TeslaMate drive-recovery toolset

Recover the trips TeslaMate loses when a Tesla API outage leaves your car stuck in driving.

When Tesla's API goes down (403 errors etc. like we saw mid June 2026), the streaming API keeps working, so TeslaMate keeps logging GPS positions but stops recording battery range, temperatures etc. The result: your car shows as driving for hours or days, and a whole run of separate real trips gets swallowed into one giant, never-ending "drive" with no range, and no distance, duration, temperature, etc.

This toolset splits that mess back into the individual trips it should have been, fills in the missing battery range, and closes each trip the same way TeslaMate normally would — so they show up correctly in your dashboards.

This might be for you if:

• Your car has been stuck showing as driving for hours or days.
• The trips you actually took are missing — swallowed into that one never-ending drive.
• It lines up with a known Tesla API or TeslaMate outage.

Nothing is car-specific or hardcoded — everything (your car, the outage window, the battery calibration, the location) is detected from your own database.

⚠️ This writes to your TeslaMate database. It is read-only until you add --commit, and it takes a database backup before its first write — but you are editing real data. Take your own backup first and run the read-only preview before committing.

Unofficial community tooling — not affiliated with or endorsed by the TeslaMate project. No warranty; use at your own risk.

---

What you need

• The machine where TeslaMate runs (your Raspberry Pi, NAS, server, etc.) and a terminal on it. These are command-line scripts — there's no app to click.
• Python 3.8+ and the psycopg2 database driver (installed below).
• Access to your TeslaMate PostgreSQL database (you already have this if you run TeslaMate).
• For the final step, docker compose access to your TeslaMate stack — which is why you run these on the same machine as TeslaMate.

---

Built with Claude Code. So far this has only been tested on the author's own TeslaMate database and hasn't been independently verified by anyone else — please read the code and back up your database before running it.

Quick start

Run these in a terminal on the machine where TeslaMate is installed.

1. Get the code. If you have git, clone it:

git clone https://github.com/yanou-g/teslamate-stuck-drive-recovery.git
cd teslamate-stuck-drive-recovery

No git? On the project's GitHub page click Code → Download ZIP, unzip it, then cd into the unzipped folder.

2. Install the one dependency:

python3 -m pip install psycopg2-binary --break-system-packages

3. Tell the scripts how to reach your database. These are TeslaMate's defaults; PGPASSWORD is the DATABASE_PASS value from your TeslaMate .env file:

export PGHOST=127.0.0.1 PGPORT=5432 PGDATABASE=teslamate PGUSER=teslamate
export PGPASSWORD=your-database-password

4. See what it would do. This only reads — it writes nothing. Always run it first:

python3 preview.py

5. If the trips, distances and ranges look right, do it for real. This backs up the database first, then asks once before writing anything:

python3 run.py --commit --compose-dir /path/to/teslamate

That's it. run.py runs the whole recovery in order — detect the outage, calibrate, split the stuck drive into real trips, fill in the range, and close each trip — and prints what it did at every step.

Run it from your TeslaMate folder (the one with docker-compose.yml), or point it there with --compose-dir. If you have more than one car, add --car <id>.

Optional: backfill the outside temperature

The car's temperature sensor also goes blank during an outage. You can fill in an approximate outside temperature from historical weather data (free, no API key) by adding --weather:

python3 run.py --commit --compose-dir /path/to/teslamate --weather

This is clearly-labelled ambient air temperature, not real sensor readings.

---

Is it safe?

It's built to be careful:

• Read-only by default. Nothing is written unless you add --commit.
• Backup first. run.py --commit takes a pg_dump before the first write.
• One confirmation, then it runs unattended (add --yes to skip even that).
• Surgical writes. It only fills in missing data on the trips it recovered. Real recorded data and healthy drives are never overwritten.
• It refuses to do something dumb. If the battery calibration looks unreliable, or a trip would be damaged by closing it too early, it stops and tells you.

If you'd like to rehearse the whole thing risk-free first, restore a copy of your database into the throwaway test stack in docker_test/ and run the scripts against that before touching your live data.

---

Want more control?

run.py is just an orchestrator. Each step is a standalone script you can run by hand if you want to inspect or adjust things between stages:

Step	Script	What it does
—	preview.py	Read-only full dry run — shows the trips it would create
1	1_diagnose.py	Report the outage and the stuck drive (read-only)
2	2_calibrate.py	Work out this car's battery range per % charge
3	3_segment.py	Split the stuck drive into the real individual trips
4	4_reconstruct.py	Fill in the missing battery range on each trip
5	5_finalize.py	Close each trip natively (distance, addresses, etc.)
6	6_weather.py	Optional approximate outside temperature
—	selftest.py	Offline math check, no database needed

Every step is dry-run by default and prints a NEXT: hint for the following one. Run any of them with --help for its options.

---

How does it actually work?

The interesting parts — how it tells a real outage apart from a sleeping car, why battery calibration is trickier than "range ÷ percent", how the missing range is reconstructed, and the one TeslaMate quirk that can silently delete a drive — are written up in HOW_IT_WORKS.md.

---

License

BSD Zero Clause License (0BSD) — do anything you like with this, no attribution required. Provided as-is, with no warranty.