Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
44 commits
Select commit Hold shift + click to select a range
01ab443
fix(live): Home use as the plain balance; display and energy load as …
frahlg-agents[bot] Oct 21, 2026
b2018e8
chore(release): version packages (#1769)
frahlg Oct 21, 2026
7ad6948
fix(backup): remove scratch that a stopped backup left behind (#1671)
frahlg Oct 21, 2026
df21447
fix(backup): a failed full backup says why (#1670)
frahlg Oct 21, 2026
aa8bc17
Report bounded phase and step timings for history failures (#1617)
frahlg Oct 21, 2026
b271a76
fix(drivers): restart a driver that stays offline (#1628)
frahlg Oct 21, 2026
5e8e417
fix(backup): stop backup creation and verification after client disco…
frahlg Oct 21, 2026
5730b44
Revert "fix(drivers): restart a driver that stays offline (#1628)" (#…
frahlg Oct 21, 2026
18f75ac
fix(live): one Home use; the page follows load_quality alone (#1773)
frahlg Oct 21, 2026
dcdbd27
chore(release): version packages (#1770)
frahlg Oct 21, 2026
200da78
test(control): closed-loop site simulation with a real fuse (#1775)
frahlg Oct 21, 2026
fc621be
fix(loadpoint): the site-meter limit is the only fuse guard for charg…
frahlg Oct 21, 2026
6aa082f
fix(control): count EV raises per phase until the charger can apply t…
frahlg Oct 21, 2026
a2330c0
fix(live): Home use is an estimate once a device read fails (#1780)
frahlg Oct 21, 2026
41675e4
docs: link public project statistics (#1781)
frahlg Oct 21, 2026
98e5c79
fix(ocpp): an Available charger is a measured 0 W (#1779)
frahlg Oct 21, 2026
8d5c6bf
fix(loadpoint): Charge now keeps charging after a restart (#1778)
frahlg Oct 21, 2026
767b144
feat(ha): open FTW through HA ingress without a second password (#1786)
frahlg Oct 21, 2026
b5702e7
fix(control): the hourly import ceiling reports import_limit, not fus…
frahlg Oct 21, 2026
db5aec8
fix(ocpp): a charger socket may stay silent for five heartbeats (#1790)
frahlg Oct 21, 2026
c14d31d
fix(loadpoint): a charger is never asked for more current per phase t…
frahlg Oct 21, 2026
b8ab2c2
chore(release): version packages (#1777)
frahlg Oct 21, 2026
bbeac31
fix(control): a cut ends the reservation of the higher orders before …
frahlg Oct 22, 2026
7202ace
test(control): field-configuration matrix for EV charging and fuse pr…
frahlg Oct 22, 2026
04dbdd2
fix(loadpoint): planned charging runs on three phases, and room never…
frahlg Oct 22, 2026
5ae590a
fix(feedback): a charger that draws against FTW's order is not follow…
frahlg Oct 22, 2026
62ed7ea
fix(control): a cut on its way covers the overage it was sent for (#1…
frahlg Oct 22, 2026
2e84710
fix(loadpoint): a one-phase car on a three-phase charger is seen, pla…
frahlg Oct 22, 2026
8ac2750
fix(loadpoint): Charge now runs on one phase on a charger that cannot…
frahlg Oct 22, 2026
5b697a5
fix(control): a battery order on its way counts on each phase (#1805)
frahlg Oct 22, 2026
b0cb7d4
fix(loadpoint): plan a one-phase car above the charger's minimum, not…
frahlg Oct 22, 2026
16c4ad0
chore(release): version packages (#1794)
frahlg Oct 22, 2026
0cb9bf2
feat(loadpoint): tell the driver Core's current cap per phase (#1809)
frahlg Oct 22, 2026
dae4f41
fix(drivers): bundle easee_cloud 1.3.8, the 7 A start for a 6 A order…
frahlg Oct 23, 2026
7cccf99
test(sitesim): judge Core with device-simulator's fuse model (#1812)
frahlg Oct 23, 2026
1883582
test(site): FTW's scenario pack on device-simulator's house, run on a…
frahlg Oct 23, 2026
b3af718
fix(control): interchangeable batteries share the plan's total (#1814)
frahlg Oct 23, 2026
7b99353
fix(control): a down battery's PV no longer takes part of the charge …
frahlg Oct 23, 2026
bac597f
feat(cloud): send FTW Cloud 48 hours of history tiles (C6) (#1817)
frahlg Oct 23, 2026
21e3d16
feat(cloudwire): x carries a watch's exchange through FTW Cloud (#1818)
frahlg Oct 23, 2026
d68dedc
feat(cloud): a watch's exchange rides FTW Cloud (#1819)
frahlg Oct 23, 2026
5cf4a80
feat(stats): one statistics answer, and anonymous daily reports throu…
frahlg Oct 24, 2026
2d7c8eb
feat(fleetping): say coarsely how the box ran in the anonymous daily …
frahlg Oct 24, 2026
e225339
Merge current master into load-balancer feedback preserving current E…
segran2 Oct 24, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions .changeset/anonymous-daily-energy.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,7 @@
---
"ftw": minor
---

