`FlowStore._git` now passes `-c gc.auto=0`, so `git commit` no longer forks a
background `gc --auto` that reparents onto PID 1 and stays there unreaped. With
auto-gc off nothing packs on its own, so `_commit` runs a foreground `git gc`
every 500 commits — the trigger a long-lived seeding session actually reaches.
`docker/compose.yml` takes the frontend's `VITE_API_URL` build arg from the
environment, keeping `https://api.${DOMAIN}` only as the fallback. `setup.sh`
already derives the scheme from `ENVIRONMENT`, so an `up --build` that does not
layer `compose.local.yml` stops shipping a bundle that calls `https://api.localhost`.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01StpRc2C6au1WJ1EUU7fsfu
A node's stop() and a supervised task's cancellation are both waited on
inside the rebuild lock, and neither had a deadline: an MQTT client whose
broker never acknowledges the disconnect leaves aiomqtt's __aexit__
waiting forever, so reload() never returned and every start, stop or
publish behind it hung until the container was restarted.
Each node now gets five seconds to close and is abandoned after that, and
cancel_all reports what is still running rather than waiting on it — it
also no longer swallows a cancellation aimed at the caller, which used to
make the lock holder unkillable. A rebuild asked for by a request gives up
on the lock after fifteen seconds with RebuildBusy, answered as a 503.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01StpRc2C6au1WJ1EUU7fsfu
On a CSS-scaled dashboard panel uPlot reads the pointer in visual pixels
and measures it against its own unscaled plot width. Refining the result
in cursor.move left uPlot's earlier arithmetic wrong: cacheMouse snaps an
offset within 1px of the plot edge to the plot width, and on a panel
scaled up the raw visual offset passes that edge at 1/drawn of the way
across, so the readout stopped advancing partway over the chart and stuck
to the last point.
cursor.bind wraps the three events that carry a position and hands uPlot
a corrected clientX/clientY, so every step after it — the snap, the drag
detection, the redraw path — comes out right by construction. It is also
stateless, which removes the cursor.move memo and the drag.click
workaround that the mousedown/mouseup mismatch needed.
cursor.check.ts now replays uPlot 1.6.32's own pipeline against the
shipped cursor config, for panels scaled both down and up.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
Reopened with the actual trigger: the base compose builds the frontend
against https://api.${DOMAIN} and only compose.local.yml overrides it,
so any build without that layer breaks login locally. I closed this
earlier on the mistaken belief a rebuild had settled it.
Every service builds with `context: ..`, and Docker only consults the
.dockerignore at the context root — the copies under frontend/ and backend/
never applied, so the host's dist/, node_modules/ and caches were shipped
into each build context.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CB8uwHkH52qxLxAkGN4gTo
Eleven "to be sorted" entries are done. Three that survive are amended
rather than removed: the house panel's chart budget (the range picker
gave its row back), layoutGraph (height is derived now, width is still
guessed), and the dark:-variant scope, which turned out to share a root
with the React Flow colorMode bug. Adds the portal's missing
not-reporting state, and drops the frontend-origin entry the rebuild
settled.
uPlot swallows the click that ends a drag, and decides a drag happened by
comparing the position it took at mousedown against the one it holds at
mouseup. It refines the first through `cursor.move` and re-reads the second
raw, never refining it (`mouseUp` -> `cacheMouse(initial: false)`), so with any
correction in place the two never agree: on a scaled panel every click on a
plot read as a drag and was stopped before it reached the page. A chart tile
could not be selected in the dashboard editor by clicking the chart, and a
Health chart could not be clicked to pin a moment.
Not the cursor correction being applied twice — that was a separate defect,
and idempotence could not have fixed this one, since the mouseup path never
applies the correction at all. Not `cursor.y` either: that only decides
whether the y cursor element is created, and both axes are read from the event
regardless.
Drag-to-zoom goes with it, and loses nothing: `setData` re-ranges the scales
from the data on every render, so a dragged range was erased by the next
reading. With no drag there is no click to protect from one.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
A chart that named no palette took the ramp in order, so three lines drew
slots 1, 2 and 3 — adjacent steps of a ramp that carries identity by lightness
alone, which at a 2px stroke read as close to one picture. Each count now
takes the slots furthest apart that the ramp allows: 1 -> 1, 2 -> 1/5,
3 -> 1/3/5, 4 -> 1/2/4/5. Five lines are unchanged.
Moving a default is only safe because the mechanism around it is inert: a
palette a dashboard wrote down is still drawn exactly as written, so no stored
document means anything different than it did. `palette.check.ts` now pins the
spread table and says which of the two properties is permanent.
An unnamed palette is empty rather than the whole ramp, so the settings panel
shows nothing picked when a dashboard is on automatic — and deselecting the
last colour is now the way back to it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
React Flow stamps `colorMode` on its wrapper as a class and defaults it to
`light`. The app's own token scopes are named `.light` / `.dark` — the classes
that let a dashboard be forced to one theme inside a shell on the other — so
every canvas was silently redeclaring the light palette on its own subtree:
white node cards and a white canvas in dark mode, with the inherited
`--foreground` text still near-white and therefore invisible. Home's neurons
had the same fault, drawn with white ring gaps and the light `--primary`.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
The comment said the canvas follows the theme without a `colorMode` prop. It
does not — React Flow stamps that prop on the wrapper as a class and defaults
it to `light`, which collided with the app's own `.light` token scope and
re-themed the whole canvas. The prop landed in 4e1898a; this is its reason.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
Reserving a slot for all three status controls left 72px for a node's title,
against about 140px before. Only the stop button and the dot come and go on
their own; the traceback button appears when a node newly fails and goes when
someone acknowledges it, which is a thing to notice rather than a flicker. It
goes back in the row, and the title gets 106px.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
uPlot writes what `cursor.move` returns back into the value it hands in next
time and calls it again on every redraw — and a chart sets its data on every
render. The correction was therefore applied repeatedly: the cursor walked
left while the pointer stood still, and the position taken at mousedown no
longer matched the one held at mouseup, which uPlot reads as a drag and
answers by swallowing the click. That is what stopped a chart tile being
selectable in the dashboard editor.
