a7719fca6a5b7cac4cb827ad22ffed4aea669156
77
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 tajniak81andClaude Opus 5
|
a7719fca6a |
A line per frame, for the frames nobody has named
ANKER_MQTT_FRAME_LOG logs every inbound cloud frame with its bytes, decoded or not — the ones this package drops are exactly the ones worth naming, so they are logged before the drop. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
245870a96a |
The add and remove buttons, and the read that checks them
Anker documents neither rfid write, so the bodies are inferred from the field names get_device_cards answers with, and every write re-reads the list: what the card shows is what the account holds, never what an undocumented endpoint claimed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
4ff6242c8f |
The last message in the map, and it reboots the charger
0108 was the one thing in the MQTT inventory nobody had wired: the device power mode, whose single documented value restarts the charger. It is the only way to reboot a charger that is on neither a CSMS nor the local network — which is most of them — so the cloud transport sends it now, and "reset" reaches it too, since that is what the OCPP path has always called the same act. Nothing waits for a confirmation: the device that would send it is the device rebooting, so the command answers at once and says the charger drops off the cloud for about a minute. The gate is unchanged and now covers both spellings — an explicit confirm plus a password step-up, audited either way. Modbus still refuses, because no register does this, but its refusal now names both transports that can rather than only the CSMS. Both clients already had the reset button and its password prompt; they were hidden in every mode that reads the device, which is why the cloud never showed one. Modbus is now the only mode without it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
b2d333a63f |
The charger was never asked what it is set to
The trigger buys telemetry and only telemetry, so a charger that has been read a hundred times and commanded none reports amps, volts and nothing else: no schedule, no balancing, no Modbus server, not even its firmware. The message that asks for that half is 0040, and the reference keeps it commented out because the app sends its timestamp without a value type. The app is what the charger answers, so the oddity is reproduced rather than corrected — sent when the settings half is missing or older than ten minutes, waited four seconds for, and after three unanswered requests still sent but no longer waited on. The three settings the panel has and the writer did not — swipe up, swipe down, smart touch — are writable now, which is all eleven of the 0100 commands. Nothing else in the map was missing: every named field of every message was already decoded, and the raw keys the card shows are fields the reference does not name either. Both cards drop a row with nothing in it, which turned a charger that reports only its ceiling into a charger that reports no current range at all. Half a range is still a bound. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
3f9d5b943f |
Four questions asked of an account that answers twenty-one
The connector called ten endpoints of the read surface the map lists, and the charger card showed four views. Everything else an EV charger can reach is now a capability too: the sessions and the history, the savings, the sharing, the binding, the group, the Wi-Fi, the firmware and its update log, the tamper records, the site's own detail, price, networks and energy — plus the vehicle catalogue, dynamic pricing, the currencies and the notification views. Thirty-eight endpoints, one action each. The two message views are GET, so the request path grew a GET half that shares the login retry with the POST one. The per-charger fan-out asks all of them, six at a time rather than one after another, and a charger that belongs to a site brings that site's four views with it once the by-serial lookup has found it. A view that answers with nothing now says so instead of vanishing: the station record is empty for a standalone charger because it has no station, which is an answer worth reading. And "source 0" in the OCPP box carries the address the account's endpoint list gives it. Anker's account-level writes stay out, as do the endpoints whose payloads were only ever read out of the app package. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
1a7f04cba0 |
A dash printed above the value it was missing
The Charger information card read "—" beside State and OCPP status while the raw
block three rows below it printed chargerStatus 1 and ocpp_connect_status 2. The
account had answered both. The merge asked for the state as evChargerStatus,
operating_state or status, which is how the standalone and station views spell
it, and the bound-device view — the one this account actually answers from —
spells it chargerStatus. The OCPP state it never asked that view for at all. All
three views now read through one fillDevice, which tries every spelling a view is
known to use, so a value any of them sends reaches the row that was drawing a
dash for want of it.
The same views were carrying the whole box-on-the-wall half unread: the Wi-Fi
network and its MAC, the signal strength, the Bluetooth MAC, the time zone, when
the account bound the charger, how the app can reach it — BLE, Wi-Fi — and the
product shot for the model, which now sits beside the charger's name in both
apps. Named rows, in three languages, the way the register map's readings are
named.
One field wanted the opposite treatment. The device record carries blue_password,
the charger's own Bluetooth pairing password, and the card was printing it in
clear into every screenshot anyone takes of that page. Any leaf key holding a
password, secret, token, private key or certificate is now masked in the raw
block: that the field exists is worth reporting, its value is not.
Four endpoints answer only when a serial is named, so none of them could belong
to the list the card is drawn from, and nothing had ever called them. The station
record, the charging totals, the OCPP backend and the RFID cards now arrive
through a charger-details capability behind
GET …/anker-solix/chargers/{sn}/details, asked for the charger being looked at,
best effort, each view reporting its own failure — an account that is not the
owner cannot read the cards, which is a fact about the account rather than an
error in the read.
Those four are shown under the cloud's own keys, and that is not an oversight.
The REST map documents which endpoints exist and what each is for; it does not
document a single one of their payloads. Naming those fields is the next commit,
made from what actually comes back, now that there is somewhere to see it.
Not touched: the endpoints the map marks ready but unwired — session history,
site price, OTA, sharing, notifications — each a feature rather than a row on this
card; and the unmapped ones, which the map warns delete sessions and unbind
devices with payloads nobody has ever seen.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
tajniak81andClaude Opus 5
|
8e2073fc4c |
The map knew the names the card was showing as hex
Every field in the cloud MQTT map that has a documented meaning now reads as a named row, on the same labels the register map uses for the same quantities. The raw block stays, and shrinks to what genuinely nobody has identified — which is the only honest reason for a key like 0410.b9 to be on screen at all. Three fields the reference decodes for nobody are decoded here. ac is where the charge is coming from — off or paused, grid, solar — and it is called chargingSource rather than chargingMode, because that name already belongs to a Modbus register and the last time a cloud field borrowed one, d9 spent a release reporting the wrong thing under the right name. b6 is the session's order id. f1, f2 and f3 are the identity fields the reference marks multi-value: four bytes read as the parts of a version, in the order they arrive, which is what the account view's own firmware string looks like. If the panel shows those parts reversed, the order is the thing to flip — it is the one assumption here that the wire has not yet confirmed. The rest was already decoded and simply never drawn. The readings card now shows the session's start, its id, the charging source, whether a cable is in, the charging window, and — since a reading is worth what its age is — the live-stream flag and both stream clocks, because telemetry and settings arrive on different messages with different triggers. Per-phase session energy joins the phase matrix as a fourth column, appearing on the transport that counts a session and staying away from the one that does not, exactly as the reactive and apparent pair does. The settings block gains the fourteen the register map has no address for: plug lock, auto restart, random delay, the schedule and its mode, the weekend window and how the weekend is handled, the light-off schedule and window, the breaker limit, the solar mode and its minimum current, automatic phase switching, the three panel gestures, and what the two balancing features are watching — the meter and monitor serials by name, their two unpinned numbers as the numbers they are. A local network block says whether the charger's own Modbus server is on and where, which is the answer the Modbus mode's setup screen otherwise has to be given by hand. The device block gains the controller version. A test now holds the line the projection quietly drew: every name in the message maps must reach a snapshot field. A name added to a map without a field to land in would otherwise surface in the raw block looking like something we understood. Left raw: a1, the frame opener the charger echoes back; b7, which the map itself calls unidentified; b9, bc and bd, which appear in no map; the five-minute 0400; and 0857 — a message type the reference's closed inventory of fourteen does not contain and this charger publishes anyway. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
2425a8d3d6 |
A field we have no name for is still a field it sent
Two cards on the Charging page were answering with a fraction of what the charger and the account actually report, and in both the losses happened quietly, in a parse that kept the fields it recognised and dropped the rest on the floor. Charger information asked three account-wide views and kept fourteen fields. A charger registered on its own is absent from the site view, which is the only one of the three carrying state, charge power and OCPP status — so exactly the charger that stands alone got the column of dashes, and nothing said why. The per-charger station record, get_evcharger_station_info, is what the mobile app opens when you tap a charger, and it is the one view that answers for a charger outside a station; it is now the fourth view, asked per charger, a failure there costing that charger's row and no more. Alongside it, every field each of the four views sent is kept as attrs, under the cloud's own key, nested objects joined with a dot and arrays carrying their index. First view to answer a key wins, which is the rule the named fields already merged by. Two hundred keys and two hundred and forty runes per value keep a station record with a session list from becoming the whole card. Charger readings lost data twice over. The frame decoder skipped any field byte its per-message map could not name, and a message type with no map decoded to nothing at all; those fields are now kept under the message and the byte they arrived in — 0410.c9 — decoded but unscaled, because a factor is half of a meaning and we do not have the other half. Then the projection read forty-odd names into typed fields and dropped the remainder: sessionStartedAt, the per-phase session energies, the three touch modes, the load-balance monitor and its meter flag, the solar monitor. Those land in extra, and the list maintains itself — the four accessors note every key they read, extra is what is left, and a field modelled later stops appearing there without anyone remembering to remove it. Keeping unnamed fields had one consequence worth guarding. An unmapped message now decodes to something rather than nothing, and ingest stamped settingsAt for anything that was not telemetry — the timestamp a control command waits on to say the charger acknowledged it. A frame we cannot read is not an acknowledgement, so the stamp is now conditional on the message type being one we map, while its fields are kept either way. Both cards show the remainder as what it is: the service's own key, no unit, no translation, no renaming, under a heading that says whose words these are. The blocks appear only when there is something in them, so a Modbus charger's readings card and a charger the cloud says nothing more about are unchanged. Naming one of these fields is a later commit, made from evidence; inventing a label for it today would only make a guess look settled. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
