Node source (Python) and the JSON dialog share one CodeMirror editor with
find and replace, autocomplete, folding, indent guides, comment toggle and a
JSON lint, themed from the CSS tokens so light and dark follow the .dark class
without a second palette. Programmatic value changes no longer fire onChange,
blur ignores focus moving into the find bar, readOnly keeps selection and copy.
The editor chunk is 470 kB (157 kB gzipped) against 2.7 MB (704 kB) for Monaco,
which also pulled in a 1.2 MB JSON mode, two workers and a font. At 4x CPU on
slow 4G a node panel is ready in 5.7 s instead of 10.8 s, the JSON dialog in
1.5 s instead of 6.7 s. Typing without a key delay keeps every character.
monaco-editor, @monaco-editor/react and the dompurify override are removed.
Specs: typed code reaches the node source, a JSON setting saves only text that
parses. Docs name Tab and Ctrl-M.
The channel sends through the Bot API with httpx only: HTML with escaped text,
cut to fit a message, silent for a warning, a short 429 waited out once, and
errors that say what to do about a blocked bot, a lost chat or a bad token and
never quote the request URL. The bot token is a stored secret: saving a
channel refuses a literal one, the card's "Store token" writes the secret and
a reference, and no endpoint returns it.
A chat id is never typed. Connect starts a ten minute code, and the chat that
sends `/start <code>` back through the bot is the one connected, so it is a
chat somebody opened a link for. The engine fetches updates only while a code
is waiting (no standing poll, no webhook, no inbound address) and ignores
everything else without a reply. Connect needs a saved form, and the form
takes the connected chat in without touching what was typed since.
The token is part of every Bot API URL, so the httpx logger gets the same
filter the request lines have, which now also masks an httpx.URL argument.
The alert manager kept its cooldowns and its ceiling per hour in memory, so a
node that stayed down was announced again by every restart. Each delivered
alert is now a row in the new alert_sent table (additive migration), loaded
for the last seven days when the manager starts and pruned as new ones come
in. A ledger that fails never stops an alert. The manager's clock is wall
time for the same reason.
A node_health "down" alert also waits 60 s (HEALTH_GRACE_S) in a pending task
per node; a later event that is not "down" cancels it, as does stopping the
manager, so a broker restarting with the stack alerts nobody.
The cooldown used to be stamped before the ceiling per hour was checked, so
an alert the ceiling held back still spent its cooldown. It is stamped only
once the alert is sent now.
The summary read degraded for 15 minutes after any single node_error, even
once the node had run well again and including one bad input from a visitor.
The controller now counts a node's failures since it last ran well and notes
when it recovered; both carry across rebuilds and an acknowledgement resets
the count. Two in a row inside the 15 minute window is "failing" (a problem),
one is listed under warnings, and nodes.failing keeps meaning "failing now".
Any signed-in account could save an ntfy channel with its own `server` and a
`{"$secret": ...}` token and press Test, which sent the stored secret to that
host. The config GET/PUT, the test route and the global search's alert entries
now answer to superusers, like /secrets. The webpush key and subscribe routes
stay open to every account, and the Alerts page shows a non-superuser only the
button for its own browser. The events table in the docs gains run_finished.
/search gains a `setting` category, `<node>.<setting>` under its flow with the
node type and the value in a new `detail` field, so Ctrl-K finds which node
uses a broker or a timeout and opens the node. `setting_text` is the one place
that decides what may be said: a reference to a stored secret shows its name
(to superusers only, as the secrets list does), a plain value in an `x-secret`
field or under a key like password, token or api_key is masked and never
indexed, nested credentials in headers or query are masked, long values are
clipped. The global search ranks what contains the query as typed ahead of
loose spellings before it caps a group, which long setting values would
otherwise have crowded out.
An input row now starts from its message and ends as the argument the code
takes it as ("as prompt"); an output keeps its key on top and the message
below. The name the code uses is muted text with a pencil while it follows the
message's last segment, and a field only when it differs. Device Follows ports
get the same order (message, then the port driven). A list port's item type
moves to the second row so it no longer squeezes the name, two ports deriving
one name get a warning, and the rows rebuilt use 44px touch targets on phones.
