You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
An EV owner can lower the requested current while the car still draws 11 kW without knowing whether the charger received the change. Charging now has one view that separates the request, the charger response and measured power. Opening a section does not change charging mode.
Battery level, goal and battery size save on release, with pending, failure and readback feedback in both clients. A stopped car never becomes a made-up 100% battery reading.
A failed replacement plan blocks old automatic dispatch until a new plan succeeds. Normal recalculation keeps the current plan; explicit manual controls remain available within safety limits.
Goal and SoC writes acknowledge storage before planning. A bounded queue keeps the latest waiting request. Both clients distinguish saved settings, a pending plan and a failed plan; they hide old charge windows after a new request.
Pause persists as a zero-power choice even before hardware identity is available. One durable intent marker prevents an older clear from overriding a newer pause across interrupted writes. OCPP requests fresh identity after reconnect and distinguishes unknown network state from a physical unplug. A socket generation catches reconnects between polls without becoming physical-session proof. Explicit Pause survives; an old Start requires a new choice. Charge now and Resume plan are separate actions. Manual charging works without an MPC plan and obeys charger and site fuse limits.
Restored battery levels and positive charging requests require matching charger and session evidence. Old acknowledgements and automatic stop decisions cannot overwrite a newer choice.
Charger setup leads from a charger-only catalogue through connection and saved configuration. Ordinary charger fields autosave; advanced OCPP and car profiles have explicit controls.
Restart keeps the existing container image. Full backup supports internal managed-driver links and isolated restore. Optimizer update and rollback persist their version pin.
App cancellation reaches savings database reads. The optimizer release workflow probes package write access before builds and defaults to a dry run.
This includes the EV allocation work in #957 and restart work in #1073; coordinate those existing PRs before merge. Easee 1.3.0 is pinned from srcfl/device-drivers#108. This beta uses Core DP. Core and updater need builds from the same release commit; a new external optimizer release is not required. The paired Webapp branch includes srcfl/ftw-webapp#55.
Validation: full make verify, focused race tests and all 533 Core web tests pass. Driver verification passes 4037 tests with 1309 skips. Tests cover command acknowledgement, fuse limits, persistence failures, session changes, cancellation, cloud delays, autosave and planning failure. Held-planner tests prove that writes return before planning finishes and that outdated windows stay hidden. Paired Webapp verification passes 1017 tests with 1 skip on Node 24.20.0, with no Svelte errors or warnings. Build and test checks pass on the exact final commits in all three code PRs.
A mobile owner/passkey walkthrough passed on Core 6843ddd and Webapp 8ead63a3 against the real box. The goal slider saved 85% by 07:00 in 1.46 seconds while planning continued, then reached a current plan. Both clients retained the goal after reload; an early car tap during reconnect opened the panel. Remove showed Goal removed and cleared the target and deadline. Both clients saved battery size 60→61→60 kWh without Save while unplugged. The App retained each value after a full reload; Core retained the restored value after reload. No browser errors occurred. The temporary identity was revoked and the original four identities remain.
The local test installed Core, updater and a separate shadow optimizer from this head. The active planner is Core DP; that three-image test setup is not a release dependency. Real UI restart kept the same container and image. A full backup was created, separately verified and restored to an isolated directory; SQLite quick_check and the restored driver hash passed.
Limits: the car is unplugged, so the final walkthrough cannot prove physical current reduction or delivery to a closed phone app. Earlier current tests used an older candidate. The external optimizer runs in shadow mode while Core DP supplies the active plan. The separate optimizer publication probe returned HTTP 403, but beta.yml only publishes Core and updater. That failure does not block this Core DP beta and needs no package-permission change for this release. No release has been published.