90558d60b2 |
What the app can set, the cloud connection can set
The broker transport could move a session along — start, stop, boost, skip the delay, cap the current — and nothing else. Everything the charger is actually configured with sat one field away in the same messages we were already decoding: the schedule it charges on, the plug lock, auto-start, the LED, load balancing, solar charging, and the Modbus server the local transport depends on. Readable, and unreachable. The obstacle was never the cloud, it was the shape of the protocol. A setting is not a register write. It is a *command*, and a command owns a set of fields inside a message type — mostly one, but five own several, and the charger reads the whole command as the new truth. A light-off schedule sent carrying only its switch is a schedule whose start and end have just been set to midnight. So a grouped write resends the siblings the caller did not name, using the values the charger itself last reported, and refuses when it has never reported them. That last part is not caution for its own sake: load balancing and solar charging carry the serial of the meter they watch, and nothing outside the charger knows it. An empty one would be adopted. Those values do not arrive with the telemetry, either. The fast 0410 stream a realtime trigger turns on carries none of them — the settings come on 0405, 0840 and 0900, which the charger sends when it has something to acknowledge. So a grouped write may have to send a trigger first purely to make the charger talk about itself, and says so plainly when even that produces nothing. Everything a caller supplies is encoded before the cloud is touched at all. A request naming one bad value changes nothing rather than half of what it asked for, and a mistyped setting costs a validation error instead of a sign-in, a certificate fetch and a broker connection to be told no. mqttsettings.go holds one table and it is the only place a setting is defined: the wire field, the name a caller uses, the state key its current value comes from, and how a value becomes bytes. The names are the snapshot's own, so a caller can read a status, change one entry and send it back. The existing limit command now builds its frame from that table too rather than encoding field a8 a second time. Reading grew to match. The frame decoder gains the fields the grouped writes must carry back — the two load-balance settings, both monitor serials, the solar monitoring mode — plus the swipe gestures, and the snapshot exposes the rest of what is now writable. One name was wrong and is corrected: field d9 was called chargingMode after the Modbus register at 20088, but the reference has it as the solar charging mode, so it becomes solarChargeMode and moves in beside the solar settings. A mislabelled reading is bad; a mislabelled writable field is worse. Over HTTP it is one action rather than a dozen, because the charger groups the fields anyway: POST .../settings with a settings object, and settings sharing a command travel in one frame instead of overwriting each other. The other two transports refuse it by name and say which one has it, the way they already refuse each other's commands. The audit trail records the values, not just that a write happened — a setting that changes what the charger will draw, or whether it answers on the LAN at all, is worth being able to trace afterwards. Two things worth saying plainly. This is built from the reference project's message maps and checked against its own frame layout, not against hardware — there is no charger on this end to point it at. And modbusEnabled is a loaded gun: writing it off stops the charger serving the register map, and the way back is this transport, or the app. The ignore rule for the local Modbus map artifact widens to the protocol maps that now sit beside it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
576df58776 |
Go the way the owner's phone already goes
Control had two transports and neither fitted the ordinary customer. OCPP waits for the charger to dial in, which needs a public endpoint it can reach, a certificate, and a firmware willing to talk to our CSMS. Modbus TCP dials the charger, which needs the server on the charger's own network. Between them they cover a charger we host and a charger we stand next to; the common case is a charger behind someone else's router, and that had nothing. It was never unreachable, though. The charger holds a connection open to Anker's own broker — it is how the mobile app drives it from anywhere, and it is the mqttStatus register the Modbus snapshot has been reporting all along. So a third control mode joins that broker as the account: get_user_mqtt_info issues a client certificate, mTLS to aiot-mqtt-eu.anker.com:8883, and commands go out on the same topics the app publishes on. Nothing on the customer's side has to be forwarded, addressed or certificated. What travels is not an API call. The payload is a JSON envelope around a base64 binary frame the device itself speaks — marker, little-endian length, message type, name/length/type/value fields, XOR checksum — so mqttframe.go is a codec rather than a client, written from the message maps in anker-solix-api and anchored on the one frame that project documents byte for byte. A frame whose fields do not tile exactly up to the checksum is refused rather than half-read: these arrive over a link we do not control, and a truncated frame must not read as a charger reporting zeros. Two of the charger's habits shape the rest. It publishes nothing unless asked, so a status read arms a telemetry trigger and waits for the next frame, and a poll inside that window answers from what has since arrived. And a broker connection costs a fetched certificate and a TLS handshake while the plugin manager builds a throwaway instance per request — so the connection lives on the account's shared session beside the auth token, for exactly the reason the token lives there, and closes itself after five idle minutes. The transport also sees two signals no other one does: the boost flag, and the plug and start countdowns. The package doc has said since the first commit that they are never set and the derived mode must do without them. Here they are set, so a charger that has been told to start and is counting down a delay says so rather than sitting in "preparing", and "skip the delay" is offered only while there is a delay to skip. The clients generalise instead of growing a second layout. Both snapshots name the same quantities the same way, so what was Modbus-only in the readouts is now whichever transport read the charger — ModbusStatus becomes ChargerStatus on the phone, mb becomes dev on the web. What each transport can be *told* still differs, and the buttons branch on that: reset and clear-limit stay with OCPP, the timeout and phase registers with Modbus, skip-delay with the cloud. A command a transport has no equivalent for is refused by name, saying which one has it. The cost is worth saying plainly. This leans on Anker's cloud being up and on an unofficial protocol the app may change under us, where Modbus leans on nothing but the LAN. And it is checked against the reference implementation's own worked example rather than against hardware — there is no charger on this end to point it at. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
f3235c403c |
A page opens where you put its tabs
Every tab bar in the app drags into the order you want, and then all three of them opened on a tab picked in the source anyway: "public" on Charging, "info" on a car, "personal" in Settings. Dragging Home chargers to the front of the charging bar rearranged the bar and changed nothing about where the page landed, which is the opposite of what dragging it there says. So the front of the bar is now the landing tab, everywhere. An arrangement is already the statement of what you want to see first; it just wasn't being read as one. Settings > Appearance overrides it per page for the case where reading order and landing tab are two different wishes, with "First in the bar" as the default and the meaning of no override at all. The rule lives in one place, lib/tabs.js, because it is one rule and three pages: the saved choice if that tab is actually on the bar, otherwise whatever leads it. The bar it is given is the one that will really render, hidden tabs and inapplicable ones already dropped, so a default that no longer has a button - a tab switched off for that car, Users on a non-admin - falls back to the front instead of opening nothing. The tab key lists moved there too, since the picker needs all three and would otherwise have copied them. Each page starts on no tab and keeps following the profile until the user says otherwise, rather than guessing and then correcting itself: the arrangement and the default both arrive with /api/me, which on a hard refresh lands after the view has mounted. A click ends the following, and so does the start of a drag - rearranging a bar must not pull the content out from under the pointer. In Settings ?tab= still wins over both, since that is what /admin redirects to. Stored as defaultTabs on the profile, one page->tab map validated per page: a tab that exists but on another page is an error, and an empty value is stored as an absent key so "no default" has a single representation. Also adds charger_tab_order and charger_card_order to the PocketBase setup script. They were never there - the arrangements of the last two commits had no column to persist into on a freshly set-up server - and default_tabs would have gone the same way beside them. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
3c64d6e84c |
The cards drag too, held by their headings
Same arrangement the provider panel gives its readings, applied to the four charging cards: drag one and the column reorders live under the pointer, the card being dragged goes half-transparent, the one it is over takes a ring, and the rail's lock holds the lot still. Two things differ from the readings row, both because these are four different things rather than four of one. A card is placed with the CSS order property instead of by moving markup, so each keeps its own template and its own v-if. And the handle is the card's heading rather than the whole card — a card that was draggable everywhere would fight the current-limit slider and the address fields for the pointer. Saved on the profile as charger_card_order, beside the tab order. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
1418a566fd |
The charging tabs drag, like a car's do
Reordering the bar meant editing the template, which is a poor way to ask for Home chargers first. The tabs now drag into either order on the same native drag events as a car's tabs and the garage, down to the live reorder as the pointer crosses a tab, the grab cursor, and the rail's lock holding the bar still for anyone who would rather not nudge it on the way to a tab. The arrangement is saved on the profile as charger_tab_order, beside the garage order and for the same reason: it is a layout choice that should follow the account rather than the browser, unlike which cards are folded. The field is reconciled onto the users collection at boot, so no migration step. Its normalizer is the garage's, which now takes the field name and cap as arguments instead of being copied. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
62cb691f98 |
Everything the register map carries, sorted the way it gets asked about
The Modbus snapshot reported about half of what one poll already brings back. The rest was read into the block and thrown away: line-to-line voltages, reactive and apparent power per phase, the PWM flag, the control-pilot voltage, and the identity block's product number, rated power and current range. All of it now decodes — no extra requests, the registers were in hand already. Added alongside it: the control block, read back over FC03. It answers a question the live registers cannot, which is what the charger is *set* to as opposed to what it is doing — a boost that was asked for reads there while the live block still reports none running. Best effort, so a charger that refuses it still reports its state. Two registers the spec leaves blank are decoded on the hardware's evidence. The control-pilot voltage reads 11873 while the CP signal register reports state A, which that enum names as 12 V, so the register is millivolts. The identity block's current range is in amps, whatever its unit column says about watts and kVA. The charging card lays this out in sections rather than a wall of forty numbers: per-phase measurements as the matrix they are, then live state, then settings, then the device itself, with alarms surfacing only when a word is non-zero. Strings in en/da/pl. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
