The canvas placed nodes with dagre, which orders nodes but knows nothing
about ports: an edge leaves a particular handle and lands on a particular
one, so two producers feeding a consumer in the other order crossed just
before they landed however well the nodes were arranged. The graph was not
a multigraph either, so two messages between one pair of nodes collapsed
into one before crossings were counted, and the endpoint labels were laid
out as 220-wide cards with every handle stacked on a single pixel.
ELK's layered algorithm replaces it. Every declared port is handed to it as
a fixed point on the node's rim, at the fraction `portFraction` puts the
handle at — which the node components now render from the same function —
so what the crossing count is minimised over is what ends up on the screen.
Measured on this installation's flows, with the labels included: `home`
goes from 51 crossings to 6 across and 70 to 29 down, `demo_training` 27 to
5 and 59 to 32, and every graph is the same size or smaller.
The engine answers asynchronously and is a chunk of its own, so positions
became state: the canvas draws nothing until the first layout lands, and an
edit keeps the arrangement it had rather than flashing through the corner.
The entry chunk is untouched and the flow route's own chunk came down 167
to 121 kB, since dagre used to be in it.
Two things the crossings made worse come with it. A hovered edge resolves
to full strength so one line can be followed through a busy rank, and a
feedback edge — one whose target the layout put behind its source — swings
out into a lane beside the graph instead of being drawn through everything
between its ends, which is what "House history" was reported for.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01U6vyQvfX249hsv9YPmsYYg
`make test-frontend` built PLAYWRIGHT_BASE_URL from DOMAIN in .env, which in a
checkout configured for a deployment is that deployment's domain — so the suite
that creates and deletes flows, dashboards and users was pointed at
app.fluksio.com, held local only by --add-host and tests/guard.ts.
The hostname now comes off the running stack (the frontend container's own
Traefik rule), so a name no local container answers to cannot be reached at
all, and the local targets default to *.localhost instead of reading .env.
Target-specific on purpose: an exported DOMAIN outranks --env-file in compose
interpolation and would put the production targets on localhost.
`rebuild-frontend` replaces the raw compose line CLAUDE.md spelled out, taking
the same domain so the baked VITE_API_URL cannot disagree with what Traefik
serves. Also: a coverage HTML report that cannot be written no longer fails
test-backend after a green suite, and both artifact actions in playwright.yml
drop to @v3, which is the only version without the github.com-only guard that
failed every run.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Dragging the brightness slider on the wall panel set the handle and
published nothing; tapping a point on the track worked. The control is a
native range input laid transparent over the drawn track, and only the
release publishes — but the input never said the drag was its own. On a
touch panel a sideways swipe is a pan, or a back-navigation, so the
browser took the pointer over mid-drag and ended it in `pointercancel`.
The value had followed the finger and was never sent.
`touch-action: none`, as the colour disk beside it has always had. The
release also answers `pointercancel` and `lostpointercapture` now, which
covers a mouse let go outside the input and leaves no way for a draft to
sit there unpublished.
Two checks: a drag across the track reaches the engine, and the input
still owns its gesture. The second fails against a build without the
CSS, which is what makes it worth having.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012n6CehUsHYYUXJ48rxaD18
A node runs when something it reads was published, not merely when something
upstream of it ran — worth saying now that it is true, since it is what a
reader relies on when a rate-limited port goes quiet. Plus `FLOW_MAX_CASCADES`
in the engine's configuration table.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BpfSinyCBfjuieikyfMPbf
And stop its drain loop asking for the whole of `stats()` every twenty
milliseconds — four round trips a poll, competing with the engine for the
connection it is measuring.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BpfSinyCBfjuieikyfMPbf
The engine was I/O-bound on its own state backend. `RedisState.lock()` is one
key — `pipeline:_lock` — for the whole process, taken five times a message at
two round trips each, and every cascade and every node read queued behind it.
Inside it, reading a node's inputs was three round trips per input (an EXISTS
for `in`, then EXISTS and GET for the value), writing was two updates that a
single transaction already gives, and the version counters went one INCR at a
time.
Replaced with the atomic command that was always available: `get_present` is
one MGET and tells a missing key from one holding null, so the lock it used to
be read under bought nothing; value and timestamp land in one `update`, which
is a MULTI/EXEC; `increment_multi` pipelines the counters. `values()` — what
every websocket snapshot calls — is two reads whatever the message count
instead of two per message.