Statistics are one answer on the box, for two anonymous daily reports. With FTW Cloud they are part of Cloud. Without Cloud they are on until you turn them off under Settings → FTW app → Anonymous statistics, which shows both messages; nobody is asked any more. A No, a fleet report switched off, or Cloud turned off on an earlier release stays off.

Beside the fleet report, the box can now send one energy report a day: the date and that UTC day's solar generation, grid import and export, battery charge and discharge and vehicle charge, in whole kWh, never in the same message as the fleet report and at its own time. It skips any day with a gap in the energy record and any site a partner controls. Once a release names an Oblivious HTTP relay, both reports go through it, so FTW Cloud never sees the box's address. This release names none yet: the energy report is not sent, a box that said yes sends the fleet report straight to FTW Cloud as before, and a box that shares only by default sends nothing until a relay is named.
5 changes: 5 additions & 0 deletions .changeset/bundle-easee-cloud-138.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
"ftw": patch
---

An Easee Cloud charger now starts a 6 A charge at once. Easee's firmware waits about five minutes before it starts a car at 6 A. The bundled `easee_cloud` driver (1.3.8) offers 7 A per phase for that start, and only when the per-phase cap that FTW sends with each order (`cap_amps_per_phase`) allows 7 A. It steps down to 6 A as soon as the charger reports charging, or after 90 s. A charger on a phase without room for 7 A starts at 6 A as before. `start_boost: false` in the driver config turns the 7 A start off.
5 changes: 5 additions & 0 deletions .changeset/cloud-history-tiles.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
"ftw": patch
---

With FTW Cloud on and history in the owner's grant, the box sends FTW Cloud its last 48 hours of 5-minute history, so the iPhone app can draw a node's history while the box is offline. Each closed 12-hour tile is the same bytes the app gets from the box directly. The box adds each 5-minute bucket 30 seconds after it ends, and sends the tiles FTW Cloud lacks when it reconnects. It sends no history without NTP time, and none past the daily byte budget. The extra traffic is about 100 KB a day.
5 changes: 5 additions & 0 deletions .changeset/cloudwire-x.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
"ftw": patch
---

FTW Cloud's protocol gains `x`, which carries a watch's exchange with its box through a one-shot request, so a watch can charge the car without its iPhone near.
5 changes: 5 additions & 0 deletions .changeset/down-group-pv-share.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
"ftw": patch
---

Batteries now take the whole charge FTW asks for when another battery is down. On a site with inverter groups, the PV of an inverter whose battery is offline, faulted or cannot charge was still counted as charge for that battery, and that part of the target went to no battery. With 3.3 kW of PV on the down battery's inverter and a 9.6 kW charge target, the other two batteries got 6.3 kW in total; when the target was smaller than that PV, they got nothing and the surplus went to the grid. That PV now counts as power on the AC bus, so the other batteries share the whole target by capacity, within their limits.
5 changes: 5 additions & 0 deletions .changeset/ev-command-cap-amps.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
"ftw": patch
---

Each charging order now tells the charger's driver how much current per phase FTW allows that charger right now, in a new `cap_amps_per_phase` field. It is the charger's share of the room below the fuse and the import limit, after the split between chargers, within the order's `max_amps_per_phase` and the charger's maximum. Until now a driver saw only the fuse rating. The Easee Cloud driver needs this to start a 6 A charge at 7 A only when 7 A fits, so a car on a tight phase is not stopped and restarted. The field adds no new limit, and FTW sends the same orders as before.
5 changes: 5 additions & 0 deletions .changeset/fleet-report-health.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
"ftw": minor
---

The anonymous daily fleet report also says, coarsely, how the box ran: how slow its slowest plan of the last day was, how many plans fell back, how often its drivers failed to read, and whether its newest update of the last week failed. Each is a bucket of a few words, under the same switch as the rest of the report. Settings shows them in the message preview.
5 changes: 5 additions & 0 deletions .changeset/interchangeable-batteries-share.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
"ftw": patch
---