aaa89dfe10 |
Relays that run at 33 degrees, not 331
The two relay temperatures came back as 331 and 319 from a charger sitting idle with nothing plugged in. The spec's gain column says 1 for both, so we reported them as 331 °C and 319 °C — a reading that would have meant a fire rather than a wallbox at room temperature. The gain is 10. The same table hands the maximum current setting a unit of watts and the timeout a unit of amps, so its unit and gain columns are not load-bearing here; what settles the alignment is the LED brightness two registers earlier, which reads exactly 100 at gain 1, and the fact that the neighbouring registers all decode as tabulated. Read back from the charger afterwards: 33.1 °C and 31.9 °C. The field becomes a float, as the voltages and currents beside it already are. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
cf4fd14b56 |
The table the charger actually keeps its measurements in
Modbus mode never returned a reading: every status poll came back as "the charger did not answer", though the charger was answering all along. It was refusing the question. The A5191 splits its map across two tables where the spec's single 2xxxx column suggests one — 20000-20100 are input registers and reject FC03 with an illegal-address exception at every address in the range, while 21000-21005 really are holding registers and read back over FC03. We inferred one space from the spec's layout and asked for all of it with FC03. The client learns FC04, sharing a body with FC03 since the two differ only in which table the server consults, and the plugin's two measurement reads move to it. Writes stay on FC06, where the controls already live. Confirmed against an A5191 on firmware 1.0.6.1: identity, live block and the control registers all decode as the spec tabulates them. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
f7472bada3 |
Reach the charger where it is, instead of waiting for it to call
OCPP asks the charger to dial us: a public endpoint, a TLS certificate, and a route in through the customer's router. Our own handler then demanded two more things the V1 does not offer — TLS on a charger that connects over ws://, and Basic auth credentials the Anker app has no field for — so every connection was turned away before the upgrade. Anker publishes a Modbus TCP register map for this charger, and it inverts the problem: we dial the charger, on its own network, with no inbound reachability to arrange. That works for a charger behind a router that OCPP cannot reach at all. internal/modbus is the protocol, hand-rolled against the spec like the MQTT and WebSocket clients beside it. The plugin's modbus.go is the V1's map: the same 0-8 status enum the cloud already reports, per-phase measurements, and the writable registers behind start, stop, current limit, boost and phase mode. A new "modbus" control mode routes the existing control endpoints down it, so the REST surface, the rate limit, the confirmation step and the audit trail are the ones already there. The commands the register map has no equivalent for say so by name rather than failing as unknown, and a current below the charger's 6 A floor is refused because it pauses the charge rather than slowing it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
ad785ee9f8 |
The fields the cloud sends, kept all the way to the card
Normalizing a provider's charger list threw most of the answer away: firmware, the site id, how the charger is registered, the charge power and the cloud's own OCPP reading all arrived from Anker and none of them got past providerCharger, which carried seven fields and dropped the rest. The information card could not show what it was never handed. It carries them now, and the card lays them out: firmware beside the model, site and site id where the charger lives, "Registered as" for standalone / in a system / bound, and — when the service knows — state, charge power and OCPP status. Charge power is relayed exactly as worded upstream, since the unit is theirs and putting one on it here would be inventing it. Settings' own list gains the firmware in its subtitle. A field no view supplied still leaves no row, so an account whose chargers stand outside a system reads shorter rather than emptier. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
d71c1b4691 |
The panel header stops folding "Sign out" onto two lines
The console shell was capped at max-w-4xl (896px), narrow enough that the header row ran out of room and the last button wrapped mid-word. It now uses the same 1368px cap as the Web App shell, so the header has the space it always assumed it had. Rebuilt the embedded dist to match. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
190ae923a6 |
The Anker token outlives the request that fetched it
The manager builds a throwaway plugin instance for every per-user call —
HealthCheckWith, InvokeWith, InvokeBatchWith each construct, Init, probe and
Shutdown. The auth token lived on that instance, so it died with the HTTP request
that fetched it: opening the Anker panel signed in once for the health probe and
again for the charger list, and a page that also asked for OCPP info signed in a
third time. Every refresh, a fresh login.
Anker throttles passport/login per IP per minute and answers code 26161 ("Failed
to request.") once tripped, so this is the shape of the failure the panel has been
reporting; the cloud has also historically kept one token per account, so each of
those logins could evict the one the mobile app was holding.
Tokens and the login backoff now live in a package-level session keyed by the
account signing in, so every instance configured for that account shares one
login. Re-configuring the same credentials keeps the token; a different account,
or the same account on the other regional server, gets its own session. Sessions
unused for a fortnight are pruned, so an edited password does not leave its entry
behind for the life of the process.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
tajniak81andClaude Opus 5
|
c1aee0fac1 |
One refused sign-in, not five: the chargers poll no longer locks the account
A chargers poll asks four cloud views. Each called apiRequest, each found no token, and each ran its own login — so a login Anker refuses was offered four times in one poll, and the next poll spent the fifth. Five is what disables the account for ten minutes, which is how "code 26161: Failed to request." turned into "your account has been disabled" on the very next attempt. The plugin now remembers a refused login instead of repeating it: the failure is cached and replayed to every caller until a backoff window passes — a minute at first, doubling to fifteen, or the full ten minutes when Anker says it has already locked the account (code 10019). New credentials clear it, so a fixed password is tried at once. chargerInventory signs in once up front. A login the cloud refuses is not four views failing, so it is reported as itself rather than as three warnings with the lockout notice buried in the last one. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
423bc2ab16 |
The charger list sorted by a field the collection never had
Opening Home chargers answered {"status":400,"message":"Something went wrong
while processing your request."} — PocketBase's generic refusal, here for an
unknown sort field. home_chargers declares its own fields and nothing else:
PocketBase adds no created field to a collection defined through the API, which
is exactly why control_audit and organizations declare theirs. The list handler
sorted by created anyway. It was the only handler in the server that sorts by
created — every other one sorts by name, km or date, fields their collections
actually declare — so the gap had never had a chance to show.
The field is now declared, and reconcile adds it to the collection already
standing on the next boot, since home_chargers is in reconcileOrder. Import order
is the only order a charger has: it carries no date of its own, and a wallbox
bolted to a wall does not accumulate events the way a car does.
The list also stops depending on that. A rejected sort now falls back to the
unsorted query rather than failing the request: the order is a nicety, the list
is not, and an owner reading a database error about a field they cannot see is
the worst of both. It also makes the deploy order stop mattering — the page works
before the bootstrap has run, and the sorted query wins once it has.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
tajniak81andClaude Opus 5
|
a3f69fa5ef |
Home chargers: your own wallbox as a record, imported the way a car is
The Home chargers tab has been showing a hardcoded "Home charger · 11 kW · NACS" since it was drawn, and the control card asked for a serial as free text — a number printed on a box hanging in a garage, typed in by hand while the connected account already knew it. The garage solved the same problem for cars a while ago, so this is that solution aimed at the wall: pick the charger off a service you have connected, press Import, and it becomes a record of yours. A charger is a record rather than a live listing because it has to outlive the account it came from. Disconnect Anker and the wallbox is still on the wall; the integration is how the charger was found, not what it is. Hence home_chargers, owned by a person and not related to any car — it charges whichever car is plugged into it, and it outlives all of them — and hence no sharing: a charger is one household's business in a way a car shared with a partner is not. The provider layer is vehicleproviders.go's shape on purpose, down to the soft gate: a listing answers 200 with an empty list and the sentence that says what to do about a closed gate, a write answers 400, because there the caller asked for something that did not happen. Anker and Greencell are two adapters over plugins that already exist, so the next charger service is an adapter appended to chargerSources() and nothing else. What is deliberately absent is the car import's checkbox panel: a charger is a name, a serial and the hardware behind it, all of which the list already carries, so there is nothing to choose and the whole screen is pick one, press Import. Only the name is editable afterwards. The rest describes hardware and came from the service, and the provider link is written by the import endpoint alone, so renaming a charger cannot quietly orphan it from the account it tracks. Deleting one says as much in its confirmation: the charger is untouched, and importing it again brings the record straight back. The Anker gate moved into ankerGate() beside greencellGate(), because the same four-case switch was about to exist in a third place. Behaviour is unchanged — the same sentences, and the probe still skips the personal opt-in, since checking credentials is what you do before switching the integration on. The phone app still has the old tab; parity there is a separate change. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
a809980d8b |
Anker health: count the chargers the panel lists, not the ones one endpoint admits to
The probe still asked get_user_bind_and_not_in_station_evchargers and read its userBindEvChargersCount, so it reported "0 EV charger(s) bound to account" for an account whose two chargers the panel was listing directly underneath — the same blind spot the capability was just moved off, left behind in the health check. It now takes the same inventory the chargers capability returns and counts that. Authenticated with nothing on the account is degraded rather than ok, following Greencell's rule: the half we address answers, and the empty half is the account or the country that picks the regional server, so the message says so instead of reporting a healthy connection to nothing. A count reached with some view missing says how many views stayed silent, because the number is then a floor rather than a total. The web panel colours degraded amber, as it already did for Greencell. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
e138fad3f4 |
Anker: every charger on the account, not just the ones outside a station