Beside that: every webhook did its blocking XADD on the asyncio event loop
(MQTT already used `to_thread`); the per-execution `NodeOutcome` was built and
validated even with no run watching; `_minute` built a tz-aware datetime per
event on the loop thread to key a dict, and now keys on an int; `move_due`
promoted delayed items one round trip each, every second; `FLOW_MAX_CASCADES`
makes the in-flight ceiling a setting rather than a constant.
`orjson` replaces stdlib json where a message pays for it — state, the
journal, the engine side of the worker pipe. `fluksio-worker` stays
dependency-free, and the run-cache digest stays on stdlib so no stored key is
invalidated. A non-finite number now stores as `null` rather than the bare
`NaN` that was never JSON.
Measured with `scripts/bench_engine.py` against a real Redis, 200 messages:
a five-node chain went from 43.9 to 103.1 msg/s with p50 latency 2110ms →
782ms and p95 3913ms → 1439ms; one source into twenty consumers went from 5.4
to 33.7 msg/s. In memory, twenty consumers went from 187 to 448 msg/s.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BpfSinyCBfjuieikyfMPbf
The limit was applied in `apply_outputs`, which the executor reaches after the
item is off the queue — so a subscriber told to publish every 15s still cost a
queue entry, a `cascade_started`, a run record and a walk of everything
reachable from it per inbound message. Seven relay nodes behind one inverter
ran 192 times a minute to publish six.
Two halves, matching the two shapes it takes:
`trigger()` now keeps a value whose every port is inside its window and
journals nothing at all. The window split came out of `_throttled` as a
read-only `_window_split`, so the question is asked the same way in both
places and the exact split is still made once, at claim time.
A cascade carries the names it actually published, and the wave runs only the
nodes something in that set feeds. A node whose triggering inputs were all
held back is completed without running, which frees its own consumers to be
judged the same way — the case where a node re-published 619 messages a minute
off inputs that changed six times. Redeliveries and emissions carry no such
set and still walk everything, since one has a half-finished wave to finish
and the other is the value already being in state.
Skipping a node can make one ready that the scheduling pass has already walked
past, so `submit_ready` runs to a fixpoint. That also closes the same latent
hole on the replay path, where a done-marker skip could strand a join with no
future outstanding to come back for it.
Measured with the new `scripts/bench_engine.py`, 500 messages through the
house's shape: a limited source went from 500 cascades / 3500 node runs /
5009 events to 1 / 7 / 19, publishing the same 8 values; an unlimited source
into limited relays took the node reading them from 500 runs to 1.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BpfSinyCBfjuieikyfMPbf
`RedisWorkQueue.stats` read XPENDING, which counts entries delivered to a
consumer and not yet acknowledged — work in progress. Entries sitting in the
stream undelivered were counted nowhere, so an engine hours behind reported
itself idle: on the house, `pending: 4` while the group's lag was 1554.
The group's own `lag` is the missing number. `backlog` now carries it on both
queues (`len(_items)` in memory), leads the health tile, and a sustained one
publishes `engine_degraded` from the timer thread — named with the flow most
of the waiting work belongs to, sampled from the undelivered tail, since that
is the actionable half. It is a summary problem rather than a /utils/health
503: a backlog should not restart the container.
Also drops the keyspace `scan_iter` `stats()` did per poll to count parked
items — it walked every state and idempotency key twice per ten seconds — for
a set the park/unpark path maintains.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BpfSinyCBfjuieikyfMPbf
Clearing a node's last error on the engine published nothing, so a second
browser kept the marker until its next snapshot. One event carries the
qualified node; the receiving client drops the marker without refetching.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Gf7WaExcJ9bs3kfJXB3nK
Both were a 2px inset --primary ring, so a widget mid-publish and one
picked in the editor looked the same. Status has three colours and none
of them means "picked", so the stroke tells them apart instead.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Gf7WaExcJ9bs3kfJXB3nK
The lock notice was on `PanelSurface` alone, so /dashboards/{name} showed
disabled controls with nothing naming the state. Same pill, not a second
one; edit mode keeps its controls live, so it stays out of there.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Gf7WaExcJ9bs3kfJXB3nK
The popover showed a non-numeric value twice: only numbers are recorded,
so MessageSparkline fell through to a collapsed ValuePreview and the
inspector rendered its own expanded one below it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Gf7WaExcJ9bs3kfJXB3nK
`POST /runs/flows/{name}` hardcoded `cause: "api"`, so every row in the
history claimed the same origin. The body now carries an optional `cause`,
closed to the values the column knows — the dashboard sends nothing and stays
"api", `fluksio run` says "cli", and the SDK client says "sdk".