When all batteries have the same power per kWh, they share the plan's total again. In a planner mode FTW followed each battery's own share of the plan. The planner pays the same for a kWh on either battery, so it often put a whole slot on one and the next on the other: one battery worked while the other stood at 0 W, their charge levels drifted apart, and on a large load only one battery's planned rate was used and the rest came from the grid. FTW now splits the live battery target across such batteries by the energy each can still give or take. The fuller battery discharges faster and the emptier one charges faster, so their charge levels draw closer while they work. Both cover a large load within their power limits: a battery at its limit stays there and the others take the rest. A battery that reports it cannot charge or discharge hands its share to the others. In an idle slot, or when FTW does not know a battery's charge level, the split is by capacity as before. Batteries whose charge or discharge limit per kWh differs by more than 5 %, such as one on a smaller hybrid inverter, still follow their own planned energy, and so does every battery in a fleet where any one differs. There is no new setting.
5 changes: 5 additions & 0 deletions .changeset/watch-exchange-through-cloud.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
"ftw": minor
---

An Apple Watch can charge the car without its iPhone near. Its commands reach the box through FTW Cloud as short encrypted exchanges the box answers itself, and an owner's phone adds the watch by its key, with a passkey step-up. The box lists `cmd.exchange` while FTW Cloud carries them.
51 changes: 51 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,56 @@
# Changelog

## 0.155.1

### Patch Changes

- 8ac2750: Charge now works on a one-phase charger again when the app asks for three phases. On a charger whose levels stop below 6 A on three phases, such as a 32 A one-phase charger with the levels 1.4, 2.3 and 3.7 kW, the app sent Charge now as three-phase, because FTW reported three phases for an auto charger. A three-phase order there has no level, so the car got 0 W. FTW now runs it on one phase unless the charger is set to 3p, and reports one phase for such a charger, so the app sends what it can run.
- 5ae590a: A charger that keeps charging after FTW tells it to stop or slow down now shows as Not following, not Limited. In 0.155.0 an Easee Cloud charger that went on drawing 3.1 kW after FTW cut it to 0 W for the main fuse showed "Limited by the main fuse" for as long as it drew. A limit explains why FTW ordered less; it does not explain a charger that takes more. FTW now warns once the charger has drawn more than its order for four minutes, and raises an alarm ten minutes later. A car that takes less than a limited order is judged as it is under a plan order. Sending the same order on every tick no longer pushes the earlier order out of the two-minute response window, so a slow cloud charger is not judged on an order it could not yet have applied.
- 62ed7ea: Charging is no longer cut twice when a cloud charger is slow to apply a cut. Easee Cloud takes 30–120 s to apply a new current, but FTW cut again after 20 s, before the first cut could show on the site meter, so a kettle on top could take the car from 10,400 W to 0 W. FTW now lets a cut it has sent cover the overage until the charger can have applied it: the time the driver declares, or 3 minutes. Only what a new load adds is cut. A raise still on its way now counts in a cut, so the phase is not left over the limit when the raise lands. Once a phase's averaged current passes 1.25 times the fuse rating, which a gG fuse carries for an hour, FTW cuts at once whatever is on its way.
- bbeac31: Charging starts again promptly after a pause. In 0.155.0, an Easee Cloud charger that was paused and then asked to start waited up to three minutes, shown as limited by the main fuse, while the phases had plenty of room: FTW still counted the higher current it had ordered before the pause as on its way. A pause now ends that reservation, because the charger follows its latest order and the site meter shows its draw until the pause lands. In the reported case the restart now comes after 12 s instead of 2 min 47 s.