get_user_bind_and_not_in_station_evchargers is the only list the connector ever asked for, and its name says exactly what it withholds. A charger that belongs to a system is not in it. Its userBindEvChargersCount, though, counts every charger bound to the account — so an owner with two chargers in a system got "authenticated; 2 EV charger(s) bound to account" from the health probe and an empty list from the capability that is supposed to show them. A working login that finds nothing. So the capability now asks every view the cloud has and merges them by serial. The standalone list still answers for chargers standing on their own; get_site_list walks the systems and reads each one through get_scen_info, falling back to get_system_running_info where that is silent — the power-service / HES split charger-state already knows; and get_relate_and_bind_devices contributes model, firmware and the Wi-Fi flag, and discovers anything in the A519 family that the first two missed. Whichever way a charger was registered, one of the three has it. The merge is first-writer-wins per field rather than last view overwriting: the standalone record knows the name, the site record knows the live state, and neither should blank what the other established. A view that fails is a warning on the document instead of an error on the call, because one dead endpoint should not cost the chargers the other two found. Only losing all three is a failure. When nothing comes back at all the response says so in its own words and names the remaining suspect — country picks the regional server, and the wrong one authenticates happily and shows an empty account. The other half of "not showing any chargers" was that neither client ever showed a list. The serial was a text box, and the number is printed on a charger hanging on a wall. Both apps now list what the account holds — name, serial, model, site, state, an offline badge — and hand the serial to the OCPP control card instead of asking anyone to go and read it. Where control is off the list still stands on its own, as the answer to the first question an owner has after entering credentials. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
a7cab50e06 |
Apprise: a gateway to hand a message to, not a hundred protocols to carry
Apprise is a Python library that speaks 100+ notification services behind one URL grammar — mailto://, tgram://, ntfy://, discord://. None of that is portable to a server that takes no dependencies, and none of it needs to be: caronc/apprise-api wraps the library in HTTP and is meant to run as a container beside us. So the connector carries no notification protocols of its own. It posts a body to an endpoint the operator runs and lets Apprise fan it out, which is also why adding a service later costs nothing here. Targets are addressed one of two ways and configKey is the switch. Stateful means the URLs live on the Apprise server under a key, narrowed by a tag expression, and recipients are then edited there — no credential for any downstream service is ever held in DriverVault. Stateless means the URLs travel with the request, from a secret config field, which is simpler for one destination and worse for ten. A call that names its own key or urls takes that destination alone rather than merging with the configured one: honouring a caller's URLs while still falling back to the configured key would deliver the message somewhere nobody asked for. baseUrl is Required, which no other connector's address is. Toyota, Anker and Greencell leave everything blank at the global layer because the superadmin → org → user cascade exists to fill it in, and a blank there means "let the user choose". There is no cascade behind this one — a notification gateway is infrastructure the operator runs, not an account a driver owns — so nothing further down can supply the address, and a blank is simply a plugin that cannot work. Better to fail at enable than at the first notification nobody sees. Three limits are choices rather than gaps. /add and /del are not implemented: the Apprise config belongs to the operator, we post to it, and a connector that can delete a notification config has a wider blast radius than one that can only send through it. privacy=1 is forced on /json/urls rather than offered as a parameter, so a target listing reads mailto://user:****@host and downstream tokens stay on the Apprise side of the wire. Attachments are remote URLs the Apprise server fetches; multipart upload is the API's own path for files and not ours. Health follows the rule Greencell set. A reachable server whose config holds nothing to notify is degraded, not down: the half we address works and the missing half is the operator's config. Two cases earn their own line — a config key set against a server running with stateful mode disabled can never resolve, and /status answers 417 rather than 500 when Apprise finds a problem with itself, so that is a parsed answer and not a transport failure. A proxy that strips our Accept header gets the same codes back as plain text, which is read rather than called unreadable; an HTML error page from something that is not Apprise is not, and a test pins the difference. Notifications needed a category of their own, and that is the one change outside the plugin: the constant, the tab order in PluginsCard.vue, and the label in all three panel languages. The cost is now written down in the plugins README beside the Descriptor example, since the previous five categories predate anyone having to add a sixth. The plugin's tests run against an apprise-api stand-in built from that project's views.py — both notify paths, the override rules, 204-as-empty against 424-as-failure, and every health branch. builtin_test.go is the other half: the blank-import list in builtin.go is a silent failure mode, since a connector left out of it compiles, passes its own tests, and never appears in the panel. What is not covered is a live instance; there is no Docker on this machine, so the wire contract comes from reading upstream's source rather than from running it, and a smoke test against a real deployment is still worth doing. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
a49d48f659 |
The panel's webfonts, in the one position CSS accepts them
The lockup was the visible symptom and the wrong suspect. Matching it to the Web App's component changed nothing a reader would notice, because the panel was not rendering Archivo at all — it was rendering system-ui's italic bold, which is a different letterform at the same size, and had been since the stylesheet was written. The Google Fonts @import sat after @import "tailwindcss". Tailwind v4 inlines its import into the rules it generates, so anything importing after it is no longer at the top of the sheet, and CSS drops an @import that follows real rules. The built stylesheet carried zero occurrences of fonts.googleapis.com; the build had been saying so on every run, in a warning easy to read as noise about a comment. Moving the font import above Tailwind's is the whole fix, and the Web App's own stylesheet has always had that order with a comment explaining it — that comment comes across, plus what it cost here. This was never only the wordmark. Every rule reaching for --font-sans or --font-mono was falling back too, which is the entire panel: the section nav, the card titles, and the endpoint tables whose monospace is how a path reads as a path. Checked against the built bundle rather than the dev server, since the dev pipeline is exactly what was hiding it: the page now reports Archivo italic 800 loaded, and the wordmark measures 115.05px — the same width the Web App's rail lockup measures. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
e648634ce1 |
The plugin list, grouped by what a plugin actually is
Category has been in the plugin contract since it was written — apis-external, drives-external, drives-local — and every builtin declared the same one, so it grouped nothing. Two of the three talk to a wallbox and one talks to a car manufacturer, and those are different questions an operator arrives with: the Toyota card is where a driver's account gets linked, the Anker and Greencell cards are where a charger's broker and credentials live. So vehicles and chargers join the constants and the three builtins say which they are. The panel groups on that field rather than on a list of names, which is what keeps an external plugin from needing panel code. Tab order mirrors the constants; a category with nothing in it gets no tab, and a single group hides the bar entirely, so an install with one connector looks exactly as it did. A category the panel does not recognise — or an empty one — falls to the external-APIs tab rather than vanishing, because a plugin nobody can see is a plugin nobody can disable. The selected tab falls back to the first group when its own goes away, which is what removing the last external plugin does. Registration still asks only for name, base URL and provider, so a plugin registered at runtime lands under Other APIs until its manifest names a category. That path already works and is the honest default: the panel is guessing about a service it has never spoken to, and the service can say. The header lockup is the other half. It was a copy of the Web App's mark rather than the same mark, and copies drift — a 32px icon against 28, a 24px wordmark against 21.6, "Driver" at text-strong instead of white, "Vault" a step lighter than brand-400. The Web App's Logo.vue moves in verbatim, props included. The one thing it cannot inherit is which variant to render: the Web App's rail is always dark, while this panel flips with its own theme toggle, so on-dark is bound to the theme and the hand-rolled bar fills that existed to survive that flip are gone. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
340a81b0d6 |
Greencell: the charger on your own broker, not a cloud it never had
The HabuDen has no cloud API to connect to. It is commissioned over Bluetooth in
the Greencell GC app, pointed at an MQTT broker the owner runs, and from then on
publishes there — so the connector is an MQTT client rather than an HTTP one,
and nothing in it reaches Greencell. The wire contract is Home Assistant's own
greencell component and the greencell_client 1.0.3 library beneath it, which is
the only published description of the topics: a BROADCAST on /greencell/broadcast
draws device announcements, and /greencell/evse/{sn}/ carries current in
milliamps, voltage, power under "momentary", the EVSE state, and the access level
chosen in the app.
That meant an MQTT client, and the server takes no dependencies, so internal/mqtt
is hand-rolled the way internal/ocpp's RFC 6455 layer is. It is scoped to what
this connector needs and says so: QoS 0 for everything we send, clean session,
no reconnect — a connection lives for one plugin call, which is exactly how the
manager builds and tears down an instance. Inbound PUBLISH is accepted at QoS 0,
1 and 2 with the acknowledgements each requires, because the QoS of a delivery is
the broker's choice and not ours; an unacknowledged QoS 1 is redelivered forever.
Read-only, and the reason is worth writing down rather than rediscovering. A
device in EXECUTE mode accepts START, STOP, SET_CURRENT and QUERY — but the topic
those go to appears in no source: not Greencell's integration page, not
greencell_client, and Home Assistant ships sensor-only for that same reason.
Publishing to a guessed topic would be a control feature whose failure mode is a
driver believing they stopped a charge. So the access level is reported, and
commandTopic is the seam: an operator who has watched their own broker and found
theirs sets it, and a state read then sends QUERY — the one command a READ-mode
device also honours — instead of waiting out the charger's publish cadence. The
day the topic is public, control is a payload away from the same field.
What the cascade resolves here is a broker, not an account, so host, port, TLS and
credentials resolve together from the highest layer that names a host: an
organization's address paired with a user's password would address a broker with
credentials never meant for it. The serial, the QUERY topic and the listen window
each describe the charger rather than the endpoint, so each resolves on its own.
Two reading rules the tests pin. A phase the device did not report stays nil
rather than zero, because zero amps on a charger is a real measurement — a JSON
null decoding to 0.0 was a live bug until a test caught it — and a partial read
returns with received/complete flags instead of failing, since a device that
publishes some topics on a slower cadence is still worth reading. And a reachable
broker with no charger on it is degraded, not down: the half we configure works
and the missing half is the device. The plugin's end-to-end tests run against an
in-process broker written to the raw wire format, so a bug in the client cannot
hide behind a matching bug in the fixture.
