and install.sh set up, also leftovers of
older versions:
- netssh blocks and unmarked legacy lines in bash, zsh, and fish
profiles
- completion files for bash, zsh (incl. the oh-my-zsh cache), fish,
and the command alias; files not generated by netssh are kept
- the cache directory with hosts, aliases, and per-instance caches
- the config file and local notes; the config is first copied to a
private netssh.yaml.backup-<timestamp> because it holds API tokens
- the netssh binary; for protected paths the sudo command is printed
--keep-config keeps config and notes, --dry-run only lists the plan,
and --yes skips the confirmation.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`netssh --version` used to be passed to ssh and failed with
"unknown option -- -". Release builds now embed the tag via
-X main.version; other builds fall back to the module build info.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Add SSH settings that apply only to netssh connections, so
~/.ssh/config does not need global changes:
- proxy_jump (passed as -J; "none" disables an inherited jump host)
- host_key_checking: strict | accept-new | off (off prints a notice)
- target: ip (default) | name | hostname; "name" connects to the
NetBox name with -o HostName=<ip> and -o HostKeyAlias=<name>, so
`Host *.example.com` rules apply and known_hosts entries follow the
device instead of a reused IP
- address_family: any | inet | inet6
- options (-o Key=Value) and extra_args
Settings layer global -> netboxes[].ssh -> ssh.rules (match by tag,
instance, kind, or name glob). Options given on the command line
always win: -o values are inserted after the user's own, and
extra_args before them.
New wrapper flags: --jump HOST, --no-jump, and --dry-run, which prints
the final ssh command instead of connecting. The TUI uses the same
options. The setup wizard keeps all new settings when it saves.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sebi