The stored node spec is unchanged.
The command line no longer imports a research module, so import-time GPU work
ends with a short-lived child instead of holding a CUDA context for the length
of a followed run. fluksio.sdk.declare starts the child, which writes each
flow's document and shims as JSON; Declared stands in for Flow in sync().
Errors keep their type and message, a child's traceback reaches the user.
save_artifact/load_artifact outside a node use the worker's own store code in
$FLUKSIO_ARTIFACT_DIR or <tmp>/fluksio-artifacts, so a node can be called by
hand. --follow prints status changes and each node's resource wait; the
placer's waiting snapshot now carries the run. fluksio cancel <run>...
The phone header carries a single speech bubble: Control on Home, the
assistant elsewhere. Chips above the input offer the user's most sent
messages per tab, then starters; a tap or a double-click on a sent or
heard message fills the input without sending.
POST/GET /flows/{name}/nodes/{id}/train submit wren_train on the node's
targets, examples and tests and read its latest run (run.for_node,
cause "train"); "Use this model" writes `model` into the draft.
Every output is a target and every input besides prompt a reading. A
targets setting gives descriptions, values, ranges and per-target
guards, model takes a trained pack's artifact digest, min_confidence
defaults to the model's threshold. A command is applied whole or not at
all, and targets the model was not trained on read degraded. The home
path (area/kind ports, groups, brightness, temperature answers) is gone.
wren's model answers "turn off all lights" with one call, and at 0.98 a
living room light switched on. The node now applies such a command to every
wired target of the kind. Number words become digits ("in five minutes"
lost its delay), aliases put the house's words and known mishearings in the
model's, and sensors.temperature is answered from <area>_temperature or
outdoor. tts writes out units and clock times before speaking.
wren drops a call on an area the command does not name (other than the
user's or, with none named, the living room): "switch on the appliances"
turned on the bathroom light at 0.97. The assistant's `values` setting gives
a str target its only values, explains them in the tool and confirms in the
word set for each ("Canopy open.").
Shown only the wired areas, the model took "turn on the bathroom light"
for the living room's light and "the office light" for the kitchen's,
both above 0.93: 5 of the 28 recorded 7B commands switched a wrong light.
Every area is shown now, with a light each, so such a call names an
unwired device and goes on to `unhandled`.
With every area shown, a transcript's "Turn on the kitchen light." fell
to 0.80. The command is passed lower case without sentence marks, as the
model was trained, which brings it to 0.99.
run_turn takes answer_tools: false ends the turn once the tools have
answered. The assistant node uses it and says what was set ("Kitchen light
off, living room light off."), which saves one model round, about 1.1 s on
the P100. A turn without a call still streams the model's own words.
rebuild-frontend keeps the running stack's FRONTEND_HOST, so the token
issuer no longer flips between the LAN address and app.localhost.
run_turn takes its tools, system prompt and tool caller; the chat passes the
MCP tools as before. The assistant node offers one tool per wired target,
reads its other inputs as context, streams its reply a sentence at a time,
warms the provider on `listening` and says so when the provider is not there.
Completions share the per-server HTTP client. wren replies when nothing reads
`unhandled`, stt shows a note when nothing was heard. /ai/control lists one
conversation per entry node with the messages to watch.
transcribe and speak share an httpx client per speech server with a 300 s
keep-alive, and run on the engine's loop. stt warms it with GET /v1/models
when an utterance opens, so a reply's first sentence finds the connection
open: a new one over a lossy link cost up to a second (openpanel V4).
- `stt`: one realtime transcription session per utterance on the speech
server's url, opened at `listening` true or the first chunk, committed at
false; a whole clip goes to /v1/audio/transcriptions. Returns at once and
publishes `partial` and `text` when the server answers.
- `tts`: one /v1/audio/speech call per sentence, cached by voice and text in
node state; a reference is marked `append` while earlier ones still play.