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Gf7WaExcJ9bs3kfJXB3nK
The suite runs in a random order and a run left marked running by an
earlier module lands in the same list, so asserting on data[0] failed on
whichever seed put that module first.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Gf7WaExcJ9bs3kfJXB3nK
`ADVISORY_ISSUES` moves next to `ValidationIssue` in pipeline.py, and the
model derives an `advisory` flag from its own code, so the distinction the
engine already made ships to the client instead of being re-guessed there.
The dock keeps its summary in `--destructive` only when a real fault is
among the issues and paints an advisory row `--muted-foreground`; the
canvas leaves advisories off a node's dot and border entirely, since node
status has three colours and no warning tier.
biome checks the generated `openapi.json`, which nothing formats since the
SDK script dropped its format pass — ignore it like the other generated
files.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Gf7WaExcJ9bs3kfJXB3nK
Saving panels replaces the whole list, so the spec's setup deleted every
panel it had not created — and with it the credential of the screen on
it. It now appends to the stored list and puts the original back in
teardown, which also removes its own panel.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Gf7WaExcJ9bs3kfJXB3nK
A dashboard whose canvas matches the screen came out at 95% of it: the
panel and view routes inset the canvas by a further 16px, on top of the
margin the grid already keeps from its own edges. Two margins, one of
them costing the scale. The padding now applies only to the stacked
phone layout, which has no canvas to shrink.
A slider's tick labels are the third row of a control that has three,
and the first thing a short tile runs out of; `ticks: false` drops them
and keeps the reading and the track. The house's two sliders set it.
The 1024x600 preset is named for what it is rather than for a diagonal
it may not have — the same resolution is sold as a seven-inch panel and
as a ten-inch one.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012n6CehUsHYYUXJ48rxaD18
A cache hit still replays no emissions — those values were the story of an
execution that is not happening — but the run they were recorded in is now
written on the row (`run_node.cached_from`), and the metrics endpoints read the
series back from there. So a reused run answers `run.metrics("train.loss")`
with the same points the run that trained did, rather than looking like a run
that produced no numbers at all. Pointed at rather than copied: a sweep of 500
reusing one frozen node would otherwise duplicate its curve 500 times.
That needed the cross-flow restore fixed first. The cache key has no flow in
it while the stored outputs are named for the flow that produced them, so
`quick.prepare` getting a hit from `train` wrote `train.dataset` into `quick`'s
state and the next node was called without its argument. One rule now covers
both halves: `requalify` reads a name owned by one flow as the same name in
another, applied to the restored outputs, to the node id behind the pointer,
and to the series names on the way out. Reuse across flows is kept.
Also: `@run:<id>.<output>` and a bare `sha256:` digest resolve on every input,
not only artifacts. Chaining a run's json config into the next one from a shell
meant pasting the whole object inline, and the CLI could not even send the
spelling — `_coerce` died in `json.loads` before the engine saw it. Both
spellings are reserved on every input now, `str` included, and `_from_run`
returns whatever the run's result holds rather than only a reference.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Dp9L6gakMVro1K2C5zdtBE
Four things from a testing pass.
`fluksio serve` printed its own lines through the root logger, which has no
handler and falls back to `INFO:fluksio.cloud.connector:...` — beside uvicorn's
aligned output it reads like something went wrong. The engine's loggers and
alembic's now use uvicorn's own handler. Named rather than configuring the
root: httpx logs every portal call at INFO and none of that is printed today.
`fluksio enroll` writes its config from another process, so an engine already
serving never learned it had been paired. It now looks for one every few
seconds and dials when it appears. `load()` rather than `exists()`, or a file
that does not parse would be restarted forever.