The apps get the third connector card. The panel needed nothing — it renders a
plugin's ConfigFields itself — but the per-user panes are still hand-written per
integration, which is now three near-copies and the argument for the generic
version already noted in the plugins README. The web form splits the broker from
the charger because the server resolves them differently. The phone card is a
declarative config against the shared widget, which gained a number field type, a
degraded state that reads amber rather than red, and a fix for a locked field
that was covering its own displayed value with dots. Twenty keys in three
languages across both apps; Greencell, HabuDen and the literal QUERY join the
proper nouns that stay in English.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
tajniak81andClaude Opus 5
|
5e6b8b4b1c |
Anker Solix: the charger's mode, and the modes it can be moved into
The connector was written against anker-solix-api v3.7.0 and upstream is at 3.8.1 now. The reassuring half of the check first: nothing we depend on moved. The passport/login ECDH exchange, the headers, and every endpoint path this plugin calls are identical across v3.7.0...v3.8.1 — the only apitypes movement touching an EV charger was get_device_rfid_cards being reordered within its own dict. The 400 new lines in charger.py are the A2345 USB charger, which shares a filename with our device and nothing else. What did land for the V1 is two entries in the release notes, and both are MQTT: 3.8.0 gave standalone chargers the usage-mode entity they were missing, 3.8.1 added a switch that reads those modes as a plain on/off so EVCC and its like have a binary to hold. We control chargers over OCPP, not MQTT, so the command path is not ours to port. The reading of state underneath it is, and that half does come over the cloud. So charger-state. The status code arrives under two different names depending on which system family a site belongs to — operating_state inside a scene's charging_pile_list, evChargerStatus inside HES system running info — and upstream's poller quietly renames both to ev_charger_status on ingest, which is the tell that they are the same number. We ask both and merge, because a site answering only one of them is the normal case rather than a fault; the call fails only when neither view is there. chargerMode and chargerModeOptions then follow ev_charger_mode_state and ev_charger_mode_options as written, including the rule that a stopped charger is startable only from standby, and the binary is the same one 3.8.1 chose: everything that is not stop_charge counts as on. The gap worth naming is that the boost flag and the plug and start countdowns reach upstream over MQTT and never over the cloud, so three of the six modes cannot occur here. That is not a bug to be found later — chargerMode takes them as parameters and the callers pass their zero values, so the day an MQTT source exists the derivation is already correct and only its inputs change. The package doc says so in the scope list beside the other limits. Five endpoints upstream has had all along and we never exposed come with it, all EV-charger-scoped: the site scene, energy_analysis under device_type ev_charger, a charger's RFID cards, Anker's own OCPP endpoint list, and one vehicle's details. charger-status takes the featuretype it was hardcoding at 1, since upstream's exporter asks for both 1 and 2 and there was never a reason for us to see only half. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
fd75833707 |
The cabin's temperature, and the one it is heading for
The climate cards landed with the endpoint migration, but only as two more folded dumps of key/value pairs. What a driver opens that tab for in January is one number, and it was three taps down inside a card called Climate. So currentTemperature and targetTemperature join the headline readings, beside the pair of electric ranges and for the same stated reason: neither figure answers the question on its own. A cabin at 12° means nothing until you know it is climbing towards 21°, and the gap between them is how long to leave the scraper in the boot. Being derived from headlineMetricSpecs, both are arrangeable the moment they exist — a car's saved order of readings can name them without anything else being told they are there, and a test now says so rather than leaving it to be noticed when a PATCH starts rejecting a key. The unit is fixed at Celsius, because Toyota Connected is the European service and there is no imperial reading to convert from. That is a default and not a claim: a payload that names its own unit is still believed over it, the way every other reading here works, so a service that one day reports Fahrenheit is labelled Fahrenheit rather than relabelled into a wrong Celsius. The two apps needed the two labels in three languages each and nothing else. That is the shape working: a section is an id the app localizes and a reading is a key it localizes, so a card added on the server arrives in both clients already folded, already arrangeable, already translated. The one thing the Web App did need was a corrected comment — the note explaining why cards fold still said Toyota reports eight sections, and it is the argument for folding them, so it should count the ten there now are. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
22a22ec43a |
Toyota: the status route the car answers, not the one it retired
Toyota put the /v1/global/remote read routes behind AWS SigV4 in mid-2026. A bearer token is no longer a credential there, so the doors-and-windows card has been asking a gateway that answers 403 — the one section of the provider tab that could only ever have been in error. The MyToyota app reads that state from /v1/vehicle/status now and pytoyoda followed it in 5.2.0; so does the connector. The electric route did not move, and the comment above the endpoint block says which of the two namespaces each one lives in, because the obvious tidy — sweep the rest onto /v1/vehicle/* — would break the ones that still work. The same migration gave the climate reads a home worth porting: /v1/vehicle/ climate-status is what the cabin is doing, climate-settings the preset it was told to do it at. Both are GETs with a vin, both are new cards on the tab, and their headings are in all three languages on both apps. Nothing about the tab's plumbing changed to hold them — a section is an id, an action, and whatever JSON comes back, which is the point of that shape. Left where they are: the POST wake calls. Upstream refreshes a stale reading by waking the modem, and this connector is documented as read-only, so climate and status show what the car last reported rather than what it would say if asked twice. The cost is a reading that can be hours old, and it is the honest one to pay for a connector that promises not to touch the vehicle. Two tests keep the migration from being undone by hand: one fails if any advertised capability points back at a retired route, the other if a capability is advertised without being wired into Invoke, which is the way the next endpoint would go missing. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
fa569c4030 |
Sign in before the first call, not after a 401 that never comes
Logging in failed on both the Web App and the Phone App with PocketBase's
{"data":{},"message":"The requested resource wasn't found.","status":404}.
The login itself succeeded; the profile fetch right after it — GET /api/me
— was what 404ed, and the web app renders the relayed body on the login
form, so it read as a rejected sign-in.
PocketBase answers a record read it will not allow with 404 rather than
401: it hides the record instead of refusing the credentials. The pb
client only re-authenticated on a 401, so with no cached token the first
request went out carrying no Authorization header at all, came back 404,
and the retry never fired. Nothing ever tried again — the server kept
404ing long after PocketBase was healthy.
The cache is empty in exactly the two cases that matter: a startup where
the up-front Authenticate failed because PocketBase wasn't up yet, which
main.go treats as non-fatal on purpose so a superadmin can still log in
and fix the connection; and a Reconfigure from the panel, which clears
the token so the new credentials get used.
So acquire the token before the first attempt rather than hoping for a
401 to prompt it, across all four superuser paths. Bad credentials now
surface as the authentication failure they are instead of masquerading
as a missing record. With no service account configured there is nothing
to acquire and the call proceeds as before, since the endpoints that need
superuser access already answer 503 on their own.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
tajniak81andClaude Opus 5
|
35e6c511b7 |
Changed parts: the list is the car's, not the app's
The Changed parts section offered all three parts to every car. An EV changes no
oil, and a checkbox nobody will ever tick is one more thing to read past on every
service — so which parts a car records now belongs to the car, the same way its
tabs, its Information rows and its Service history columns already do.
It works the way those three do because a fourth mechanism for the same idea
would be a fourth to keep in step: hidden_service_parts on the car, validated by
the endpoint that already does this, stored as the hidden set so a part added in
a later release is on by default, and needing write access because the choice
belongs to the car and everyone it is shared with sees it.
There is no order beside it, which is the one place this departs from the other
three. Those arrange things whose position means something — a tab bar reads left
to right, a table's columns are read across. The parts are a checkbox list inside
a single column, and moving Cabin air filter above Oil says nothing. Adding one
later is the same shape as the others if that turns out to be wrong.
A part switched off leaves the form and the history together — the chips on the
phone's cards, the web column's summary and the panel it opens. "I don't record
this" means it stops taking up room, not that it takes up room saying nothing,
which is the rule a hidden column already follows. That is the judgment call
here: a car with five years of oil changes hides them all by switching the part
off. Nothing is written to the records, so switching it back on brings every one
of those chips back, which is what makes the call safe to reverse.
The part that would have been a silent data bug: the API rewrites all three
booleans from the body of a service update, so a form that simply stopped
sending a hidden part would set it false on the next edit of any old record.
Both forms therefore keep every part in their state and submit every one — only
the checkboxes are filtered. The mirror of that is a *new* record, where a hidden
part starts false rather than at its `initial`, since ticking a box nobody was
shown is not a default, it's a guess. Oil is the only part with initial: true, so
that case is live the moment anyone hides it.
Verified: go vet and go test ./... pass, with a new test covering that every part
is hideable (unlike the tabs and the columns — a service that changed nothing is
a real service), that the "parts" column key is refused as a part key and a part
key as a column key, and that no part is also a column. flutter analyze is clean
and flutter test passes 32 to 35, the new ones covering visibleParts, that a
hidden part's chips go while its stored boolean stays, and the picker's fourth
section. npm run build is clean.
Both apps were driven against throwaway stub APIs. Web: the picker saved
{"hiddenServiceParts":["oil"]}, the table's parts cell went from "Oil & Oil
filter +2" to "Engine air filter, Cabin air filter", the record whose only part
was oil went to an empty cell, the panel dropped to two rows, the add form
offered two unticked boxes where oil's initial: true would have ticked one, and
editing the three-part record sent changedOil:true back with a box that was never
on screen. Phone: the same car rendered chips "Engine air, Cabin air", "Changed
parts —" for the oil-only record, and an add sheet with exactly two unticked
boxes.