- wren: `reply` says what was applied, `unhandled` carries the prompt when
nothing was. The install hint names the git pin (PyPI's `wren` is another
package).
- ai/client: transcribe, speak and transcribe_stream beside stream_completion.
- A device socket publishes scalar `in` ports as {"type": "publish"} text,
in order with its binary frames.
- Pipeline.publish walks only what its values reach: a node downstream of a
consumer that published nothing no longer runs again (wren re-acted on
every microphone chunk).
- `service` node type: command, port, worker label, health path, env and
weights (an artifact digest the worker fetches as {weights}). Its flow
starts it; it restarts with backoff, reads degraded while starting or
without a worker and down after three starts that never got healthy, and
publishes http://<worker>:<port> once /health answers 200, "" when it ends.
- Worker protocol 3: a `serve` call, every call in its own process group,
killed when the engine ends the call, the socket drops or the agent gets
SIGTERM. `--service-host` overrides the address the engine sees.
- Services and compiles take no `--parallel` slot, and a flow's detail is
built off the event loop, so a save no longer stalls the engine.
C4: a device's `viewers` (superusers by default, or members) covers the
image and video messages its ports publish. Anyone else gets none of their
values on the socket, in the catalogue or a flow's state, 403 for their
history, for the frames they hold (the last 8 included) and for TURN
credentials, no bytes when asking for them, and `bye: forbidden` for a live
view. A device credential stays bounded by its dashboards. The device page
sets it in its Camera section, and a tile the viewer may not see says
"No access to this camera" and asks the camera for nothing.
C3:
- A volatile upload is held to MAX_FRAME_BYTES (4 MiB, 413 above it, read
without extra copies), and so is a frame on a device's socket.
- A frame the engine holds in memory is written to the ring directory
before a local worker is handed it, so load_artifact finds it.
- A socket the bus dropped events for catches up with a snapshot, demand
included, and a device's credential is checked again per frame, cached by
the flow store revision and its nonce, so a lost event never keeps an
unpaired device's frames coming.
- A report is at most 64 KB once its named config_schemas are put in place.
- A device away for 60 s has its true bool messages set to false.
- Refused frames are logged by fixed keys, a frame is named after its
message, `fluksio serve` waits 5 s for open connections, a tile with a
stream URL creates no demand, the 8-name cap keeps the newest names.
- Every container's log rotates (json-file, 3 x 10 MB).
A device reporting a `live` port streams its camera to the browser's own
RTCPeerConnection. The flows socket carries the signalling as text, so it
crosses the portal: `live` messages on a person's socket, `signal` on a
device's (start, sdp, candidate, bye, busy). The server routes by session,
holds a device to two viewers, ends a session when either socket closes or
no answer comes within 10 s, and gives each socket an outbox of its own
that ends a session rather than drop a message. TURN credentials are
minted per session in coturn's REST scheme (STUN_URLS, TURN_URLS,
TURN_SECRET), the viewer's with the device's offer; GET /devices/{id}/ice
hands them out too. The tunnel keeps a stream's early frames instead of
dropping them past eight and forgets them when the stream or link goes.
A media tile on a device's camera, and the Devices tab's preview, play
live when tapped, pause frames meanwhile, and fall back to frames with a
short note on busy, failure or no picture within 10 s. coturn is an
optional compose service behind the `turn` profile.
- Volatile frames are held in memory with a byte-bounded queue instead of
the /dev/shm ring scanned on every put; workers keep writing to the
directory, which is trimmed on a timer. A volatile upload no longer spools
to disk. ARTIFACT_VOLATILE_BYTES=0 turns the ring off.
- A device sends camera frames as binary messages on its socket, held and
published in one step and credited to its device node; refused frames are
logged and dropped. Its snapshot carries demand and the uplink (binary, or
http through the portal).
- One demand registry for every socket's wanted media, the tunnel's
included, plus what running flows read, mapped onto on-demand ports and
published as device_demand; at most 8 media names per socket.
- Per-message deflate off in serve and dev; a device may fetch the last 8
digests of a message; published dashboards cached per store revision; a
revoked device socket is closed with 1008.