The refiner now returns its previous answer unchanged when handed it back, so
a redraw is a no-op. Lifted out of the config as `cursorRefiner` so the branch
can be checked without a browser; `cursor.check.ts` replays uPlot's own
sequences, including the press-is-not-a-drag one that regressed.
The Playwright hover test now also asserts the cursor holds its place across a
redraw, which is what the first version missed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
A dashboard names the data colours its charts draw with: an ordered, distinct
subset of the `--chart-1…5` ramp, stored on the settings channel that already
carries `theme` and `locked`. No backend or client change — `SettingDef.value`
is free-form and a name this build does not wire up is left alone rather than
refused.
Naming none is the whole ramp, which resolves to the identical token per
series as the `--chart-${(index % 5) + 1}` charts drew with before, so every
existing dashboard is unaffected. `palette.check.ts` asserts that equivalence
rather than trusting it.
Distinct slots, not free assignment with repeats: only slot 1 against slot 5
clears 3:1 within the ramp, so two traces on one slot could not be told apart.
Status colour is deliberately outside the palette — a fault is `--destructive`
because of what it means, not because of where it sits.
The provider is mounted by the editor as well as the view, so picking a
palette repaints the charts beside the panel instead of describing them.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
- A node's height follows the ports on its busiest side. It is a function of
the document, so `layoutGraph` reserves exactly what is drawn and nothing
measured is fed back into the layout.
- The three status controls now sit in slots that are there whether the
control is or not. A node running many times a second mounted and unmounted
the stop button on every execution, resizing the card each time.
- A port bound to another flow's message is drawn as a label, naming the node
at the far end and its type. Only the opposite direction was answered
before. The scan behind both is now cached on the store's commit counter
rather than reading every flow per request.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
A dashboard canvas is CSS-scaled to fit its panel while uPlot maps the
pointer against its own unscaled plot width, so the cursor drifted further
right the further into a chart it went. A `cursor.move` refiner divides the
visual offset back into layout pixels; unscaled hosts get a no-op.
A querying chart's range picker moves onto the frame's title line through a
new `useHeaderSlot`, giving the plot back the row it spent. The editor's drag
handle is the header, so the picker is exempted from it.
Charts can be drawn as a monotone cubic spline — uPlot's own path builder,
monotone so a smoothed line never invents a reading between two real ones.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
The portal card's "Failures 24h" had read 0 since failures_24h left
HealthSummary: that commit noted nothing read the field, which was true of this
repo and not of the portal in index/, where health is an opaque JSON blob with
no schema to catch the removal.
Restored on the connector rather than in /summary — no screen here reads it, and
it is folded from /observability/flows?hours=24, the same endpoint and window the
app's Home tile sums, so the two cannot drift apart again.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
`--xy-connectionline-stroke` was never set, so the line dragged between two
ports took xyflow's `#b1b1b7` literal in both themes and was a hair thinner
than the edge it became. It is a selection-like state, so it takes `--primary`
and the canvas's own edge width.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
--brain-decay 90s -> 150s and the wire floor 32% -> 16%, so a quiet
connection dims further and takes longer doing it. The pulse ring's inset
and band become custom properties (defaults unchanged for the editor's
card), and the neuron passes its own ring and gap so the arc runs the full
width of the gap; it moves after the disc, which otherwise paints over it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
`flow.css` set `--xy-background-pattern-dots-color`, which nothing reads:
xyflow's `.react-flow__background-pattern.dots` reads
`--xy-background-pattern-color` and only falls back to the `-dots-color-default`
literal. The grid was therefore pinned to `#91919a` in both themes — and with no
`colorMode` prop the `.react-flow.dark` default never applies either, so it read
as a light-mode grid on a dark canvas.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZeGnqVsf5VHQqvz4HdUhN
Both screens the house is looked at on are 1280x800, so that is what the three
dashboards are laid out for: twelve columns of 96px, twelve rows of 51px, and
nothing past the bottom, because a panel does not scroll.
The motors are one control each instead of three buttons. A button could only
publish; a segmented control reads back as well — so the motor writes what it
is doing to the same message the control sets, and the segment that is held is
the direction it actually went. Up, Stop, Down for the shutters; Close/Open for
the window and In/Out for the awning, which is what those two are for.
A run stopped part way now leaves the position unknown rather than claiming the
target it never reached, so the next command in either direction moves it.
The preflight gained the two checks this needed. One runs each sample shape
past the port that would receive it. The other is arithmetic: every tile inside
the panel and none on top of another — both silent failures on a screen with no
scrollbar, and both caught before anything is written.
Sizes were settled by looking. A slider needs three rows or its tick labels
fall off; a status icon needs three or it loses the word under the glyph; a
gauge in two rows has no arc worth reading, so the battery is a bar on Home and
a gauge on Energy where there is height for one. A chart spends eighty pixels
on its chrome whatever it is given, so two of them read on this panel and three
did not — the temperature history is the one that went, and `history` still
answers for it.
`capture-panels.mjs` is how that was checked: the three panels at the screen's
own pixels, in both themes, reporting whether anything spilled.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two things the installation found that the checks did not.