Not verified: no automated test guards the web behaviour — the web app still has
no test runner, so the above was read out of the live DOM and the outgoing
request bodies by hand. The phone's picker was checked by widget test and by
rendering, but its Save was not driven end to end. Neither app was run against
the real API Server: bootstrap appends the new field on the next start, and until
that start a client sending hiddenServiceParts takes a 400 — they deploy together
from this repo, but the server must go first.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
tajniak81andClaude Opus 5
|
d4033dbcef |
A build date you may only half know; one look for an empty cell
Two changes, both about showing what is actually known rather than a tidier version of it. The build date asked for a day. A car's build date is often only a year, or a month and a year - the VIN plate is stamped with a month, the papers carry a day, a grey import neither - so a field insisting on all three is answered either with an invented day or with nothing, and both throw away what the owner did know. The field now picks its own precision: a full date, a month and year, or a year, each with the control that suits it. A year is typed rather than picked, because a date picker that makes you walk back to 1998 is worse than four keystrokes. Stored as the ISO prefix - "2015", "2015-03", "2015-03-10" - which is ISO 8601 reduced precision, and printed back at exactly that precision. The three shapes sort and compare as strings in date order, which is why the prefix is stored rather than a date with a precision field beside it. The formatter takes the string apart rather than parsing it: "2015-03" read as a UTC instant and printed in local time hands back February west of Greenwich. Narrowing the precision keeps what is still true, so a day dropped from "2015-03-10" leaves "2015-03". Widening clears the field. That is the awkward half of the control and it is deliberate: there is nothing to widen a year with, and leaving "2015" behind an empty month box would store a date the screen is not showing. The column was free text with no validation at all, which was tolerable while only a date picker could write it and is not now that three shapes are legal. normalizeBuildDate parses rather than pattern-matches, so "2015-13" and "2015-02-31" are refused instead of stored as something no reader can print. The phone needed changing to avoid destroying this. It parsed buildDate with DateTime.tryParse, which returns null for "2015" - so a half-known date would have shown as a dash, and saving the car from the phone would have written "" back over it. It holds both date fields as the string they arrived as now, prints them at their own precision, and hands back anything it cannot set. Its picker still only makes full dates; a precision control there is a separate job. Separately: an empty cell of the service table had three different looks in one row. The dash under Notes was body-coloured, as though it were content; the one under File was 12px, having borrowed the size of the Download button that would otherwise be there; the one under Changed parts was muted at 14px. They are one constant now, muted at the row's own size, which is what Next date and Next km already did for a missing value. The Download link keeps its own styling - it is an action, not a value. Verified in a browser: a stored "2015-03" loads as month precision in a month picker, month to year narrows to "2015", year to day clears, "19x98abc" typed into the year box sanitises to "1998", saving sends buildDate:"1998" and the Information tab then reads "1998" - while a full first-registration date beside it still reads 06-08-2026. All five empty cells across the three columns now compute to the same size, colour and weight, with the filled ones unchanged. go vet and go test ./... pass with a new test over the three valid shapes and six rejects; flutter analyze is clean and 22 tests pass, one new, covering a half-known date in two date formats and the time zone that could shift it; npm run build is clean. Not verified: First registration still demands a full date. The same argument applies to it and the field is now a reusable component, but it was not asked for and is one line away. The web formatter's month-name paths - the DMY and MDY formats, which spell the month out - are covered only by the phone's mirror of the logic, the web app still having no test runner. A car created through the Toyota import bypasses the new validation; it only ever produces full dates, so nothing invalid gets in that way, but it is not guarded. Both apps need redeploying before any of this is visible. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
c5d431c560 |
Changed parts: one column in the service table, not one each
Oil & Oil filter, Engine air filter and Cabin air filter had a column of Yes/No
each in the Service history table, 375px of the 1022px table between them for
three bits of information. The form has always kept them together in one Changed
parts section, which is the honest shape: they are one answer to one question
about a service, not three unrelated readings. The table said otherwise, and the
list is going to grow - every part added would have taken another column and
pushed the table into a sideways scroll.
They are one column now, 234px with the widest summary on screen, and its width
no longer depends on how many parts exist. The cell names what was changed
rather than counting it, because a history is read down the page and "2" tells
you nothing about which two; past two names it becomes the first part and a
tally, which is what keeps one line one line as the list grows. Nothing changed
reads as an em dash.
The detail is a dropdown, not a dialog. This is read-only detail about one row
of a table you are reading down: a modal would black out the rows being compared
against and charge an open-and-close for each one. It is pinned under the button
it was opened from, closes on an outside click, Escape or a scroll - it is fixed
to a point on the screen, so a table that moves underneath would leave it
pointing at the wrong row - and there is one panel rather than one per row. It
lists every part with a Yes or a No, the unchanged ones included, so the em-dash
row still answers the question instead of being a dead cell.
One list in lib/serviceParts.js now drives the form's checkboxes, the cell's
summary and the panel. That is the point of the change as much as the width is:
adding a part was three edits that had to agree, and is now one entry plus its
boolean on the API's service_records collection. The form builds its state and
its payload from the list rather than naming the three fields twice - the save
payload is unchanged in shape, which was checked against the wire rather than by
reading it.
This walks back part of the previous commit, which had just made all three
hideable separately: the server's column set drops oil/engineFilter/cabinFilter
for a single "parts" key, and a test now asserts those three are not columns of
their own, so the table cannot drift back. A car with ["oil"] stored as hidden
would quietly get the combined column - nothing has that stored, the deployed
stack predating the feature, and stale keys are dropped on read rather than
erroring.
Verified in a browser against a stub API: all four summary cases (one part
named, two named, three as "Oil & Oil filter +2", none as an em dash); the panel
opens anchored under its button with the right Yes/No for the row, stays inside
the window, and closes on outside click, Escape, scroll and a second click,
switching rows without leaving a second panel behind; the picker offers "Changed
parts" as one entry and hiding it sends {"hiddenServiceColumns":["parts"]};
dragging sends "parts" in the order with the hidden column holding its slot; the
Edit dialog renders from the shared list and its PATCH still carries all three
booleans with the unticked one false. go vet, go test ./... and npm run build
are clean.
Not verified: no automated test covers any of it - the web app still has no test
runner, so the cases above were driven by hand. The drag and the panel were
exercised through dispatched events rather than a pointer, the browser pane not
compositing, so the native drag image and the panel's behaviour under a real
click-and-hold are unchecked. The dropdown overlaps the rows beneath it, which
is what a dropdown does but was not weighed against a taller table. The deployed
Web App still shows three columns until it is redeployed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
tajniak81andClaude Opus 5
|
b60d929ed6 |
Service history: columns you can switch off and rearrange
A car's page has let you choose and arrange two things for a while - which tabs
it shows, and which rows the Information tab lists, both dragged into whatever
order you like. The Service history table was left out of that: nine columns,
hardcoded, in one order, for every car. An EV shows Oil & Oil filter and Engine
air filter on every row of a history that will never record either, and a reader
who mostly wants Notes has to look past four columns of dates and distances to
reach it.
It works the way the other two do, because a third mechanism for the same idea
would be one to keep in step. Both lists are properties of the car, so everyone
it is shared with sees the same table, and both need write access to set. The
columns are stored as the hidden set rather than the visible one, so a column
added in a later release is on by default. The arrangement covers the hidden
columns too, which is what makes a column switched back on return to where it
was instead of reappearing at the end - verified below, since that is the part
of this shape that is easy to get wrong and invisible until somebody hits it.
Date cannot be switched off. Every row of that table is work done on a day, and
a history with the day taken out stops being a history; it can still be dragged
anywhere, which is exactly the rule Information already follows in the tab bar.
That is a judgment call and the annotation that prompted this only circled the
other eight columns - moving "date" into hideableServiceColumns and dropping the
filter in the picker would reverse it in two lines if it turns out to be wrong.
Server: hidden_service_columns and service_column_order on the car, validated
against their own key sets by the endpoint that already does this for tabs,
fields and readings. The arrangeable set is derived from the hideable one plus
the date rather than written out again, so the two cannot drift as columns are
added. Bootstrap appends missing fields to existing collections, so the two
columns appear on the next server start with no migration to run.
Web: the table stopped being nine hardcoded th/td pairs and is now driven by one
list of columns, head and body from the same source, which is what stops a moved
or hidden column from shifting the headings out of line with the cells. The
cells are built a row at a time rather than a call per cell, so a long history
doesn't rebuild every cell three times to read its text, its classes and whether
it is the file column. The column headings kept their existing car.services.col*
translations - the keys are mapped rather than derived, because renaming a dozen
strings in three languages to save a lookup table would be the wrong trade. Four
new strings in all three languages.
Verified: go vet and go test ./... pass, with new tests covering both key sets -
that hiding the date is refused, that a field key is not a column key, and that
the arrangeable set is the hideable one plus the date. npm run build is clean.
The page itself was driven in a browser against a throwaway stub API: the
rewritten table renders identically to the hardcoded one, switching two columns
off removed exactly those two from head and body with the rest still aligned and
sent {"hiddenServiceColumns":["oil","engineFilter"]}, dragging Notes onto Km
reordered head and body live and saved an order with the hidden columns still
holding their places, switching Oil back on returned it between Next km and
Cabin air filter rather than to the end, and a read-only share gets no gear
button, no draggable headings and no drag hint.
Not verified: the drag was exercised by dispatching drag events at the
component's own handlers, not by a pointer - the browser pane was not
compositing, which rules out both screenshots and a real drag - so the native
drag image and cursor are unchecked. No automated test guards any of the web
behaviour; the web app still has no test runner. The API rejects unknown JSON
fields, so this web build against an older API Server would take a 400 when
saving the picker: they deploy together from this repo, but one must not ship
without the other. The phone app is deliberately untouched, having no column
table to arrange, and ignores both new fields.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
tajniak81andClaude Opus 5
|
b4e99240c6 |
API panel: document the whole REST surface, not half of it
The API section's tables are hand-maintained and carry a comment saying they
mirror the routes in server.go. They had stopped: 60 rows against 141 routes,
so 81 endpoints were reachable and undocumented. Everything added since the
tables were written is in that gap - technical checks, fuel, charging,
maintenance, documents, reminders, attachments, the vehicle-provider and
integration surfaces, and the OCPP socket - along with the per-car
sub-resources (fuel-stats, charging-stats, the provider routes) and
PUT /api/cars/{id}/view. Ten new groups, 60 rows to 123, and nothing listed
that no longer exists.