`fluksio status` says where the installation stands with its portal — never
paired, linked, or paired and unreachable, which is the one worth acting on.
`--seed` and `--timeout` had no help text at all. Both say what they are for
now, and the docs say what a seed is actually for: recorded on the run, part of
its input digest, and passed to an input named `seed` when the flow declares
one, so the number a run is labelled with is the one the code drew from.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019V5bsYGNxcgPs4xXmTPx69
`runs --local` already exists for the reason status needs it: the flows and the
history are in this directory, and looking at them should not need a server.
Watching still does — an in-process engine is the command itself, so nothing
could change under it — and says so rather than sitting on a still frame.
Found setting up a sandbox that drives the SDK entirely through --local.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019V5bsYGNxcgPs4xXmTPx69
Two halves of the same gap: the CLI could start work but not show you any.
`fluksio status` draws the home screen's top half in a terminal — health and
what is wrong with it, every flow with its state and node count, and the
recent runs and failures under them. `--watch` keeps it there. Rich does the
drawing; it was already installed under fastapi's own CLI, and is named now
because a command depends on it.
`fluksio run` with no parameters at a terminal asks for them, one line per
declared input with its declared value in brackets — so Enter through the lot
is what running the defaults looks like, and an artifact input takes the
`@run:` spelling the engine now resolves. A scripted run is untouched: passing
any parameter, or piping the command, skips the questions, as does --defaults.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019V5bsYGNxcgPs4xXmTPx69
Four things the python SDK turned up, each fixed where every client sees it.
A key no port declares is now an error rather than a silent drop, on the
return, the yield and the emit alike — the contract the docs already stated.
The SDK reads literal yields at sync time, so a typo fails before anything
runs, and an emission of one fails the call rather than being logged where
nobody looks.
NaN and infinity are refused at the port. JSON cannot spell either, so one
that travelled came back as a 500, a socket frame that stopped the canvas, or
a metric batch the database dropped whole.
An artifact input takes `@run:<id>.<output>` or a bare digest, resolved on the
engine — so the CLI, the run dialog and a python caller mean the same thing,
and a sweep can pass one at all.
Node timeouts are off by default. The clock measured silence, which a training
node is full of, and remote workers had already stopped enforcing it — their
heartbeat reset it. Now a heartbeat proves the agent rather than the node,
ninety seconds of nothing fails the call either way, and the engine touches
work it is still running so a long node is not redelivered at sixty seconds.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019V5bsYGNxcgPs4xXmTPx69
compose.dev.yml mounts the source over the image's own copy so `--reload`
means something in a detached local stack, and mounts the Playwright report
out to the host. Under act_runner the daemon is a sibling of the job rather
than its host: it resolves those paths against its own filesystem, finds
nothing, and mounts empty directories. So the backend reported "Path does not
exist fluksio/main.py" for a file the image plainly ships, and every shard
uploaded nothing — "No files were found with the provided path:
frontend/blob-report".
compose.ci.yml now puts backend's volumes back to what compose.yml declares
and drops Playwright's entirely; the shard copies its report out of the
container before the teardown instead.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A code node in a batch run is now fingerprinted by its source, its raw
settings and the values it reads — an artifact input counting as its digest,
which is what the content addressing was always for. A run that finds the key
restores what the earlier one returned and skips the node, recorded as
`cached`. The run history is the cache: `run_node.outputs` beside the
`cache_key` the schema already had, no second store. On for code nodes, never
for the built-in and connector types that have side effects; off per node with
`@node(cache=False)` and per run with `--no-cache`.
Emissions are not replayed on a hit, so a cached training node returns its
result without redrawing its curve. Recorded in NOTEPAD.md with the two other
deliberate limits.
`fluksio run --local` boots the real app in the command's own process and
drives it through its ASGI interface behind the ordinary client, so a run no
longer needs a `serve` terminal beside it — same data directory, same history,
and the cache carries between the two. It always waits, because the engine it
starts lives exactly as long as the command.
Also: `fluksio sweep --param lr=0.1,0.01` for the product of the lists,
`run --follow` for a run's numbers as they arrive, Ctrl-C cancelling a waited
run rather than abandoning it, coloured statuses on a terminal, and `name`
made optional on the metrics endpoint so a follower can ask for every series.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Four things had drifted. It told you to run `fluksio login` two lines above
saying `serve` signs you in — one of the two had to go, and it is the login.