Review follow-up: the capacity/restart finding was a false positive. Manager.Load already reanchors and calls persistSession when capacity changes. TestConfirmedBatteryLevelSurvivesDatabaseCloseAndReopen now proves 60→100 kWh, a real SQLite close/reopen, and 84%→84.6% after another 600 Wh. See go/internal/loadpoint/session_sqlite_test.go and b5b0eae. All 12 review findings have code and regression evidence: 11 fixed and one proven false positive. Immutable SDK socket aliases bind old requests, replies and commands to their original connection. Real SDK tests prove that they cannot mutate a replacement connection in either OCPP version. Fresh OCPP recovery has real protocol tests for 1.6 and 2.0.1, adoption, failed triggers, late status and reconnects between controller ticks. These are protocol tests, not a new physical charger test.
…arge
Surplus-only is an EV policy, not a site import ban. The car may use
PV left after house load in the same slot the home battery buys from
the grid. Live surplus and the near-term 3Φ gate follow the same
accounting.
Add a joined EV site harness (plan → charger Tick → ComputeDispatch →
site identity). Isolated planner, charger and dispatch suites never
ran on one clock, which is why the combo stayed invisible.
Signed-off-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Fredrik Ahlgren <fredrik@sourceful-labs.com>
Give leftover PV, house residual and grid identity a home in
loadpoint.GridW. DP, ValidatePlan, main.go and the joined site clock
all read that contract.
Plan→EMS mapping lives in SlotDirectiveFromMPC and
SlotDirective.LoadpointDirective — the same adapters main.go uses.
The site clock publishes via Service.InstallPlan and reads
SlotDirectiveAt, so charger and battery cannot be given two mappings
of one slot.
Signed-off-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Fredrik Ahlgren <fredrik@sourceful-labs.com>
…r import
Meter import is not grid-funded battery charge. Soak plus EV can import
together while the battery is still taking leftover PV. Live surplus and
the 3Φ gate now treat grid minus EV as the battery's grid, the same
identity PlannedPVSoakW already used on a soak-only tick.
ValidatePlan rejects leftover breach. Python leftover-assert matches the
new policy. Ship a new optimizer image with this core.
Signed-off-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Fredrik Ahlgren <fredrik@sourceful-labs.com>
Field report 2026-09-05 22:00 (#1002): Charge now on an Easee over the
cloud driver, and the Manual tab said "Charging at 16 A until the car is
full" a tenth of a second later. That line was written at click time and
never refreshed; the Easee cloud takes 5–15 s to act, the status table
showed 0 W, and the operator removed the charger to charge by hand.
Everything needed to say what the charger did with the order was already
on the wire in /api/ev/status; nothing read it on this path, and the box
kept no record of when the hold was installed.
The box now keeps that record. ManualHold carries StartedAt, the manager
records when its current order was first given (commanded_since_ms), and
GET /api/loadpoints carries `manual`: state (sent, accepted, charging,
not_drawing, stalled, limited), started_at_ms, since_ms, requested and
commanded watts and amps, and the charger's reported limit and reason.
ManualStatusFrom is pure and tested tick by tick.
The Manual tab is redrawn on every poll from that account, the plan
strip says the same sentence while a manual charge runs, and a refused
Start reads as a failure with the server's reason instead of the success
sentence.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The reason will be displayed to describe this comment to others. Learn more.
Stale comment
Left a non-blocking comment; not approved, and no reviewers were assigned. Cursor Bugbot reported one unresolved medium-severity finding, and APPROVAL_POLICY.md still requires CODEOWNERS review plus a visual check of the web changes. The only matching CODEOWNER is the PR author, so nobody was requested.
Sent by Cursor Approval Agent: Pull Request Router and Approver
The reason will be displayed to describe this comment to others. Learn more.
Stale comment
Left a non-blocking comment; not approved, and no reviewers were assigned. Cursor Bugbot was present on this head but finished as skipped, so the required automated-review signal did not complete successfully. APPROVAL_POLICY.md also does not authorize auto-approval here; CODEOWNERS and the web visual check still need a human.
Sent by Cursor Approval Agent: Pull Request Router and Approver
The reason will be displayed to describe this comment to others. Learn more.