Checked by extracting every mux.Handle route from server.go and diffing it
against every row in the tables rather than by reading both lists: the seven
attachmentRoutes calls expand to their 21 concrete routes, the three static
panel paths are excluded, and the diff is empty in both directions. That
script is not committed - it is a one-off, and a real guard belongs in the Go
tests where it can see the mux, which is worth doing if these tables drift
again.
The flat list endpoints require ?car={id} - they refuse a cross-car listing
so access can be enforced - so the ones that do now say it. Attachments are
documented once with {records} standing for the seven collections that take a
file, because all seven behave identically and seven copies of three rows
would bury that. EndpointTable grew an optional note line under the header to
say what {records} means.
Two things fixed while in there. PUT rendered uncolored - methodClass has had
no PUT entry since the superadmin group started using it - and the longest
paths, the per-charger OCPP control routes, overran the card and were clipped
by its overflow-hidden rather than scrolling; the table sits in an
overflow-x-auto wrapper now. Thirteen of the eighteen tables scroll at phone
width, four of them before this commit's rows were added.
Group titles are translated in all three languages, plus one new auth label
for the OCPP socket, which authenticates the charger with OCPP Basic auth
rather than a bearer token. The endpoint descriptions stay English:
TRANSLATIONS.md calls them developer reference documentation, and this commit
does not reopen that.
Verified by go build ./... and go vet ./internal/api, and by driving the
panel in a browser against a mock API Server standing in for the real one -
all eighteen tables render, the {records} note and the PUT color show, the
Polish titles resolve, and no table clips at 375px. dist is rebuilt so the
embedded panel matches the source.
Not verified: nothing ran against a live API Server, so the descriptions are
checked against the handlers' code and comments rather than against
responses. The mock only answers /api/identity and /api/status, which is
enough to get past the login gate and reach the section.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
tajniak81andClaude Opus 5
|
a25b31842d |
Round the connected service's readings; keep sheet buttons off the nav bar
Both found by driving the installed app on a phone rather than by reading the code, which is worth noting: the second one is invisible in a simulator with gesture navigation turned off. The bZ4X's tab showed "Electric range (A/C on) 99.744 km" beside "Electric range (A/C off) 103.9 km". The long number is a reading converted out of miles: headlineMetrics multiplied by 1.609344 and printed whatever came out, so a range estimate claimed to know the distance to the metre, and the two readings disagreed about their own precision on the same card. Distances now keep one decimal and percentages none, applied by the reading's kind rather than by whether it was converted - a provider reporting 99.744 km natively gets the same treatment. Anything else is left alone, because without knowing what it measures there is no safe place to cut. The odometer already rounded to a whole number on its own path; this only changes the headline readings. The Add-user sheet's "Create user" button sat underneath the system navigation bar. Every one of these sheets padded its bottom with viewInsets.bottom, which is the keyboard - correct while typing and wrong the rest of the time, because with the keyboard down that inset is zero and the navigation bar is still there. They take the larger of the keyboard and the navigation bar now, since a raised keyboard covers the bar and the two must not be added. One helper on DriverVault rather than the same expression in six files, which is how the six drifted into being identical and identically wrong. Verified: go build, go vet and go test ./... pass, with a new test covering the conversion (62 mi reads 99.8 km), a native over-precise reading, a percentage, and the odometer's whole number surviving. flutter analyze clean, 21 tests pass, and the rebuilt release APK was installed on the phone - the Create user button now sits clear of the navigation bar, where the screenshot that prompted this showed it clipped. Not verified: the rounding is not visible on the phone yet. It talks to a deployed API Server that has not been rebuilt from this commit, so that tab will keep reading 99.744 until the server is redeployed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
215c027ada |
Panel: give the API Server a name, and a tab to set it in
The Web App can be pointed at more than one DriverVault, but a server it
adds is only ever identified by the URL that was typed into the connect
dialog. Nothing on the other end says what it is called, so the switcher
has no name to show that the operator did not invent locally.
So the server now carries one. SERVER_NAME joins the config, defaulting to
"DriverVault API Server" so /api/health always has something a client can
display rather than an empty string every caller has to special-case.
GET/PUT /api/admin/server-config follow the pb-config and webapp-config
shape exactly: superadmin only, applied at runtime and then persisted to
.env, with the same "applied but could not be saved" warning when the write
fails. There is no /test sibling, because a name is a label and not an
address - there is nothing to probe. The length cap counts runes rather
than bytes, so a 64-character Polish or Danish name is not cut off at the
halfway mark.
/api/health reports it, unauthenticated, which is the point of the whole
change: a client adding this server by URL can label it from the probe it
already makes, instead of needing a second and authenticated call before it
can draw the entry.
In the panel it is a new API Server tab, first in the superadmin group
since it is this server itself, ahead of the PocketBase and Web App tabs
that describe what it talks to. Strings in all three languages, and the
route table in the README and the API reference tab both grow the two new
endpoints.
Known gap, deliberately not closed here: the compose files do not pass
SERVER_NAME, so under Docker a rename from the panel writes the container's
.env and no volume keeps it - it reverts to the default the next time the
container is recreated. Wiring it as ${SERVER_NAME:-} would make the host
.env authoritative, at the cost of the other trap the previous commit
documented, where the environment silently overrides the panel on every
restart. That is a call about the deployment, not about this endpoint.
Verified by new tests over the handler: the rename applies at runtime,
lands in .env, reaches /api/health, is rejected without touching .env when
blank or over-long, and accepts a name of exactly the limit in multi-byte
runes. go build, go vet and go test ./... pass. Drove the built panel in a
browser against a stub backend - the tab renders, loads the current name,
saves, and reads correctly in Polish - and ran the rebuilt api-server.exe
and webapp.exe end to end, confirming the embedded bundle really contains
the new tab and that SERVER_NAME reaches /api/health through both the
server itself and the Web App's proxy.
Not verified: no Docker build, so the images still serve the old panel
until they are rebuilt and pushed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
tajniak81andClaude Opus 5
|
ee4ac441be |
Plugins: drop the plugins.json migration, and the volume it needed
The project has no public installs, so there is nothing to migrate from. MigrateLegacyFile, the file-backed Store it read through, PLUGINS_FILE and the legacy path threaded through the Server all go. What is left is one store, PocketBase, and a plugins package that touches no filesystem at all. That was the last thing keeping api_data alive, so the volume goes too. All four compose files now declare exactly one volume, pb_data, and the standalone API Server compose declares none - it talks to an external PocketBase and has nothing of its own to keep. Backing up the stack is backing up one path again. Both images get simpler for it. The API Server image loses VOLUME /data and the su-exec entrypoint that existed only to fix a mounted volume's ownership, so it goes back to a plain USER app; its working directory is now /app and holds nothing. The AIO image loses its second volume and chowns only /pb/pb_data. One consequence worth stating plainly, because it is a small regression rather than a no-op. The panel's Settings -> PocketBase and Settings -> Web App screens write .env in the working directory, which is now ephemeral. In the multi-container stack that changes nothing: compose sets all five of those keys as container environment, and loadDotEnv only applies a key that is not already set, so the file could never win a restart there anyway. In the AIO image it did win for POCKETBASE_ADMIN_EMAIL/_PASSWORD, which are not in that container's environment - so a service account fixed from the panel now lasts only until the container is recreated. Both READMEs say so. Moving those two screens into the app_settings singleton would close it properly; the PocketBase URL and credentials cannot follow, since they are how the database is reached in the first place. go build, go vet and go test ./... pass; the compose files parse and each resolves to a single pb_data volume. Not verified: no Docker CLI here, so neither image was built. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
660af5736a |
Plugins: create the settings collection instead of waiting for it forever
|
||
|
tajniak81andClaude Opus 5
|
01a8fecf40 |
Docker: stop telling operators to turn off the bootstrap that upgrades them
Every deployment file advised setting PB_BOOTSTRAP=false "once the database
is established". That was harmless while the schema was static. It stopped
being harmless in
|
||
|
tajniak81andClaude Opus 5
|
9bd5c523c4 |
Plugins: the global layer moves into the database, beside the other two
The integration cascade stored its top layer differently from the two below it: org (L2) and user (L3) plugin config lived in PocketBase, in a pluginSettings field, while the global (L1) layer sat in a plugins.json next to the binary. That split was accretion rather than design - the file was the whole store in the v1 MVP, and the per-tenant layers were later built on PocketBase and layered on top of it instead of replacing it. It also cost something real. plugins.json was a second state store with different durability from pb_data: its own volume, its own ownership, its own backup. Losing pb_data is unmissable; losing api_data was silent, which is how "every plugin comes back disabled after a redeploy" happened. L1 now lives in the app_settings collection - one record keyed "global", holding its settings in a pluginSettings field, the same mechanism and the same field name the layers below use. The documents still differ in shape, because only L1 carries enable state and the registration of external plugins, but the storage is no longer a special case. The Manager grows a Store seam (PocketBase in production, file for the import, memory for tests) and, more importantly, a loaded gate. Settings in a database mean the store can be unreachable at boot - a cold stack, or a service account still to be set from the panel. That must not read as "no plugins configured", or the first save would write emptiness over real settings. So until a read succeeds the Manager stays unloaded, every mutation is refused, /api/admin/plugins* answers 503, and a background retry backs off to two minutes. The same gate covers a document that will not parse: it is never replaced by one built from an empty map, which is a stronger guarantee than the .corrupt backup it replaces. Writing to a store also revealed a hole in the previous fix. Classifying a save failure as errPersist was left to each Store, and a store that returned a plain error would fall through to the "saved, but the plugin failed to start" branch and be reported as a 200 - the same silent-success bug through a different door. The Manager now classifies, whatever the Store returns; a test pins it. Upgrades are automatic: on the first boot that finds no settings in the database, an existing plugins.json is imported and renamed to plugins.json.migrated. The import is refused if the store is merely unreachable, or if the file does not parse, so a stale or broken file can never overwrite live settings. /data is still needed - the panel rewrites .env there when it retargets PocketBase - but plugin settings no longer depend on it. 21 tests in internal/plugins cover both stores, including the production path against a fake PocketBase: create-then-update of the singleton, round-trip across a restart, an outage that leaves settings intact, a missing collection reading as not-ready rather than empty, and the import running exactly once. go build, go vet and go test ./... pass. Schema changes are mirrored into scripts/setup-pocketbase.mjs as that file requires. Not verified: no Docker CLI here, so no image was built and the bootstrap of app_settings against a real PocketBase is untested outside the fake. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