It ran `fluksio sync` before `fluksio run`, which `run` now does itself. The
pipeline snippet passed `prepare` and `evaluate` to `Flow` without importing
them, so copying it got a NameError on the one example that matters. And the
site link was relative, which resolves to nothing on the page this file exists
to be — pypi.org.
Added one sentence, on installing into the environment you already work in,
because it is the reason a reader's own imports keep working and there is
nowhere else on this page they would learn it.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012ue1tkFWB1bcGy3aWhCKpU
**The portal link puts itself back up.** It already retried a connection that
raised, but a session that ended *cleanly* — a portal restarting, a proxy
closing an idle socket — returned normally and went straight back round the
loop with no wait at all, so an engine could spin against a portal that was
merely saying goodbye politely. Every ending now reconnects on a delay, and
the delay turns on whether the attempt got as far as attaching: one that stood
up and dropped is a network event and retries at once, one that never stood up
waits longer each time. Jittered, so a portal coming back is not met by every
installation it serves in the same instant. Ping timeouts are named rather
than defaulted, since they are what bounds how long a suspended laptop's dead
socket looks alive, and the keepalive task is awaited so the reason a link
went reaches the log instead of the garbage collector.
**`fluksio enroll <code>`** is the whole command now; hub.fluksio.com is the
default and `--portal` names another. The one command run before anything
works should not need two flags.
**`fluksio run` syncs first.** The reason a run exists is usually the edit
before it, so remembering to sync was remembering to do something the computer
could do — including the worker refresh, which is what makes an edit to your
own package take effect at all. `--no-sync` opts out for a tight loop.
That last one needed discovery fixed: it only ever looked at top-level `*.py`,
so a repository whose code is in a package — the ordinary shape — found
nothing from its own root. It now descends into the packages it holds.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012ue1tkFWB1bcGy3aWhCKpU
Two things a local install should not have asked for.
`fluksio serve` now signs you in. Logging in to your own machine was a
formality — the password was printed by the same process that would have
checked it, and the database it authenticates against sits in the directory
the token goes into — so `serve` mints the token itself and says where it put
it. `fluksio login` is left for an engine somewhere else.
And an installation is `.fluksio` beside the code, found the way `.git` is,
rather than one `~/.fluksio` for the machine. A repository with its own venv
was already getting its own engine; it now gets its own flows, run history and
token too, instead of three repositories sharing one database and fighting
over one port. `--global` asks for the shared one, `--data-dir` still names
any directory, and when both exist the banner says which you are looking at
and how to reach the other.
The directory ignores itself from within — a `.gitignore` of `*`, the way uv
writes one into `.venv` — because it holds a credential and a database, and
neither belongs in anybody's history. The token is written mode 600. A login
an older version wrote to ~/.config/fluksio is still read, so nothing that
worked stops working.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012ue1tkFWB1bcGy3aWhCKpU
The gates have never gone green on the new runners. Three separate reasons:
- backend/Dockerfile shipped Python 3.10 while the code imports typing.Self
and datetime.UTC, so the container exited on import and the suite could not
even load its conftest. The image moves to 3.13 and the packages declare
>=3.12, which is the floor the tests actually pass on; ruff's target follows
and rewrites timezone.utc and asyncio.TimeoutError accordingly. Relocking
drops the 3.10 branch, which bumps FastAPI and so regenerates the SDK.
- frontend/README.md had no trailing newline and two dashboard widgets used
arbitrary text-[…] sizes. Both are em-relative on purpose, so they move to
the inline style the neighbouring ramp already uses.
- Every commit left its own run queued: without a concurrency group a runner
that was offline for a while works through a backlog nobody reads. A stack
that fails to come up now prints its logs before the teardown removes it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
poll() still calls a child that has exited but not been reaped alive, so a
node that ended itself with os._exit left its corpse on the idle queue and
the next call died on a broken pipe. Mark it the way the out-of-step paths
already do, and _release replaces it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
readme and license-files name README.md and LICENSE, but the image only ever
received pyproject.toml, so hatchling failed metadata validation the moment
the layer cache stopped hiding it:
OSError: Readme file does not exist: README.md
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two looks were somebody else's language spoken well, and neither was the
product's. A dashboard nobody has dressed yet should look like the rest
of the app, so there is now a third set that follows the root
DESIGN-GUIDELINES.md to the letter — `--card` surfaces told from the page
by a hairline and a low shadow rather than by colour, every control a
pill, 16px panels, frosted floating chrome, one slate-blue accent spent
on what a person can act on — and it is what `look` means when nothing
says otherwise.