A script with three `from()` statements in it produces three results all
called `_result`, and InfluxDB refuses that outright — so the measurements go
into one filter and the rows come back tagged with which one they are.
And the answer a database node hands back holds a *list* of rows, which a
record may not: a record is flat scalars. It was declared one, so every chart
failed on the type check the moment a real answer arrived.
The second one is now caught before anything is pushed: the preflight runs
each sample shape past the port that would receive it, which is what turns
"expected record, got dict" from a runtime surprise into a line of output.
Also records the two engine faults this seeding session surfaced.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The Node-RED installation this replaces is 865 nodes across three tabs, and
roughly a fifth of it is unreachable — the pellet stove's controller, the
scene engine and the awning's logic were all disconnected from the heartbeat
they ran on. What is here is the intent rather than the wiring: nineteen named
flows, 109 nodes, and no heartbeat at all. A sensor value is the event.
The device layer moves with it. `actor/*` and `light/*` were never a device
interface — Node-RED subscribed to its own topics, stamped a DMX channel on
each and encoded one Art-Net universe — so those topics retire with it and the
encoders are five nodes in the `dmx` flow.
Two shared library nodes carry what every actuator needs.
`arbiter` answers the thing this design was missing: a value someone sets on a
screen is not undone by the next evaluation. A manual value wins for a hold,
the house takes over when it expires, and a schedule can force past both — so
"off at two in the morning" still means off. The control binds to the message
the arbiter writes back, so one tile shows what reached the fixture and
setting it is the override.
`motor` is why a stop is now commanded once. A rollershutter has no position
sensor, so time is the only feedback: it says how long to run and a trigger
sends the single STOP that ends it. The reference sent STOP forever.
Everything is seeded stopped, the Art-Net node does not transmit and the heat
pump does not accept commands until house.json says so.
`--dry` checks the whole set without an installation: names nothing provides,
loops, type disagreements, widgets bound to nothing, and every Python node run
once on values of the shape it declared — including whether what it returns
goes anywhere. That last one has already caught a typo that would have
published into silence.
house.json holds this installation's addresses, MAC addresses and DMX map and
is git-ignored, as the Node-RED inventory is.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Four small things, each with a device behind it.
An MQTT filter now routes what it subscribed to. `+` and `#` reached the
broker and were then looked up in an exact-match dict, so every message a
wildcard subscription received was dropped in silence.
`json_key` lifts a value out of the object a device wraps it in — Victron
publishes `{"value": 47}` on every path, which was otherwise a Python node
per port.
The trigger node learned `passthrough` and `wait_port`, because how long to
wait can be a value rather than a constant: a rollershutter takes 26 seconds
up and 28 down. A wait of zero sends nothing afterwards and still cancels
what the last message scheduled, which is how a stop is commanded once
instead of forever.
The HTTP sender takes fixed `query` parameters, so an API key is a secret
reference rather than a message on the canvas, and `send_inputs` off for a
request whose inputs are only a trigger.
Also: `delay` accepts fractional seconds, and `TZ` reaches the container, so
a cron expression means local time. Left unset it is UTC, as before.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A dashboard could only ever receive as a set of tiles. This adds the dashboard
itself as a receiver: `settings` maps a name to a value plus an optional
binding. Unbound, the setting is simply its value — a wall panel that is always
dark costs no flow. Bound, a flow drives it live and the value is the fallback.
Two settings are wired: `theme` (system/light/dark) and `locked` (read-only).
There is no schedule field on purpose — a node publishing to the bound message
on a cron is what a schedule is here, which is the point of a channel.
- `messages_for()` now walks a dashboard's bound settings as well as its
widgets' bindings. Without this a paired screen is refused its own theme
message, on the one surface the setting exists for; it bounds the socket too.
- `locked` is gated in `usePublish`, so every control inherits it, and each
control also draws itself disabled — a dead button reads as broken otherwise.
The panel surface says Read-only in the corner.
- The theme is a class on the dashboard's own surface, never the root: inside
the app shell it must not flip the chrome. `.light` gains the tokens `.dark`
already had (mirrored in the index repo) so both directions work on a subtree.
- Settings bindings are type-checked from the document alone, the rule widget
bindings follow, and mirrored on the server.
- A bound setting is drawn on the flow canvas as a dashboard-level endpoint.
- The demo's house flow now publishes `home.panel_theme`, which the demo
dashboard's theme binds to: the panel goes dark after sunset, at no tile cost.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018tULRZJUkZsw7rMJ3h4xvu
A custom hue ring — a conic gradient, not a canvas — with saturation and
brightness sliders beside or under it depending on the tile's shape, sized
for a wall panel and reachable from a keyboard. It publishes [h, s, v] by
default, which is what the reference installation's DMX encoders read, and
`format` switches that to [r, g, b] or "#rrggbb".
`usePublish` moves to its own module so a widget in a file of its own can
reach it without importing `widgets.tsx` back.
The panel now carries one wfrac node per unit, polling as well as commanding,
so a command can be checked against the unit's own answer rather than assumed.
Three things kept a panel from driving the old unit. The seeded commands catch
is off by default, which is the usual reason a fresh panel looks dead;
AIRCON_COMMANDS=1 arms it at seed time. A unit that is off names no mode, and
a flow redelivers every bound port on each run, so that rejected value blocked
every command including power-on. And the old firmware serves one connection
at a time.