- 5b697a5: A one-phase charger now starts while FTW raises the battery. Since 0.147.2 a car on a one-phase charger, or on a charger set to 1p, could sit at 0 W shown as limited by the main fuse with every phase far below its limit: FTW took the battery's whole new order, which the site meter could not show yet, off the room on the car's one phase, so a three-phase battery counted three times there. In the reported case Charge now waited 42 s; on a busy 16 A site with the battery stepping up every half minute the car did not start for 9 minutes. The battery's order now counts as a three-phase battery puts it, the same current on each phase. A three-phase charger's limit does not change.
- b0cb7d4: A car that charges on one phase of a three-phase charger is planned again once FTW sees it. FTW plans such a car at a third of each three-phase order, but it gave the planner the charger's three-phase minimum, 4.1 kW, with those one-phase levels. The planner dropped every level below that minimum and planned no charging, so the car stopped about five minutes into the session and missed its goal. The planner now gets the charger's three-phase levels divided by three, for example 1.4–3.7 kW. Each level is a third of an order the charger accepts, so the car draws what the plan counts.
- 2e84710: A one-phase car, as many plug-in hybrids are, now gets what it can take from a three-phase charger. Such a car draws the order's current per phase on one phase only, so a 4.8 kW three-phase order gave it 1.6 kW while the plan counted 4.8 kW, and the charger panel said it was not following. FTW now sees it on the site meter: when its order rises, one phase rises by about the order's current and the other two stay put, while the charger's own reading shows the car drawing that much. For the rest of that charging session the plan counts the car's one-phase levels, up to the charger's current per phase (3.7 kW at 16 A), the charger stays on three phases, and the order names the current the car should take. The charger panel says "Charging on one phase, at most 3.7 kW" instead of "Not following". A house load, a three-phase car or a two-phase car is never taken for a one-phase car, and the next car starts unknown. The fuse limit is unchanged: it counts a three-phase order's current on every phase.
- 04dbdd2: Planned charging and Charge now run on three phases. In 0.155.0 a loadpoint set to auto could charge a three-phase car on one phase all night: with an Easee step table that also lists one-phase levels, the planner spread the charge at 3.1–3.5 kW, which only one phase can deliver, and FTW switched the charger to one phase to follow it. On a 3×20 A site the car got about 3.3 kW instead of about 10 kW. One phase never helps the fuse: it puts all the current on one phase, caps the car near 3.7 kW, and an Easee has to pause to switch back. Auto, and an unset phase mode, now mean three phases on a charger that can run 6 A on each, and a planned order is raised to that minimum (4.14 kW at 230 V). Only charging from spare solar below it uses one phase, with the same hold against flapping as before, also when that solar-only loadpoint has a goal. The planner gets only steps the charger can run on three phases. A charger whose step table, or maximum, stays below 6 A on three phases is a one-phase charger and keeps its one-phase steps and orders; so does a loadpoint set to 1p. A one-phase charger above 16 A with a longer table still needs phase mode 1p.
- 04dbdd2: FTW no longer cuts a charging car when every phase has room. In 0.155.0 a car charging on one phase under auto was cut from 3.5 kW to 1.6 kW and then to 0 W, again and again, while the busiest phase read about 16 A on a 20 A fuse: dividing the fuse room between cars read the order as three-phase, gave the car's own current no credit, and cut it to what was left on that phase. Keeping a car's current order on the same phases is now never a cut there; only a measured overage on the site meter cuts. An order that moves the same watts onto other phases is still held to the room on those phases.