That also turns the exemption the other way round. The dashboard is still
allowed to look unlike the product; it just no longer does so by default.
An existing dashboard, which has never named a look, lands on the design
it had before any of this.
Restraint is the style rather than an omission here: no ripple, no glow,
no lift, and a press answered by the colour changing. The one deliberate
departure is the selector, which holds its choice in `--primary` rather
than the `--accent` the segmented rule asks for — that is a decision
about the widget, not about the look, and a control must not change what
it signals when the drawing changes. All three sets hold it the same way.
Every other README carries the AGPL line; these two did not, and they are the
ones that become pypi.org/project/fluksio and /fluksio-worker — the most
public surface the project has, and the place a copyleft licence is least
useful to have to go looking for. §13 is named outright for the same reason.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012ue1tkFWB1bcGy3aWhCKpU
Several of them describe this installation specifically — its broker, its
aircon, its DMX wiring — and this repository is about to be public. The
test and client-generation scripts stay; CI depends on scripts/test.sh.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Adds reference/connector-contract.md and code/connectors.md to the site, with
the node-type reference finally linking the contract it describes. Strict
build passes.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Replace the empty LICENSE placeholders with the verbatim GNU AGPL v3 text,
fix the invalid "AGPLv3" SPDX string in the package metadata, and name
Melvin Strobl as the copyright holder wherever the old footers said
"Fluksio ... all rights reserved".
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The release workflow was already right; what it would have uploaded was not.
`fluksio` had no readme, so its PyPI page would have been blank — the app
repo's own README is a contributor's map of `frontend/` and `docker/`, which
is the wrong front page for `pip install fluksio`. It now has one of its own,
aimed at somebody who landed on the project page. Both distributions gain
authors, urls, keywords and classifiers; `twine check` passes clean on all
four artifacts where it warned on two before.
The workflow publishes `fluksio-worker` first, because `fluksio` depends on it
and the other order leaves a few seconds — the whole of a first release — in
which the dependency cannot be resolved. `--check-url` makes a re-run skip
what is already uploaded rather than failing on it, which matters because a
version on PyPI can never be replaced.
Licence metadata is deliberately still absent: LICENSE is MIT in somebody
else's name, inherited from the template this was scaffolded from, and whose
it should be is not a decision to make in a commit. NOTEPAD carries it.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012ue1tkFWB1bcGy3aWhCKpU
Stretched end to end, a panel with two dashboards showed a pill nine
tenths empty — which was tolerable while the rail was chrome outside the
canvas and obvious once it moved onto the panel itself. It now hugs its
entries and sits centred in the column it reserves; that column is the
same width either way, so no arrangement moves.
A rail longer than the panel is tall still scrolls, and still draws no
bar: a wall panel is swiped, and there is no room for one.
Five things a wall panel showed that a wall panel should not.
**A tile's body no longer clips.** It scrolled, and a box that scrolls
also cuts whatever crosses its edge — which took the glow off a lit
button at exactly the width where the button filled its tile, and off a
gauge's arc at exactly the height where the dial filled its own. Only
what is written or listed asks for a scroller now; everything else is a
picture drawn to fit, and what overflows is left to the frame, which
clips at the tile's edge where a shadow has already faded out. The
slider's phantom scrollbar goes with it.
**The selector is a selector.** Named for what it does rather than what
it is, and the choice it is holding is held in the dashboard's own
primary — a pill that slides between the options rather than a grey one
that had to be looked for. The stored type is untouched, so no document
changes meaning.
**The arrangement is held off the panel's edges**, by the same distance
it holds between two widgets. The ground is not held off with it: a
background covers the whole panel, and only what is arranged on it has
a margin. No stored panel loses a row to it.