The editor only ever arranged the first section, so the demo's Home
dashboard lost 16 of its 25 widgets the moment it was edited. A page's
sections are now read as one arrangement — each pushed below the one
before it — and written back as one, which is the shape a dashboard was
already heading for: one dashboard, one canvas, and the panel rail for
the several-dashboards story. The page tabs that story made dead are
gone; PageDef/SectionDef stay in the schema and a page the editor does
not show round-trips untouched.
The row height derives from the canvas as the column width already did,
so the same arrangement is the same picture on a 7" panel as on a 4K
one. The uPlot rules move beside UplotChart, where a chart on Health is
styled without a dashboard having been visited first, and the bar's
readout travels on one property instead of jumping sides at 30%.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018tULRZJUkZsw7rMJ3h4xvu
- IconPicker replaces the three icon selects (rail icon, icon-widget rule,
"Otherwise"): the glyphs in a grid, and a button that clears back to none —
which a Radix SelectItem could never offer.
- ModePicker/StylePicker drop out in favour of a shared ui/Segmented, the same
sliding-thumb shape RangePicker and the widget-side control already wear.
- PanelRail draws no scrollbars at all: hiding them also takes back the gutter
a vertical bar claimed from a column exactly as wide as its buttons, which is
what pushed a horizontal bar under them.
- The panels dialog can re-pair one screen (POST /panels/{id}/unpair) without
deleting the panel it hangs on.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018tULRZJUkZsw7rMJ3h4xvu
- Home puts the dashboards beside the flows: two equal-height columns,
capped and scrollable, most recently worked on first. Each tile is a
schematic footprint built from the stored widget placements.
- Flows and dashboards can be picked by long press or ctrl-click; the
create button becomes a trash and one dialog covers the batch.
- The offline banner is drawn on the body so it centres on the viewport,
and the live socket now releases the offline latch a stray 503 set.
- A boot spinner before React's first commit, a router pending screen for
code-split pages, and skeletons where an empty list used to flash.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018tULRZJUkZsw7rMJ3h4xvu
Three things a paired wall panel needed.
The scope check now walks the panel's widgets instead of allowing the
`/messages/` prefix wholesale: a screen may publish what its own controls and
querying charts point at, read the history of what its tiles draw, and nothing
else — the catalogue of every message in the installation included. The same
walk that already bounds its socket, so both surfaces agree.
Pending pairing codes moved out of the per-process dictionary into Redis, keyed
per code with the code's own TTL and indexed in a zset so the fifty-code cap
means the same thing to every worker. Without a Redis there is one process by
definition, and the dictionary stays.
And a per-panel nonce in the token, bumped by `POST /panels/{id}/unpair`: that
refuses the screen hanging there without touching the panel, its dashboards or
their arrangement. A save cannot write the nonce back, so a stale client cannot
undo a revocation. Only for a credential this installation signed — one the
portal minted carries no nonce and is revoked at the hub.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018tULRZJUkZsw7rMJ3h4xvu
An input widget published over HTTP and read the result back over the socket,
so between the two it drew the pre-publish value — a slider handle let go of
visibly snapped back. usePublish now holds the sent value until the echo
matches, the publish is refused, or 3 s pass, and every input widget (button,
switch, slider, input, dropdown) draws that instead of the live value.
A publish in flight pulses a primary ring over the tile: an absolutely
positioned, inert overlay, so nothing resizes or shifts. A refusal drops the
hold and toasts, naming the message. Success stays silent.
The slider also draws its own scale — min, max and a few labelled stops that
land on steps — replacing the unlabelled datalist marks that dropped out past
fifty steps.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018tULRZJUkZsw7rMJ3h4xvu
"What it can import is what the Modules screen installed" was true and
read as a wall: every example is a self-contained file, and a function
that imports half a repository looks unsupported. It is not — the
manifest is handed to `uv pip sync` verbatim, so `-e /home/you/repo`
installs the project you already have and a node becomes a three-line
wrapper over it. The code stays in your own repository, under your own
version control, importing its own siblings.
Two caveats that are easy to lose an afternoon to, so both are written
down: an editable install reaches the venv without a reinstall but not
into a worker that already imported it — the engine's workers are
long-lived, and Apply is what retires them, while an attached worker
starts a process per call and is always fresh. And the path is resolved
on whichever machine runs the node, while the manifest is committed to
the flow repository, so a laptop path means nothing in a container.
Verified rather than assumed: `uv pip sync` takes `-e`, the shim body
loads through the real `load_function`, an edit is live in a fresh
process and stale in a reused one.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The guide asked for a hostname and served the interface on
`app.${DOMAIN}` with the API on `api.${DOMAIN}` — which is not what
someone with a box in a cupboard sets up, and could not be made to work
by typing an address into a browser. `compose.lan.yml` already solves it
for the dev stack; the same file layers onto the production one, so the
guide now opens with that and keeps the hostname as the other option.
The reference gained the setting behind it. `VITE_API_URL` is a build
argument of the frontend image rather than something the stack reads,
which is worth saying once: empty means the interface addresses the API
relative to whatever origin served it, so one image answers on an
address, a hostname and an ssh tunnel alike, and no CORS list has to be
kept in step.
Verified against the production compose layered with `compose.lan.yml`:
the page, `/api` on the same port, a cross-origin-free login, the flow
websocket upgrading through nginx, and `/docs` still refused there.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
An enrolment outlives the database it was made in. Restore a backup, or
move to a different one, and the same operator is a different row —
`local_user_id` then names nobody, every portal session resolves to no
local user, and the machine answers 401 to the only route into it. That
is the lockout the welcome frame's owner exists to prevent, and it was
prevented only for the case where the row still existed.