- Media tiles want bytes only while on screen; the Devices tab previews a
wired camera; the device node warns when a flow reads an on-demand port.
Every call through the portal arrives off the tunnel, so the per-address limit
skips those (the hub limits them by the real address), and the cap of 50 counts
and evicts portal and local requests apart, so a portal flood never pushes out
a device waiting on this network.
- POST /devices/pair answers a 2-character `check`, listed by /devices/pending
and shown beside each waiting entry; five codes a minute per address, and at
50 waiting the oldest gives way instead of a 429
- portal-minted device tokens are asked for with the device's nonce (`pnc`) and
held against it, so unpair and delete revoke them; one without `pnc` stands
only while the nonce is 0
- a device credential acts as the device (`device_principal`), not as the
approving superuser, so deleting that account unpairs nothing
- the report is validated port by port: unknown enum values read as `unknown`,
unreadable ports and sections are left out and logged, `config` may name an
entry of `config_schemas`; 422 only for no schema 1 or no object
- devices.json is parsed once per change (cached by stat), delete drops the
report and keeps the nonce, a busy recheck runs in the background
- scope and io only from enabled flows (start/stop tells the devices), a port
bound twice gives one io row
- the Devices tab's saves cancel and hold back device refetches until settled
- A device credential that cannot be checked now (device store not running,
nonces unreadable) raises DeviceUnavailable: 503 over HTTP, and the
/flows/ws handshake and the tunnel stream accept and close with 1013.
1008 (a 403 before accepting) stays for a revoked or invalid one
- Device files are written through a temporary file, so a read mid-save no
longer looks like a deleted device
- The first report turns Touch friendly on only for a device paired since
pairings are noted (DeviceState.paired_at), so devices paired before keep
the switch they were given
- NodeDef reads `device`/`device_policy` from older clients as the worker
label, for a release
- Tests: transient vs revoked socket, the touch default, the old worker keys,
and a restart with a device node and a node_health rule sending no push
A device reports its board, screen, camera and ports from its own
credential (16 KB, schema 1, replacing the last report), and the flows
are validated against it again. The node panel turns a reported port's
config schema into a form, a reported screen is a canvas preset and a
touch screen's first report turns Touch friendly on. An offline device
node is degraded: amber on the canvas, a warning on Home, never a fault.
- NodeDef `device`/`device_policy` become `worker`/`worker_policy`, the
issue `ignored_device` becomes `ignored_worker`, and the SDK's `@node`
takes `worker=` and `worker_policy=`
- FlowStore._migrate renames the keys in published and draft documents
once, bumping their version; a queued run's `needs` reads the old key
- placement passes the label as `label`, client regenerated
- docs: worker placement wording, MCP issuer without DOMAIN, device rows in
the CLI data directory and the API table, device links in compose comments
- flow/nodes/device.py: `device` node with params `device` and `config`;
provides the device's `in` ports, requires its `out` ports without
waking anything, never run (Node.passive), down "Offline" from presence
- DeviceDef.io and DevicePort go; GET /devices/{id} computes `io` and
`bindings` from the device nodes of published flows, in a fixed order,
cached by FlowStore.revision; the credential's scope comes from them
- issues `unknown_device` and `device_port` (wired twice; against a report
the wrong direction or dtype, or unlisted as advisory), rechecked when a
device is created or deleted; ValidationIssue.advisory is a field
- a device publishing on a bound port is credited to its node by the
server; the panel endpoint kind is gone from `_endpoints` and the canvas
- a rebuild that moves a device's bindings sends `dashboard_changed` with
an empty name, which the firmware refetches on
- startup moves each device's `io` rows into device nodes in the flows
declaring their messages, draft and published, one commit; rows no
flow declares are logged and left out
- frontend: Device node settings (device, Publishes, Follows, a port
picker from a report), palette shortcuts per reported port, DeviceDetail
lists bindings with "Use in flow" in place of the I/O editor
- docs: devices, flow editor, node types, issues, API
stroblme
Melvin Strobl