fix(octopus): choose on purpose between conflicting catalogue entries for a car (#5438)

[bootc]

Octopus's charge-point and vehicle catalogue isn't unique by make and model, and the device lookup kept whichever match came last. So a 7.4 kW zappi was planned at 22 kW: the catalogue lists Myenergi / zappi (all models) at both 7.400 and 22.000 kW. And because fetch.py takes the larger of the Octopus figure and car_charging_rate, the user's own setting couldn't bring it back down. In the catalogue cached on my system, 142 of 311 charge-point make+model pairs and 169 of 1322 vehicle pairs map to more than one value.

Closes #5438.

The change

• catalogue_value() gathers every matching value, and when they disagree it picks one on purpose:

  • Charge point: the lowest power listed, which is the most every variant can do. Octopus's figure only ever raises input_number.predbat_car_charging_rate, so a user with a faster charger still gets their rate by setting it.
  • Vehicle: the largest battery size listed, so the car's remaining charge is over-estimated rather than under-estimated.

• The component warns once per make and model, naming the values and the one in use. For a charger the warning also names the setting that decides the rate; for a vehicle it says car_charging_battery_size can't change it. Example:

  Warn: OctopusAPI: Octopus's catalogue lists charge point Myenergi 'zappi (all models)' at more than one power (7.4, 22 kW) and Predbat can't tell which yours is, so it uses the lowest, 7.4 kW, unless input_number.predbat_car_charging_rate (_1, _2 for further cars) is higher. Set that to your charger's rate
  

• The figures are converted to float. Values that are missing, non-numeric, non-finite or not above zero are ignored. GraphQL returns them as strings ("22.000"), which the discovery coordinator's typed ratings container dropped: Coordinator: octopus ratings.charge_point_power_kw value does not fit the container's type - dropped. The planner was unaffected by that part, because fetch.py already converts them.

• The misleading # Take the max as Octopus over reports comment in fetch.py is corrected. The code under it is unchanged.

• The Octopus direct-connection section of docs/car-charging.md now explains the lookup, what happens on an ambiguous entry, and which setting to change.

Where this departs from the issue

The issue suggested skipping an ambiguous catalogue value and using the configured settings, and letting a configured value override the catalogue. I didn't do either:

• Skipping the value. For a charger, skipping is the same as taking the lowest, except when the configured rate is below every catalogue figure. For a vehicle it would fall back to car_charging_battery_size, which is 100 kWh when it isn't set. That's further off than any of the catalogue's own variants.
• Letting car_charging_rate cap the Octopus figure. I've left fetch.py's max() as it is. It was added on purpose in Fix octopus intelligent slot crash #2855, and the default (7.4) can't be told apart from a rate the user set. Changing it could lower the rate for owners of faster chargers who rely on Octopus's figure.
• Letting car_charging_battery_size win over Octopus's battery size. Four templates (givenergy_cloud, givenergy_ems, huawei, solaredge) ship car_charging_battery_size: [75] uncommented. Preferring the setting would swap the real size from Octopus for that placeholder.

Who sees a change

• Owners of a charger whose model is listed at several powers, who had been getting a higher figure only because of catalogue order. They are now planned at the lowest figure, or at their configured rate if that's higher. Someone with a genuinely 22 kW zappi who left input_number.predbat_car_charging_rate at 7.4 now gets 7.4 kW, and the warning tells them what to set.
• Owners of a vehicle whose model is listed at several battery sizes. They now get the largest size rather than the last one listed.
• Unambiguous entries behave as before, except that the HA attribute is now a number rather than a string. fetch.py reads it with float() either way.

Tests

• catalogue_value(): unique, agreeing and conflicting matches; strings converted; null, missing, non-numeric, NaN, infinite, negative and zero values ignored; no such make or model.
• async_get_intelligent_devices() with the catalogue mocked: the lowest power or largest size is chosen for conflicting entries and the unique ones come through, all as floats. There's one warning per ambiguous make and model over two polls, and the ratings survive the coordinator.
• End to end: the device goes through the component's own intelligent_dispatch sensor to the real fetch_sensor_data_cars(). It is planned at 7.4 kW, not 22. A configured 3.7 is raised to 7.4, and a configured 11 still wins.
• I checked each guard by reintroducing the bug, and each makes the new tests fail with a message naming it: last match wins, first match wins, highest charger power, raw strings, no finite/positive check, no warn-once, warn-once keyed on the message alone, and fetch.py not applying the Octopus figure. The fixtures list the conflicting values in different orders, so neither first-match nor last-match wins can pass by luck.
• test_octopus_catalogue_cache now expects the battery size as 57.5 rather than "57.5".

./run_all --quick and pre-commit run --all-files pass. One correction to the issue: the catalogue spells the make Myenergi, not myenergi.

🤖 Generated with Claude Code

[Copilot]

🟡 Changes recommended

The new warning hard-codes the default entity prefix, making its remediation incorrect for custom-prefix installations.

1 open finding

• Use configurable prefix in car charging rate warning · New

What changed in this PR

Fixes #5438 by deterministically resolving duplicate Octopus catalogue entries.

Changes:

• Selects minimum charger power and maximum vehicle battery size.
• Converts valid catalogue values to floats and logs ambiguities once.
• Adds documentation and comprehensive regression tests.

File	Description
apps/​predbat/​octopus.py	Implements catalogue resolution and warnings.
apps/​predbat/​fetch.py	Clarifies existing rate-selection behavior.
apps/​predbat/​tests/​test_octopus_intelligent_devices.py	Tests conflicts, warnings, discovery, and planning.
apps/​predbat/​tests/​test_octopus_catalogue_cache.py	Updates numeric-value expectation.
docs/​car-charging.md	Documents catalogue lookup behavior.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.