One superuser is not a guess: it is the account enrolment would have
used, so it is adopted and written back. Several is a guess, and this
says so instead. Writing it back matters beyond this: a screen paired
through the portal borrows the same field, so it was refused for the
same reason with no way to say so.
Found on the production instance after the move to SQLite, which is
exactly this case — DEPLOY.md said to enrol again, and the machine
should not need telling.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A zensical site under docs/, served by a new `docs` compose service behind
Traefik, built with --strict in CI. Same pattern the sibling n3xd workspace
uses.
Getting started splits the way the landing page does — one path is
`pip install fluksio` and a training script, the other is a Docker stack and
an afternoon in the browser — because the two audiences will not spend the same
amount of time. Everything after that is shared: the concepts, the web
interface (app and portal), the CLI and the API, and a reference for node types,
payload types and configuration.
The three flow guides move here from the docs submodule rather than being
copied, so there is one version of them.
Styling mirrors DESIGN-GUIDELINES.md: the app's token palette remapped onto
Material's variables in both schemes, Inter, the 16px panel radius, and the one
terracotta accent spent on the facility lane of the audience split.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01M7Xv3cJEW5c8AXxn2hoojV
A `pip install` on a locked-down host — the case the CLI exists for —
may have no git, and the store shelled out to it while building the flow
repository, so the engine refused to start at all. The store is files;
git is their history. Missing it is now one warning and no commits
rather than a stack trace, which is the difference between a machine
that runs your experiments and one that does not.
Found by installing the wheels into a bare python:3.12-slim and pairing
it with the portal: `fluksio enroll` took the code, `fluksio serve`
dialled out, and the hub was proxying requests through the tunnel.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`pip install fluksio && fluksio serve` on a machine with no Docker, no
database and no configuration — which is the case this is for: a node on
a cluster where ports cannot be opened. It makes its data directory, its
key and an admin account, prints the password once, and serves. Pairing
is `fluksio enroll <code> --portal …`, doing what the Settings screen
does through the same function, before the engine starts and without one
running — a machine nobody can route to has no browser pointed at it
either. The portal serves the dashboard, so nothing is served here.
Two things had to give way. `fastapi[standard]` pulls a cloud CLI that
wants sentry-sdk 2.x while we pinned below it — no pip resolution
existed, so the pin is lifted, which the comment beside it had been
waiting for and which also lets the Python cap go. And `uv` is now a
dependency rather than something to find on PATH: the Modules screen is
how a data scientist installs torch, and it was quietly falling back to
the engine's own interpreter.
The CLI imports nothing from the engine before it has set DATA_DIR — the
settings are built on the first import of core.config, and reaching it
early put the database in the working directory. There is a test for
that now, because the failure is silent.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
One process owns this database — the image has run a single uvicorn
worker for that reason since the four-engines bug — so a file beside the
flows is the honest shape for it, and it is what lets `fluksio serve`
need no infrastructure at all. Live values, node execution and the work
queue never came here anyway; what does is a rollup a minute at a time,
a row per cascade and the run history, and WAL keeps the readers going
while that one writer works.
DATA_DIR is now the one setting that moves everything an installation
keeps; the rest derive from it and the images still spell theirs out.
The schema is prepared in-process at startup, so the prestart service is
gone, and the ten Postgres-only revisions collapse into one portable
baseline.
Three things only worked because psycopg was casting for us: a token's
subject arriving as a string where the column is a UUID, `greatest`, and
`date_bin`. The timestamps needed a column type of their own — SQLite
stores no offset, and a naive datetime read back either raises against an
aware `now` or serialises as local time.
Postgres stays in the stack only for Umami, behind the analytics profile.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A cluster or GPU host installs `pip install fluksio-worker` and gets the
agent and the runner, not psycopg, numpy and the MCP SDK. The engine
depends on it as a workspace member, so the file it launches node code
with is the same file a remote worker runs — which is what keeps a node
unable to tell the difference. Copying the two files by hand still works.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A wheel whose top-level module is `app` collides with anything else in a
user's venv, so the package that is about to be published takes the name
it is published under. Only the Python package moves; the repo, the
Docker WORKDIR and the compose project keep theirs.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`cut -d: -f2` inside a plain YAML scalar reads as a mapping key, so
test-backend.yml was not valid YAML and would have been rejected the first time
a runner picked it up. A block scalar takes the whole command as text.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The sibling index frontend now prerenders its public pages, where the initial
state read of localStorage and the matchMedia probe both run with no browser
present. The design contract keeps this file byte-identical across the two
repos, so the guards land here in the same change.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A checking lint reports what a writing one silently reformatted: after any test
run, `playwright/.auth/user.json` and `test-results/.last-run.json` are written
by Playwright and failed `make lint`. Both are gitignored, and `playwright-report`
was already excluded beside them.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uq8mtNb97A7praJLyeEYgs
Twenty-three items are done: the dashboard draft/publish split and panel
freshness, the canvas clipping warning, the inject node's per-port payloads,
the sticky failure and the brain click-through, and the lint and DOMAIN
tooling fixes.
The ChartWidget entry is rewritten rather than closed. Its ring buffer would
only remove the cheap half of the cost, and it is the shape most likely to
tempt someone into re-adding the dependency array that made a full window of
chart data go stale.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uq8mtNb97A7praJLyeEYgs
`make lint-frontend` was `biome check --write --unsafe ./` — a lint target that
reformatted the whole tree rather than checking it, which is why every parallel
change in this repo has had to work around it. `lint` checks now and a new
`format` writes. The pre-commit hook and CI needed no edit at all: both call
`bun run lint`, so they became checks the moment its meaning changed.