## 0.155.0

### Minor Changes

- 767b144: Add a Home Assistant ingress mode. The HA app can use the owner's HA login for setup and settings, including installs with a saved FTW password. The ingress listener accepts Supervisor requests and local health checks; other clients cannot open the UI. Ordinary native installs keep their existing access rules.

### Patch Changes

- 8d5c6bf: Charge now keeps charging after an update or restart. Before, a restart turned it into a pause until someone pressed Start again. It still ends on Stop, when the car is unplugged or full, and a different car on the charger does not inherit it.
- c14d31d: A charger is never asked for more current per phase than it supports. When FTW moves a charger that can run three phases onto one phase, the order now stops at that charger's own current per phase instead of the site fuse. Before, a 16 A charger on a 25 A fuse could be offered 18–25 A on one phase while charging from surplus or on Charge now.
- 6aa082f: EV charging no longer pushes a phase past the fuse when a charger is slow to apply a new current, when a charge moves between three phases and one, or when one car is raised while another is cut. Each raise now counts as current on the phases its command names until the charger can have applied it: the time its driver declares, or 3 minutes when it declares none. A car takes only room that no other charger's raise holds. After a rise, a charger that declares no response time may wait up to 3 minutes before it climbs further.
- fc621be: Charging no longer stalls or switches off when the phase mode is empty or auto, when a battery-level PV unlock is saved but not active, or on a one-phase site. The site-meter limit is now the only fuse guard for chargers: it measures room on the phases each charger is told to use, and a three-phase offer is never below 6 A per phase.
- a2330c0: Home use is marked "≈" as soon as a device stops answering. Before, Core kept the device's last power for up to 90 s and still called Home use measured, so the page could show a wrong value and "% self-powered now" while every watt came from the grid. The last value is still shown until the age limit, but as an estimate that names the device. A device the watchdog has taken offline now stays offline until it delivers data again. Each failed read used to bring it back for a moment, so Home use flipped every few seconds and the watchdog reported the device lost again each time.
- b5702e7: When the hourly import limit holds a car back, FTW now says so instead of blaming the main fuse. The charger's reason is `import_limit` ("Limited by the hourly import limit") whenever this hour's import budget leaves less room than the fuse, and a battery boost stopped by that budget says "Hourly import limit stopped boost". `fuse_limit` now means the main fuse alone, and the web page says "Limited by the main fuse".
- 98e5c79: Home use is no longer marked estimated because an OCPP charger has no car. When the charger reports Available, no car is connected and no current flows, so its power counts as a measured 0 W in live Home use and in history. Each heartbeat or other message from the charger keeps that reading current, and it ends as soon as a car arrives. Unavailable and Faulted still leave the charger's power unknown. The Available a charger sends for itself (connector 0) at boot or reconnect no longer reads as an unplugged car. Core now asks OCPP chargers for a heartbeat every 30 s instead of every 60 s, so one late heartbeat no longer marks an idle charger offline: one more small request and reply a minute per charger. A heartbeat_interval_s set in the config is kept.
- db5aec8: An idle OCPP charger no longer loses its connection about once a minute. The OCPP library closed a charger socket that had sent nothing for 60 s, the same as a 60 s heartbeat, so a charger that only sends heartbeats could be dropped as each one arrived, reconnect, and briefly show Home use as "≈". A charger socket may now stay silent for five heartbeat intervals (150 s with the default 30 s heartbeat). FTW's own watchdog still decides when a quiet charger is offline.

## 0.154.4

### Patch Changes

- 5e8e417: Stop full backup creation and verification when the HTTP client disconnects. Ctrl-C releases the backup and update locks instead of leaving work running on the box.
- df21447: A failed full backup now says why. Core writes when a backup starts, ends and fails to the journal, with the phase, table and row count it reached. A backup stopped by the two-hour limit for a live copy names that limit instead of only "context deadline exceeded". GET /api/backups keeps the reason, table and row count until the next backup starts, and `ftw backup` shows the last failure before it begins. When the raw copy does not fit in /dev/shm, Core warns that it is copying on the data disk and reports how much space earlier backups' scratch files still take there.
- 7ad6948: Core now removes what a stopped full backup left behind before it accepts a new one. A backup cut short when Core stops, for example by an update, left its raw copy in /dev/shm until the next reboot. On a Raspberry Pi that held more than a gigabyte of RAM and sent the next backup's copy to the SD card, where it ran past its two-hour limit. At start Core removes such raw copies, the stage and verification folders and any unfinished archive in the backups folder, and logs each path and its size. It leaves them when another backup is running, and it removes nothing else.
- aa8bc17: Report the longest step and step count beside each total phase time when a history write or forecast read fails. Archive failures also report connection, preparation, locks, write and commit times before cleanup. This separates one slow operation from many short operations without logging sample values or changing storage behavior.
- 18f75ac: Home use on the energy flow no longer shows "≈" when every device is read, and "% self-powered now" is back. The page marked Home use as an estimate whenever /api/status carried a second, energy-averaged value, which it always did. It now follows Core's `load_quality` alone, and /api/status no longer sends `load_estimate`.

Home use has one value. 0.154.3 was never published as a beta; its change also ships here: Home use is the site meter minus every device that is read, shown when one device is offline (marked "≈"), and blank only when the site meter is missing or stale. The Home Assistant sensor "Load Power (energy average)" announced for 0.154.3 is removed before release; "Load Power" stays.

## 0.154.3

### Patch Changes

- 01ab443: Show Home use again when one device is offline. Live Home use is the energy balance as is: the site meter minus every device that is read, as before 0.141. A device that is offline, or a setup problem, no longer blanks it; the dashboard marks it "estimated" with "≈". A negative value is shown, since it points at production FTW cannot see. After sunset at the site's map pin an offline inverter counts as 0 W. Home use is blank only when the site meter is missing or stale. `/api/status` adds `load_quality` and `load_reason` as hints.

Home use now has two series. The live display value is logged as the `load_display_w` metric. The energy-based Home use from closed history intervals, with its quality flags, stays what load models train on. Home Assistant gets a second sensor, "Load Power (energy average)", next to "Load Power".

## 0.154.2

### Patch Changes
Expand Down
Loading
Loading