HEE-VIEW(1) HEE Tools HEE-VIEW(1)
NAME
hee-view - hee-view: one command, the whole real picture -- pve node,
every
SYNOPSIS
hee-view [-h] [--pve-host PVE_HOST] [--haproxy-host HAPROXY_HOST]
hee-view [--pve-host IP] [--haproxy-host IP] [--dns-host IP]
hee-view --sites [--sitemap URL_OR_PATH] [--network lab|public|all] [--nerd]
DESCRIPTION
[--dns-host DNS_HOST] [--nerd] [--sites] [--sitemap SITEMAP]
[--via [USER@]HOST] [--crawl] [--braindump] [--job JID]
[--json] [--root DIR] [--tickets] [--workspace DIR]
[--offline] [--jobs] [--md] [--live] [--publish VMID:PATH]
[--mark-seen] [--braindump-url URL] [--sites-only]
[--network {lab,public,all}]
hee-view: one command, the whole real picture -- pve node, every
container, HAProxy backend health, DNS. Same "only view to see them
all" intent as view.lab.tcos.us, as a real CLI, no do-this-do-that.
Real trigger (2026-08-24, Spencer): "not going to all this do this do
that. slick hee view for all." Reuses hee-pve-health's real data
(shells out to it --json, doesn't reimplement pvesh calls) and adds
HAProxy + DNS checks in one terse report.
Real trigger for --sites (2026-08-28, Spencer): "first thing may be to
build a complete site map, setup monitor (simple) with hee view -sites
(or similar) to make sure we get 200's everywhere, condense view
(adding -nerd to that will be verbose as fuck)" -- reuses the real,
already-established --nerd split (condensed by default, full per-URL
detail only with --nerd) rather than inventing new verbosity flags.
Default source is the real, live profile/SITEMAP.yaml on .github's
main branch -- pass --sitemap to check a local copy instead.
EXAMPLES
$ hee view --jobs --md --root tests/fixtures/dispatch # ci
$ hee view --job demo-2026 --root tests/fixtures/dispatch # ci
$ hee view --braindump # reads view.lab; not run in CI
$ hee view --jobs --live --publish 107:/www/data/jobs.json # what the kiosk timer runs; needs the lab
$ hee view --tickets --offline --root tests/fixtures/dispatch --workspace tests/fixtures/tickets-workspace # ci
$ hee view --tickets --publish 107:/www/data/tickets.json # every repo's tickets, jobs, blockers, open PRs; needs gh and the lab
options:
-h, --help show this help message and exit
--pve-host PVE_HOST
--haproxy-host HAPROXY_HOST
--dns-host DNS_HOST
--nerd per-container cpu/mem/disk-io/net-io/elapsed + totals;
also full per-URL --sites detail
--sites check every real URL in the site map answers 200-399
--sitemap SITEMAP sitemap source -- URL or local path (default: real
.github/profile/SITEMAP.yaml on main)
--via [USER@]HOST an ssh target or alias, never a URL: run the --sites
checks FROM that host (this file is sent with `python3
-`; the host needs only python3). The point is a
vantage that is not the operator's own IP: the
DigitalOcean droplet (man.tcos.us) sees tcos.us the
way the public does -- Cloudflare edge, no LAN split-
DNS, no home-IP exemptions. Implies --sites-only and
--network public unless given.
--crawl with --sites: follow same-host links from each listed
site (depth 2), check every page found, follow
blog->media redirects to a 2xx, and report reachable
pages the site map does not list
--braindump what has been written to the view.lab brain dump since
you last looked. Append-only file, so the only
question is what is NEW -- this keeps a cursor and
reports the difference. WARNING exit (1) when there is
something unread, so it drives a watch or a status
line.
--job, -job JID one dispatched agent job, by id or unique prefix: the
dispatch record and results when finished; while
running, the container's cgroup, the claude process
and out/ read from the pve host
--json with --job: also print the raw record or probe as JSON
--root DIR with --job/--jobs: the checkout whose .hee/dispatch to
read (default: this repo, then ~/git/fleet-ops)
--tickets, -tickets every repo's tickets with their flow judged, every
agent job and what blocks it, and the org's open PRs;
--json prints the feed, --publish VMID:PATH writes it
for view.lab/tickets.html
--workspace DIR with --tickets: where the repos live (default ~/git)
--offline with --tickets: read no pull requests (no gh, no
network)
--jobs, -jobs every dispatch record (current repo, then ~/git/fleet-
ops): one row per job and a per-agent scorecard
--md with --jobs: Markdown tables, for committing as
pve/agents/SCORECARD.md
--live with --jobs: also probe every running job on the pve
host (cgroup, process, out/)
--publish VMID:PATH with --jobs: write the JSON to that path inside a
container on the pve host (view.lab:
107:/www/data/jobs.json) instead of printing it --
what the 30 s timer on kiosk runs
--mark-seen with --braindump: remember the newest entry, so the
next run reports only what arrives after it. Separate
from reading on purpose -- a run that scrolled past
should not silently mark it read.
--braindump-url URL override the dump endpoint (heerc:
HEE_VIEW_BRAINDUMP_URL, default
https://view.lab.tcos.us/cgi-bin/braindump.cgi)
--sites-only skip the pve/haproxy/dns section; just the site checks
--network {lab,public,all}
only check sites on this network. A GitHub-hosted
runner can only reach 'public'; an on-host checker
inside the lab can reach 'lab'. Default all.
HEE-VIEW(1)