`app/Makefile` assigned DOMAIN from .env, and a plain assignment beats an
inherited environment variable and is not exported — so `cd app && make
dev-local` served localhost while the same checkout's tests targeted the
configured domain. `export DOMAIN ?=` gives the lattice that was intended:
command line, then environment, then .env.
Alongside: the backend's htmlcov bind mount created that directory as root, so
`make test-backend` died on the coverage step after every test had passed,
which reads like a test failure and is not one. The alerts screen's copy of
ALERTING_EVENTS is now checked by a test rather than trusted. And the shard
comment claimed two spec files where there are nine.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uq8mtNb97A7praJLyeEYgs
A node's error cleared the moment it ran again, so a failure that genuinely
fired an alert could leave no trace on the canvas by the time anyone looked.
The engine records it now — on the node's status, so it survives a reload and
every client agrees — and reading the traceback is what clears it. The seam is
the event bus, which is where every failing path already meets: a queued live
run, an explicit run, a preview, and a single triggered node all publish
`node_error`, while the controller's own observer would have seen only one of
them.
That was half the confusion. The other half: clicking a failed neuron on Home
often landed on a flow where everything looked fine. Nodes merge into one
neuron by instance key — every InfluxDB node pointing at the same bucket is one
neuron — and the click went to whichever flow contributed a member first, not
the one that failed. It now goes to the failing member and selects it, and the
canvas marks a failing node rather than leaving it to the dot alone.
The inject node emitted one payload to every port it declared, whatever their
types, so an inject on a bool port carrying the text "true" raised at publish
time. Each port gets its own field now, typed and parsed by that port's dtype,
and remembers what it last sent. A port that is renamed carries its value with
it; one that is removed takes its value with it. An inject written before this
keeps emitting exactly what it did.
The derived-cron chip also appeared on the delay node, where `interval` is a
rate limit and a schedule derived from it means nothing.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uq8mtNb97A7praJLyeEYgs
Shrinking a dashboard's canvas silently clipped whatever now fell past the
bottom edge: `maxRows` constrains a drag and nothing else, so a stored placement
is corrected against the column count alone. Nothing warned, and nothing offered
a way out.
The remedy is a notice rather than a reflow, because the canvas height is
written on every keystroke — typing 400 passes through 4 and 40, and anything
that moved widgets would flatten the arrangement while the number was still
being typed. The notice carries the reflow as its one button, and packing is
sideways because the grid already compacts vertically: nothing below the canvas
has room above it.
Dropping a widget also selected it, which opened its panel, which rescaled the
canvas under the pointer. The drag handle was simply missing from the selector
that already exempts the resize handle — which is why resizing never had this
problem.
Also: an icon rule's caption could only be set through the API, and the panel
rail drew two letters where a dashboard can now carry a lucide icon.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uq8mtNb97A7praJLyeEYgs
A dashboard went live the moment it was created — an empty document straight to
the panels — while a new flow starts as a draft. It now works the way flows do:
published means `dashboard.json` exists, so every dashboard on every running
installation is already published and nothing needs migrating. Only the ones
created from here on start as drafts.
Mirroring FlowStore turned up a latent 500: discarding the draft of a dashboard
that had never been published unlinked its only file, and the read that followed
raised out of a 200 handler. It answers 400 now, the way a flow does.
Publishing all of them was 2N requests, because a publish has to name the
version it expects and the summaries did not carry one. They do now — and so do
the flow summaries, which had the same defect nobody had written down.
A panel had no way to hear about any of this. A publish, or a change to which
dashboards a panel carries, now puts one event on the bus and the screen
refetches what changed: no reload, so a wall display never blanks or asks for
its credential again. The subtle half is that a socket's message allowlist was
computed once at handshake — a reassigned panel would have fetched its new
document and then shown tiles that never updated.
The panels dialog logged non-superusers out. Every write in it needs a
superuser, not only the checkboxes the report mentioned, so the dialog is
read-only for everyone else. The logout itself was `main.tsx` treating 403 as a
dead session, against the contract deps.py spells out: only a 401 ends a
session, and a 403 now says so rather than silently signing someone out.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uq8mtNb97A7praJLyeEYgs
An installation enrolled before per-user mapping has nobody mapped, and
fail-closed means its owner is refused. Re-enrolling fixes it and can only
be done from the machine's own network, which is the wrong thing to require
of a machine whose only route in is the portal.
The hub names the owner in the handshake now, and this takes it: if the
enrolling account has no portal identity and nobody else holds that one, it
is written once and every later attach is a no-op. A mapping somebody else
holds is never moved - enrolment was told who that is, and this is only a
repair. A failure to write one does not drop the link.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Remote access used to collapse every portal session onto the account that
performed the enrolment. That was the only thing it could do while nothing
here knew who was at the other end, and it is why letting a second person
in meant handing them the first one's account.
`user.portal_sub` is where a portal identity meets a local one: set for the
enrolling superuser at enrolment, and for each person a superuser admits
afterwards through Settings -> Remote access -> Add remote user. The code
they type comes from the newcomer's own portal account, and it is redeemed
against the hub with this installation's tunnel credential rather than with
a portal session, so being let in is not itself the power to let others in.
The account created is never a superuser, which closes the same door from
this side.
A proxy token now resolves through that mapping and nowhere else. An
identity nobody mapped resolves to no user rather than falling back on the
enroller, so deleting the local row under Admin -> Users is the whole of
the revocation: it bites on a credential already in flight, and it does not
wait on the portal being reachable to be told. Telling the portal is best
effort for exactly that reason.
The cost is stated where it lands, in DEPLOY.md: an installation enrolled
before this has no mapping, so its owner reconnects once with a fresh code.