**The rail is drawn on the panel.** It was chrome bolted to the edge of
the screen beside the canvas — in the app's own design rather than the
dashboard's, and on a scaled canvas not even lined up with it. It now
takes a column out of the canvas the way the margin does, scaled with
it and wearing its look. Which cell each widget sits in is unchanged;
only how big a cell is.
Two of these were the same mistake twice: an unlayered rule stating
`position` for everything wearing a class — `.gl-surface` on a rail
placed by a utility, and a blanket lift over every child of a pressable
on a pill placed by `layoutId`. Both now say it one element at a time.
The workflow this serves: make a venv, install what you work with, then `pip
install fluksio` into the same one. Building a second environment beside it
was exactly wrong — the packages the nodes need are already here, and the
Modules screen was asking for them a second time.
`NODE_VENV=auto` (the default) adopts that venv. It declines in the three
cases where adopting would be wrong: `managed` says otherwise, a managed venv
already exists and may hold packages somebody installed on purpose, or the
engine is not running from a venv at all. The images set `managed`, since the
venv in them holds the app and nothing of anybody else's.
An adopted venv is never written to. `uv pip sync` makes a venv hold exactly
the manifest, so pointed at somebody's own environment it uninstalls their
work and the engine with it — `sync()` refuses outright and `reconcile()`
returns before it can be called at startup, which is where that would have
happened first. The Modules screen lists what is installed and drops its
editor; `pip` is how that environment changes.
`fluksio serve` now names the interpreter node code runs on, which is the
thing a data scientist most needs to know at that moment. `fluksio-worker`
already defaulted `--python` to its own interpreter, so a GPU box works the
same way — that was only ever undocumented.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012ue1tkFWB1bcGy3aWhCKpU
- A tile no longer lifts under the pointer. A finger does not move away
afterwards the way a cursor does, so whatever hover raised stayed
raised until something else was touched: a tile stuck, not answering.
- The ground's blobs wander a closed path on their own clock instead of
sliding back and forth along one line, which read as things moving
rather than as light in a room.
- A bar is one grid now, its rows borrowing its columns, so names of
different lengths no longer start and end their tracks in different
places — two bars that share no baseline cannot be compared, which is
the one thing a stack of them is for. The names read rightward into
their tracks, with room either side.
- A reading on its way somewhere is written to as many decimals as the
value it is heading for. Without that a slider stepping in halves
passed through 22.37460937 on its way to 24: a number nobody asked
for, a different width every frame.
- The brightness column is the same control as the slider widget's,
stood on its end and thicker, and exactly as tall as the disc beside
it. Getting there meant drawing a slider's rail, fill and handle
rather than styling `::-webkit-slider-*`: those need one set of rules
per orientation, each with its own centring quirk, and the handle
landed off its track when the writing mode turned. The native input
stays, laid transparent over the top, so the keyboard, the pointer and
every `aria-` are still its.
The disc was sized against its parents in percentages, and every box
between it and the tile is sized by what is in it — so the chain never
resolved, the brightness slider grew, and the tile scrolled with half a
disc showing. Container units are the tile's own height whatever sits in
between, so the widget now fits exactly.
Two guards behind it, since both are things a picture can fail silently:
nothing on a panel of pictures may overflow its tile, and Touch has to
actually make a control bigger while leaving what it publishes alone.
A dashboard is a wall panel somebody hangs in their own hallway, so it
now wears what they choose: a look, and a palette of their own colours.
Two complete component sets live under `Dashboard/ui/` — `glass`
(translucent panes over a slowly moving ground) and `material` (Material
3 tonal cards) — behind one prop contract. Every control's state,
keyboard and `aria-` live in `ui/core` and are shared, so the two sets
are the same dashboard drawn twice rather than two products: a set only
decides what a control looks like while doing it.
Four settings join the channel, each drivable by a flow like any other:
`look`, `palette`, `background` and `touch`. A palette is an ordered list
of hex colours — background, surface, primary, accent, text, then more
chart colours — pasted from a coolors.co link or typed, written onto the
canvas as the token variables everything already reads. Trailing roles
are derived, so three colours are a whole dashboard, and derived text is
held to AA rather than trusted (`theme.check.ts` measures it). A palette
also decides light or dark, since its first colour is the ground.