Stale comment
Left a non-blocking comment; not approved, and no reviewers were assigned. Cursor Bugbot was present on this head but finished as skipped, so the required automated-review signal did not complete successfully. APPROVAL_POLICY.md also does not authorize auto-approval here; CODEOWNERS and the web visual check still need a human.
Sent by Cursor Approval Agent: Pull Request Router and Approver
The reason will be displayed to describe this comment to others. Learn more.
Left a non-blocking comment; not approved, and no reviewers were assigned. Cursor Bugbot passed on this head with no current findings, but APPROVAL_POLICY.md still does not authorize auto-approval: CODEOWNERS review and a visual check of the web changes remain required. The only matching CODEOWNER is the PR author, so nobody was requested.
Sent by Cursor Approval Agent: Pull Request Router and Approver
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
An EV owner can lower the requested current while the car still draws 11 kW without knowing whether the charger received the change. Charging now has one view that separates the request, the charger response and measured power. Opening a section does not change charging mode.
This includes the EV allocation work in #957 and restart work in #1073; coordinate those existing PRs before merge. Easee 1.3.0 is pinned from srcfl/device-drivers#108. This beta uses Core DP. Core and updater need builds from the same release commit; a new external optimizer release is not required. The paired Webapp branch includes srcfl/ftw-webapp#55.
Validation: full make verify, focused race tests and all 533 Core web tests pass. Driver verification passes 4037 tests with 1309 skips. Tests cover command acknowledgement, fuse limits, persistence failures, session changes, cancellation, cloud delays, autosave and planning failure. Held-planner tests prove that writes return before planning finishes and that outdated windows stay hidden. Paired Webapp verification passes 1017 tests with 1 skip on Node 24.20.0, with no Svelte errors or warnings. Build and test checks pass on the exact final commits in all three code PRs.
A mobile owner/passkey walkthrough passed on Core 6843ddd and Webapp 8ead63a3 against the real box. The goal slider saved 85% by 07:00 in 1.46 seconds while planning continued, then reached a current plan. Both clients retained the goal after reload; an early car tap during reconnect opened the panel. Remove showed Goal removed and cleared the target and deadline. Both clients saved battery size 60→61→60 kWh without Save while unplugged. The App retained each value after a full reload; Core retained the restored value after reload. No browser errors occurred. The temporary identity was revoked and the original four identities remain.
The local test installed Core, updater and a separate shadow optimizer from this head. The active planner is Core DP; that three-image test setup is not a release dependency. Real UI restart kept the same container and image. A full backup was created, separately verified and restored to an isolated directory; SQLite quick_check and the restored driver hash passed.
Limits: the car is unplugged, so the final walkthrough cannot prove physical current reduction or delivery to a closed phone app. Earlier current tests used an older candidate. The external optimizer runs in shadow mode while Core DP supplies the active plan. The separate optimizer publication probe returned HTTP 403, but beta.yml only publishes Core and updater. That failure does not block this Core DP beta and needs no package-permission change for this release. No release has been published.
Review follow-up: the capacity/restart finding was a false positive. Manager.Load already reanchors and calls persistSession when capacity changes. TestConfirmedBatteryLevelSurvivesDatabaseCloseAndReopen now proves 60→100 kWh, a real SQLite close/reopen, and 84%→84.6% after another 600 Wh. See go/internal/loadpoint/session_sqlite_test.go and b5b0eae. All 12 review findings have code and regression evidence: 11 fixed and one proven false positive. Immutable SDK socket aliases bind old requests, replies and commands to their original connection. Real SDK tests prove that they cannot mutate a replacement connection in either OCPP version. Fresh OCPP recovery has real protocol tests for 1.6 and 2.0.1, adoption, failed triggers, late status and reconnects between controller ticks. These are protocol tests, not a new physical charger test.
Contract-pair: srcfl/ftw-webapp@8ead63a