Panels and the health summary still act as the enrolling account - neither
of them is a person, and neither gained a way to name one.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Twenty-one items are done: the canvas viewport and logs panel, the widget
overflow and stacking work, the backup and queue-namespace changes, and the
observability fold.
What replaced them is mostly the residue of doing them — a legend styled in the
wrong CSS chunk, two overflow floors that nothing would notice regressing, and
artifact blobs that outlive the rows pointing at them.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uq8mtNb97A7praJLyeEYgs
Three things moved the viewport independently — the shape-fit effect, focusNode,
and React Flow's own fitView prop — so a fourth for "centre the node I just
selected" would have been a fourth party to the argument. There is one effect
now, and which branch it takes is decided by what changed rather than by what is
true: selecting a node brings that node into the lane the panel leaves, and
every other change — new wiring, a new endpoint, a panel opening — re-fits the
whole flow into the same lane. A selection centres once, so the port edits that
follow re-fit around it, which is what makes a new edge's far end visible.
The refit triggers on the edge count, not the bindings key: that key changes on
every keystroke in a message-name field, and refitting per character is not what
"an edge was created" means.
renderedNodes overwrote xyflow's own `selected` flag, so a box-selection of
several nodes was invisible even though delete and copy acted on all of them.
The logs panel was a popover anchored on its own button, which is why it sat off
centre, hugged the button and closed on any outside click. It is a plain surface
above the dock now, and the button is stateful. Escape still closes it.
Expanding a node's editor gives the panel the whole inset and puts the code on
the left with the settings beside it, while the toolbar and the flow name
translate off screen. Narrowing the window past `md` gives the room back — the
sheet it becomes has no second column to hold.
The zoom buttons are gone: there is a mouse, or there is a pinch.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uq8mtNb97A7praJLyeEYgs
A bar drew its nested reading on top of the outer one in --chart-5, which
measures 2.53:1 against --primary and lost the 3:1 guideline for non-text. The
readings now partition the fill end to end, up to three of them, in a token of
their own: --primary-nested, the primary hue a few steps deeper, 3.14:1 light
and 3.12:1 dark. It cannot also clear 3:1 against --muted — in dark those two
are 5.82:1 apart and a colour 3:1 from both would need a 9:1 gap — so a segment
is drawn inside a gutter of outer fill rather than ever bordering the track,
which is what separates neighbours too, and what caps the count at three. A
nested value larger than its outer used to spill onto the track; it is clamped.
`inner` still reads as a single binding, so no dashboard needs migrating.
On a phone, .widget-grid took its width from the widest thing any widget held —
a truncating flex item still offers its whole unwrapped line as a min-content
contribution — and a handful of widgets had no floor of their own: the uPlot
legend is a table, a fieldset carries min-inline-size: min-content from the UA
sheet, and buttons are whitespace-nowrap. Each is capped now. A widget's body
scrolls rather than clipping, so long text stops painting over the title.
Gauges and bars move between readings instead of jumping, and a segmented
control slides one thumb rather than recolouring cells. The gauge arc is drawn
whole and revealed by its dash, because `d` cannot be transitioned.
UplotChart pushed new readings only when the point count changed, so once a
rolling window was full a refetch left the old values on screen.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uq8mtNb97A7praJLyeEYgs
/observability/timeseries and /flows read every metric_minute row in the window
and folded them in Python, so the 7d preset pulled a week of rows on each 30 s
poll. date_bin() does the binning now — the row count drops to the slices asked
for, and to flows × 60 for the sparklines. A window of zero hours used to divide
by nothing and answer 500; windows are clamped to an hour at the low end and to
the retention period at the high end, past which there is nothing to find.
/observability/runs returns {data, count} rather than a bare list, so a minute
busier than the 200-row cap says so instead of quietly showing its newest 200.
The count is only queried when the page comes back full, which keeps the poll
from handing back what the fold just saved.
failures_24h leaves the summary — the Home tile counts errors over the selected
window from the rollups, and nothing had read the field since.
Deleting a flow now takes its Run rows and their nodes, metrics and artifacts
with it. This lives in the route rather than in forget_flow because renaming a
flow calls that too, and a rename must keep its history. The observability
rollups stay: they are the record of what ran, and retention already prunes them.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uq8mtNb97A7praJLyeEYgs
The queue shared the `pipeline:` prefix with flow state, so `RedisState.clear()`
could DEL the queue stream and `keys()` enumerated queue entries — only callers
filtering `__`-prefixed names kept it safe. It moves to `queue:` without a
migration: whatever is in flight at the upgrade is dropped once, documented in
DEPLOY.md rather than papered over.
Alongside it: `pool_pre_ping`, so a connection idle across a Postgres restart
costs a round trip instead of a failed request; the test suite pins
ENVIRONMENT=local and DOMAIN=localhost itself rather than inheriting a
deployment's .env; and `depth` leaves the queue stats, where it reported the
capped journal length as if it were a backlog.
ALERTS_FILE and PANELS_FILE now point at /data. They defaulted to a path on no
volume, so alert routing and every wall-panel pairing were living in the
container's writable layer and vanishing on each rebuild. Carrying the existing
files across is a manual step; DEPLOY.md has it.
development.md was still the upstream template — compose.override.yml,
localhost.tiangolo.com, `docker compose watch` as the dev flow — and said
nothing about the Playwright suite. Rewritten against what the Makefiles
actually do. deployment.md was template text too, duplicating the root
DEPLOY.md, and is gone.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uq8mtNb97A7praJLyeEYgs
Three things the security pass on the portal pairing turned up. The first two
were already true of a screen on the local network; what changed is that a
panel credential is now presentable from the internet, which is what makes
them worth closing rather than recording.