Widgets are measured against their own tile with container queries rather
than against the viewport, animate through `motion`, and can be drawn
without their title. The three reworks:
- a bar draws a row per reading, up to eight, each in the dashboard's own
data colours and each able to carry its own scale — replacing readings
nested in one fill, which could only ever share one colour and stop at
three. Documents written the old way are read as rows.
- a chart's range picker moved to a column down its right-hand edge, which
gives the plot back a whole row of a short tile.
- the colour wheel became a disc: hue is the angle and saturation the
distance from the middle, so a colour is one gesture rather than three,
with brightness on a slider beside it.
`index.css` and `lib/motion.ts` are untouched — the dashboard overrides
token *values* on its canvas, never the blocks the two repos share.
Nothing reads it since the engine moved its state to SQLite, but it still
carries the pre-cutover history and an alembic stamp naming a revision that
commit deleted — which reads like a broken migration chain and is not one.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012ue1tkFWB1bcGy3aWhCKpU
A data scientist keeps their code where it is and decorates it: `@node`
declares a function's ports beside the function, `Flow(name, nodes=[...])`
says which of them make a flow, and `use(fn, wire=..., **settings)` rebinds
one for a single flow. `fluksio sync` uploads the document plus a generated
import shim per node, so the store still holds a complete, runnable,
git-versioned definition while the code it imports stays theirs.
`fluksio login|run|runs` and `flow.submit().wait()` are the client half, over
the run endpoints that already existed. Runs record the user repository's
commit beside the store's, so "what code produced this number" is answerable
on the side that now holds the code.
- `fluksio/sdk/`: ports, decorators, the flow builder and its checks, the shim
generator, an HTTP client and sync. Standard library only at import, so
`from fluksio import node` in a training script pulls in no engine.
- `FlowDef.origin` marks a flow code-defined; `Run.origin_commit` carries the
repository's commit; `POST /modules/refresh` retires the workers without an
install, which every sync calls — a worker holds the imported package in
memory, so an edit to it is invisible until the process goes.
- The canvas shows a generated body read-only and names the repository to edit
instead; a body edited there stops the next sync rather than being discarded.
- The worker's reporter carries inert `Port`, `node`, `use` and `Flow`, since
the shim imports a module whose first line declares them.
- `examples/myresearch` is the worked example, `make sync-example` uploads it.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012ue1tkFWB1bcGy3aWhCKpU
The engine and the worker speak a version-matched protocol, but compose.dev.yml
mounted only backend/fluksio: a reloaded engine talked to the worker baked into
the image. Adding the request id to that protocol therefore left every python
node failing with "the worker answered request None", because the old worker
echoed no id at all.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01StpRc2C6au1WJ1EUU7fsfu
The module docstrings and the notepad still described one rebuild that
touches everything. Closes the toggle cost, the seeding cost, the
per-save rebuild, the modules/apply rebuild and the Playwright spec that
could not fit a rebuild into its five seconds; files the follow-ups the
refactor leaves behind.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01StpRc2C6au1WJ1EUU7fsfu
A stopped flow's nodes are built like any other flow's — being stopped
means having no subscriptions, schedules or webhooks, not being absent —
so a toggle only ever needed the lifecycle call and the gate that goes
with it. It was doing a whole-pipeline rebuild instead, which on a
populated installation is every node in every flow reconnecting.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01StpRc2C6au1WJ1EUU7fsfu
Publishing or deleting a flow now splices that one flow into the running
graph instead of reconnecting every node in the installation, saving a
shared node's source rebuilds the flows using it, and installing modules
rebuilds only the flows holding a node that would not load.
Renaming stays on the full rebuild — it rewrites message references in
every other flow's document — and so does startup, which has no graph to
splice into.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01StpRc2C6au1WJ1EUU7fsfu
A node's timeout now covers its body only: the pool loads the source into the
worker it picked, off the node's budget, so imports that outlast the timeout no
longer make a node impossible to run. Draft checks compile without caching, so
saving does not evict what a busy node is serving calls from. Requests carry an
id the worker echoes and the pool checks, a reply is encoded once, and the
remote-exception cache is bounded.
DELETE /flows/{name} answers 409 while the flow has a running or queued run,
which is what was letting run_node rows outlive their run.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01StpRc2C6au1WJ1EUU7fsfu
`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
stroblme
