[h8pewou]

No description provided.

[suda]

Thanks Adam, this was clearly a lot of work. The Tesla fix is good and the connector back-fill is something I wanted done for a while. I also like having a concrete list of what we are missing, even if most of it cannot go in yet.

I read through the files and here is where things stand, starting with what blocks the merge.

1. EV Database as the source

Every new file (8 manufacturers, 66 models, 284 variants) has a single source: https://ev-database.org/. The problem is not the quality of their data. It is that using it comes with conditions they publish at https://ev-database.org/data-services-api. Free use needs prior permission from them, automated collection is not allowed at all, and anything reused has to carry attribution with a live link. We never asked them for permission, and I do not want to abuse their work. This is why CONTRIBUTING.md asks people not to use the site, and why I removed every EV Database citation from the repo back in August.

What I want instead is manufacturers as the source of truth: a deep link to the official spec or price page for the exact trim and model year. When a manufacturer page does not exist, the press sources in CONTRIBUTING.md are fine as a fallback. If a record cannot be sourced either way, drop it rather than keep the EV Database link.

2. Validation fails

python scripts/validate.py --directory data --check-duplicates reports 14 errors on this branch (it passes on main). All of them are trim names that YAML reads as numbers: name: 50, name: 2.0, name: 300. The files are audi-q4-e-tron-35-2025, audi-q4-sportback-e-tron-35-2025, audi-q4-sportback-e-tron-45-2025, e-go-life-20/40/60-2022, mercedes-benz-eqe-300-2025, mercedes-benz-eqe-suv-300-2025, mercedes-benz-eqs-350-2022, skoda-elroq-50-2025, skoda-enyaq-coupe-iv-60-2025, skoda-enyaq-iv-50-2025, tesla-roadster-2-0-2010 and tesla-roadster-2-5-2012. Quoting the names (name: "50") fixes it.

3. The new records are stubs

The new variants have battery capacity, AC/DC kW, and WLTP range on 201 of the 284. Nothing else. No performance.total_power_kw, no efficiency.wltp_kwh_per_100km, no time_10_to_80_min. The 66 new models have no production_status, no dimensions, no seating. CONTRIBUTING.md names power and efficiency as the minimum for a variant to be useful. production_status matters especially here, because a good share of these cars are long out of production: C-Zero, iOn, i-MiEV, Fluence Z.E., e-Golf, e-Up, i3, Honda e, Fisker Ocean.

4. EU connectors on cars that are not sold here

Cadillac Escalade IQ, Optiq, Vistiq and Celestiq, GMC Hummer and Sierra EV, Chevrolet Silverado EV, all 28 Rivian variants, Acura ZDX, Honda Prologue, Dodge Charger Daytona, Jeep Wagoneer S, F-150 Lightning and Cybertruck all have ac_connector: Type2 and dc_connector: CCS2. Those cars ship with J1772/CCS1 or NACS. And tesla-roadster-roadster-2025 (190 kWh, CCS2) is not a production car.

5. Duplicates

Several of the new variants already exist under a different id:

• dongfeng-box and its two variants are the dongfeng-nammi-box / dongfeng-nammi-box-comfort-2025 we already have
• renault-zoe-ze-50-r110-2023 and renault-zoe-ze-50-r135-2024 vs renault-zoe-ze50-r110-2024 and renault-zoe-ze50-r135-2024
• renault-5-e-tech-ev40-120hp-2025 and renault-5-e-tech-ev52-150hp-2025 vs renault-5-e-tech-evolution-40-2025 and renault-5-e-tech-techno-52-2025
• dacia-spring-electric-45-2025 and dacia-spring-electric-65-2025 vs dacia-spring-expression-2025 and dacia-spring-extreme-2025
• toyota-urban-cruiser-48-8kwh-fwd-2025 vs toyota-urban-cruiser-49-fwd-2025

6. Connector back-fill on the 362 existing variants

This part is fine. Type2/CCS2 on EU-spec cars, CHAdeMO on the 2024 Leaf, same as the 164 records that already had connectors. The only thing missing is metadata.updated_at on each touched file. VERIFICATION.md asks for it on every changed record, so please add updated_at: "2026-09-30".

7. Tesla Model 3 Premium RWD

Good. Charging fields, heat pump, one-pedal driving, the rewritten notes and the two sources all check out, and leaving data_quality: estimated was the right call. This could merge today on its own.

What I would suggest

Split it into three PRs so the good parts do not wait on the rest:

1. The Tesla Model 3 Premium RWD fix. Ready now.
2. The connector back-fill, with updated_at added.
3. New vehicles, re-sourced from manufacturer pages, minus the duplicates and the US-only and non-production records, with the names quoted. A smaller batch done properly (the Stellantis vans, say, or the Renault and Dacia models) would be very welcome.

#9 is closed as a duplicate of this one; it had the same commits, just opened from your fork's main.