The artifact endpoint authenticates for itself, because a worker's credential
has to open it and that token is no use anywhere else. It resolved the caller
without handing over the request, so the one credential that is scoped by
route was judged by no route at all — a panel could read and write the store
as whoever approved it. It passes the request it already holds now.
The websocket has no route to judge either, and there the bound has to be on
what is sent: a panel is given the values its own dashboards draw and nothing
else — no node status, no logs, no shape of the graph. The keys stay in the
message, emptied, because a screen on a wall runs the bundle it was paired
with. `messages_for` reads that set off the published dashboards, and is the
walk the `/messages/` allowlist has wanted for a while.
And locality is no longer a header anyone can type. The marker the connector
stamps is a value minted per process, so reaching this API directly cannot buy
a device the credential meant for one that cannot reach it at all.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017F9RnYCJgASuBTcAjxmnsp
A screen somewhere this installation is not reachable from asks the portal for
a code instead, and the portal mints its credential — because a token signed
here is one such a device could never present.
Where it was minted changes nothing about what it may do. The panel gate moved
off the branch that decodes a local panel token and onto whatever claims name
a panel, so the portal's and this installation's are bounded by the same check
against the same panel's dashboards. A token of that scope naming no panel is
refused rather than left holding the account it borrows.
The connector marks what arrives on its socket, since that is the only thing
that makes it true, and the approval screen now names what is holding a code —
approving adopts whatever answers, so it is worth a look first.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017F9RnYCJgASuBTcAjxmnsp
Four controls and the unit's own answer beside them. Two catches, both on:
the flow is seeded stopped and the node's commands setting is off. The initial
values are read off the unit when the script runs, so starting the flow asks
for what it was already doing rather than commanding it to something else.
Two things stopped the engine commanding this house. ConnectorNode hardwired
its node function to a no-op, so an input message reaching a connector was
discarded and Art-Net's packet builder was unreachable; write() now carries
the input ports, which is additive so the contract version holds. And the MQTT
publisher JSON-encoded every payload, so "ON" went on the wire quoted and the
devices on a shared broker, which speak bare values, ignored it.
seed_house_control.py is the rig: a flow that drives the washing machine plug,
a dimmer and a colour fixture over MQTT, carries the same two as DMX on an
Art-Net node with transmit still off, and a dashboard to drive it by hand.
`tests/utils/api.ts` took the API origin from `VITE_API_URL`, which
`tests/config.ts` loads out of `app/.env`. In a checkout configured for a
deployment that names the deployment — so the browser went to the local stack
while every setup and teardown call, `deleteAll` included, went to the live
one. `privateApi.ts` had the same reading, and it creates users.
Both origins now come from one place: `PLAYWRIGHT_BASE_URL`, with the API
derived from it (`app.<domain>` → `api.<domain>`) or named outright by
`PLAYWRIGHT_API_URL`, which is what CI and the compose service set. Nothing in
the suite reads `VITE_API_URL` any more.
Belt and braces, since a stack served under a real domain answers to the same
names its production instance does: a global setup resolves both origins and
refuses anything that is not loopback or a private range, before a test runs.
`PLAYWRIGHT_ALLOW_PUBLIC=1` says you meant it.
`make test-frontend` is now that safe run — the Playwright image on the proxy
network with both names mapped onto Traefik by address, as the host user so it
does not leave root-owned results behind.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NUb8YpL2s3gmN9WTACTt4q
A sweep against the code. Removed as resolved: the worker pool's `_running`
is keyed by (run, node) now, `WorkItem.kind` documents only the three kinds
that exist, `routeTree.gen.ts` matches its generator, and the host's Node is
well past the version the Vite 7 note was about. Removed as disproved: the
collector's redelivery branch is reachable — Redis reclaims an entry this
process is still running.
Corrected: the demo places six of the fifteen node types, not thirteen — the
count was the widget types. Narrowed to what is left: the header/config sync
now covers settings but only on an untouched scaffold and only one way; a
declared flow input is what the dashboard-input node asked for; the flow
boundary already refits the canvas where node ports do not.
The Playwright hazard is not resolved — both names still resolve to
production — and the fix as written is incomplete: `VITE_API_URL` has to be
overridden to `http://` or every setup call fails the TLS handshake.
Also: one `asText` instead of two, and the queue docstring no longer refers
to a node item that never existed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NUb8YpL2s3gmN9WTACTt4q
A flow's inputs are the messages it takes from outside — a dashboard control,
a run, the API — and its outputs are what a batch run reports. Both existed in
the document and in the engine, and neither had any UI: the values looked
hard-coded on the canvas and the Run button always used the declared defaults.
The canvas now draws each as a labelled endpoint, the way it already draws a
dashboard tile or another flow, skipping an input something else already
accounts for. The flow panel edits them — mode, name, type, starting value,
and for a live flow the value it currently holds with a way to put a new one
in. Pressing Run on a batch flow asks for its parameters first.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NUb8YpL2s3gmN9WTACTt4q
A python node's settings are constants of its own function, so they are passed
the way its ports are: by name. The controller binds them to the compiled
function, the `params` field is gone from the worker and remote protocols, and
a setting sharing a port's name is reported as a node error rather than
shadowing it. The panel's scaffold follows suit and keeps the header in step
with both ports and settings.
The demo's `pace` moves from a flow input to a setting of the training node,
which is what it always was.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NUb8YpL2s3gmN9WTACTt4q
stroblmeandClaude Opus 5