c173ca3653 |
Plugins: a save that fails should say so, not vanish on redeploy
Reported symptom: every plugin comes back disabled after redeploying the image, having been enabled before it. The persistence design was already right - each compose file mounts api_data:/data and points PLUGINS_FILE at /data/plugins.json - so the fault was that a failed write to that file was invisible. Three defects, each confirmed with a test before being fixed: A failed write was reported as success. Upsert set rec.Enabled before it persisted, and the handler folded the resulting error into the same 200-with-warning used for "saved, but the connector failed to start". The panel reloaded, read the in-memory record and showed the plugin enabled; only a restart revealed that nothing had reached the disk. A save that fails now rolls back in memory and returns 500, so the panel row shows the error instead of "Saved". A corrupt state file silently wiped the rest. Load returned an error, main.go logged it and carried on with an empty record set, so the next toggle overwrote plugins.json and took every other plugin's config with it. An unreadable file is now moved aside to plugins.json.corrupt, and persistLocked writes through a temp file + rename so an interrupted write cannot produce that corrupt file in the first place. A state file holding "null" panicked the server with "assignment to entry in nil map" on the next save, and a null entry nil-dereferenced in Load. Both now decode to "nothing configured". Two changes make the next such failure loud rather than silent. StartPlugins probes writability at boot and warns that plugin changes will not survive a restart. And the API Server image gains the root entrypoint the AIO image already had - chown /data, then drop to app via su-exec - because a host bind mount (API_DATA=/srv/...) or a volume created before /data existed arrives root-owned, and the unprivileged process cannot write to it. Not addressed here: a deployment that never reuses the named volume (docker compose down -v, a renamed compose project, an anonymous volume from a bare docker run) loses the file whatever the code does. The new boot warning tells the two apart - writable but empty means the volume is the problem, not permissions. go build, go vet and go test ./... all pass. The Dockerfile change is reviewed but not built: there is no Docker CLI on this machine, so the su-exec privilege drop follows standard Alpine practice rather than an observed run. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
9abb03ee4f |
Service history: 0 km is a reading, not a blank
A car collected new sits at 0 km, and every km calculation in the app
quietly refused to work for it. ComputeDerived only filled NextServiceKm
when Km > 0, so a service record entered at 0 produced no next-due
distance at all — the date side worked, because it guards on IsZero(),
which is genuine absence rather than a number that happens to be low.
The same conflation had been copied outward from there. The reminder's
km signal wanted currentKm > 0 before it would count anything down, the
web badge and the service-life ring tested the odometer for truthiness,
formatKm printed an em dash for zero, and fuel and charging rejected a
0 km entry as "odometer (km) is required" — which is the first charge
of an EV on the driveway on delivery day. The phone app carried its own
copy of each. Editing such a car offered an empty odometer box, since
the forms only prefilled a reading above zero.
Everywhere the odometer is a measurement, absence is now tested as
absence: null in the clients, negative on the server, and the required
fields check that the box was filled rather than that the number cleared
zero. Fuel and charging validate Km < 0 instead, and their inputs drop
min="1". Completing a repeating km reminder rolls from the car's actual
reading in every case; the old fallback to the previous target existed
to keep an untracked car off a due date in the past, but CurrentKm +
RepeatKm is ahead of the car by construction, so it could not have
happened.
Left as it was: dueKm, repeatKm and the service intervals, where zero
really does encode "no trigger" and "use the default", and the liters
and kwh checks, since a zero fill is not a fill.
Maintenance is the exception. Its odometer is the one that is genuinely
optional, so zero there still has to mean "not recorded" and those three
sites keep the truthiness test, commented. Fixing that properly wants a
nullable field rather than an int, which is a schema change and its own
commit — the same shape of problem as the latency em dash in
|
||
|
tajniak81andClaude Opus 5
|
3c4eba87c8 |
Status: latency the panel can actually show
PocketBase and the Web App share a Docker network and answer a health probe in well under a millisecond. Milliseconds() truncated that to 0, and omitempty on an int64 dropped the zero from the JSON, so the panel saw no latencyMs at all and rendered an em dash. Latency is now a *float64 rounded to one decimal, computed from Microseconds(). The pointer keeps an absent measurement — the API Server row, which probes nothing — distinct from a genuinely fast one. pbProbe follows suit, since it copies straight off svcHealth. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
f08849e50c |
Docker: a build context that isn't 3.6GB, and health you can see
The all-in-one image builds from the project root, and Docker only reads .dockerignore from the context root — so the ones under "API Server" and "Web App" never applied to it and every AIO build shipped the whole tree, "Phone App/build" included. A root .dockerignore allow-lists the paths that build actually copies. The dev split stack passed neither PB_BOOTSTRAP nor the SUPERADMIN vars, so it created the schema and then no user to log in with. It passes them now, and .env.example says so. WEBAPP_URL was never set anywhere, leaving the panel status page probing localhost:8090 — itself — and always reporting the Web App as down. Each compose file now points it at wherever the Web App really is, and the BFF grew a real /healthz instead of letting the SPA fallback answer probes with index.html and look healthy no matter what. In the AIO, PocketBase and the API Server drop to an unprivileged user; only nginx stays root to bind :80. The entrypoint takes ownership of the two volumes first, so data written by the old root-only image stays writable. All three images carry a HEALTHCHECK, every compose file declares one too (so depends_on still gates against an older pulled image), and web-app waits for the API Server to be serving rather than merely started. Also: pinned alpine/golang/node and PocketBase 0.39.11, so a rebuild months from now produces the same image; nginx forwards WebSocket upgrades instead of stripping them, with the map in http.d where Alpine actually reads it; and a .gitattributes keeps entrypoint.sh on LF, because a CRLF shebang from a Windows clone fails at container start with "no such file or directory". Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |
||
|
tajniak81andClaude Opus 5
|
cc1dafa9f7 |
Cars: drag the tabs into order, and a lock for every arrangement
Two things, both about layouts you arrange by dragging.
A car's tab bar now takes a drag: the tabs reorder as the pointer crosses
them and the arrangement saves on drop — or on dragend, since a tab
released in the gap beside the bar never produces a drop and would
otherwise revert on the next load. Same native drag events as the garage
and the Information rows, so also pointer-only, and it needs write access.
The order belongs to the car, like the choice of which tabs show at all,
so everyone it is shared with sees the same bar. It is stored as the full
list of keys, hidden tabs included, so a tab switched off and back on
returns to where it was rather than to the end; a key the stored
arrangement doesn't mention — a tab added in a later release — follows
the arranged ones. Information is arrangeable although it cannot be
switched off, which is why the validation needs arrangeableCarTabs rather
than reusing hideableCarTabs; it is derived from that set so the two
cannot drift as tabs are added. tabOrder rides on the existing PUT
/api/cars/{id}/view, so a tab drag never has to resend what is hidden.
Where the page opens is unchanged: Information, wherever it now sits.
And a padlock in the sidebar, above the theme toggle, holds every
arrangement in the app still at once — the garage, a car's tabs, its
Information rows, the provider's readings. It is a guard against nudging
a layout while reading it, not a permission: it is the user's own setting
and says nothing about what anybody may edit, so locking hides your own
drag handles rather than stopping a co-owner rearranging a shared car.
Stored as dragLocked on the profile, like the theme it sits above, so a
locked account is still locked on the next device — where a folded
provider card stays one browser's reading habit. Unlocked by default, so
nothing changes until it is clicked, and while locked the grab cursor and
the drag hints go with the drag.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
||
|
tajniak81andClaude Opus 5
|
6191160f14 |
Cars: log what charging costs, and drag the provider readings
Three things, all on a car's page. A Charging cost tab, which is the Fuel cost tab written for an electric car: charges in kWh, consumption in kWh/100km beside km per kWh, cost per km and price per kWh, and the same summary panel over the whole history. It keeps the reference-point method too, and has to — a session records the energy that went in, not what was left in the battery, so a given number of kWh only maps to a distance between two charges that ended at the same state. A charge to the car's usual full point plays the part of the full tank; partial charges still count towards the cost and roll into the next full one; and a charge taken without logging it leaves its window uncomputed rather than reporting an implausibly good figure. Sessions live in their own collection, the figures are derived on read like the fuel ones, and logging a charge advances the odometer exactly as a refill does. The tab switches off from the gear like every other, so a petrol car need never see it. The headline readings on the connected-service tab now drag into any order, saved on drop. Stored on the car as metricOrder, like the Information rows, rather than per device the way the collapsed cards are: an arrangement is something everyone the car is shared with should see, where a folded card is one browser's reading habit. Only what the provider reported can be arranged, so a reading that turns up later joins the end rather than displacing the arrangement. Fuel is now Fuel cost, tab and heading, which is what the tab has always been about and what pairs it with Charging cost. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> |