# AI use cases for Car Watch (ThinkOffApp/CarWatch) Repo: [ThinkOffApp/CarWatch](https://github.com/ThinkOffApp/CarWatch) | Use cases: 4 | Page: https://dedupedai.com/use-cases/thinkoffapp-carwatch/ CarWatch runs a local AI agent on a Raspberry Pi 5 in a car, which joins chat rooms, answers questions from the owner's manual, and reports car events. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Ask the car about a warning light with no internet: https://dedupedai.com/use-cases/thinkoffapp-carwatch/ask-car-warning-light-offline/ - Get engine and fault events posted to the family room: https://dedupedai.com/use-cases/thinkoffapp-carwatch/car-posts-events-to-family-room/ - Pick the car's model by its measured speed on the Pi: https://dedupedai.com/use-cases/thinkoffapp-carwatch/pick-car-brain-by-measured-speed/ - Lock the car from your phone after an alert: https://dedupedai.com/use-cases/thinkoffapp-carwatch/lock-car-from-phone-after-alert/ ## How to ask the car about a warning light with no internet with Car Watch Repo: [ThinkOffApp/CarWatch](https://github.com/ThinkOffApp/CarWatch) | Difficulty: Advanced | Time: 4 hours | Basis: From the README Ask the car by voice what a warning light means; it answers from its own owner's manual. **Who it is for:** A driver with a Car Watch Pi in the car who sees the yellow tyre light while driving and has no phone signal. **The problem:** Warning lights are hard to decode mid-drive, and the owner's manual runs to 489 pages. **The outcome:** A spoken answer with a manual page citation and live tyre pressures, produced with the internet switched off. ### Steps 1. Clone the repo onto the Raspberry Pi 5 with `git clone https://github.com/ThinkOffApp/CarWatch`. (inferred, check the README) 2. Reboot the Pi so the systemd services self-start the model server, room agent and voice listener. (inferred, check the README) 3. In the driver's seat, ask the question out loud, for example 'What does the yellow tyre light mean?'. 4. Listen for the answer from the car's speakers and check that it names a manual page. 5. Ask about something the manual does not cover and confirm the car refuses to guess. ### Prompt to paste into your agent ```text Using the CarWatch repo at https://github.com/ThinkOffApp/CarWatch on , send the room agent the question 'What does the yellow tyre light mean?' and return the answer text with its page citation. Verify that the answer cites a manual page and that the test ran with no internet connection. ``` ### Caveat The README's measured speeds apply to a Pi 5 16 GB running one 14.3 GB quant, and a model that size takes about three minutes to load from microSD. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/thinkoffapp-carwatch/ask-car-warning-light-offline/ ## How to get engine and fault events posted to the family room with Car Watch Repo: [ThinkOffApp/CarWatch](https://github.com/ThinkOffApp/CarWatch) | Difficulty: Intermediate | Time: 2 hours | Basis: From the README The car posts engine reads and stored fault codes into a chat room when real events happen. **Who it is for:** A household that shares a chat room and wants to know about a car fault without someone checking the dashboard. **The problem:** A stored fault code or hybrid milestone is easy to miss if nobody is looking at the car. **The outcome:** Room messages with the event, the time and a plain-language explanation from the manual, posted only on real events. ### Steps 1. Clone the repo with `git clone https://github.com/ThinkOffApp/CarWatch` and set up the Pi as described in the README. (inferred, check the README) 2. Connect the car's agent to your GroupMind room, using the name you chose, such as @gle. (inferred, check the README) 3. Drive the car and let the engine reads run, then check the room for event posts. 4. Confirm the post names the stored code and includes the explanation from the owner's manual. ### Prompt to paste into your agent ```text Using the CarWatch repo at https://github.com/ThinkOffApp/CarWatch, configure the room agent for the GroupMind room , then report the last three event posts it made. Verify that each post has a timestamp and that no post appears without a real car event. ``` ### Caveat The README marks this use case as proven on the author's car, so other cars and OBD setups are not covered by the claim. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/thinkoffapp-carwatch/car-posts-events-to-family-room/ ## How to pick the car's model by its measured speed on the Pi with Car Watch Repo: [ThinkOffApp/CarWatch](https://github.com/ThinkOffApp/CarWatch) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README Compare each model's measured speed on the Pi, then swap the running model from your phone. **Who it is for:** An owner who wants quick answers to simple questions and the larger model for manual questions. **The problem:** A large model answers slowly on a Pi, and a small model cannot answer manual questions well. **The outcome:** The running model is chosen from the Model zone, based on speeds the Pi itself measured. ### Steps 1. Open the phone dashboard and go to the Model zone, which lists every .gguf file on the box. 2. Compare the prompt and generation columns for each model against the reference Pi 5 16GB table. 3. Tap the model you want; the swap waits until any answer being generated has finished. 4. Call `GET /api/models` through the auth gate to confirm which model is now running. (inferred, check the README) ### Prompt to paste into your agent ```text Using the CarWatch repo at https://github.com/ThinkOffApp/CarWatch on , call GET /api/models and list each model with its size and measured generation speed, then name the running model. Verify that each reported size matches the size of its .gguf file on disk. ``` ### Caveat The speeds come from one Pi 5 16GB, so another device will give different numbers and the table changes as models are benched. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/thinkoffapp-carwatch/pick-car-brain-by-measured-speed/ ## How to lock the car from your phone after an alert with Car Watch Repo: [ThinkOffApp/CarWatch](https://github.com/ThinkOffApp/CarWatch) | Difficulty: Advanced | Time: 3 hours | Basis: From the README After the car reports an open window or unlocked door, lock it from your phone or watch. **Who it is for:** A car owner away from the car who gets a room alert about an open window or a door left unlocked. **The problem:** The car can report a problem, but without a remote path you cannot act on it while away. **The outcome:** The car is locked from the phone or watch, and the room records the action. ### Steps 1. Set up Tailscale on the Pi and on your Home Assistant machine using docs/remote-access.md in the repo. 2. Confirm the Pi reaches Home Assistant at its Tailscale address from outside your home network. 3. When the car posts an alert to the room, reply with the lock request from your phone or watch. 4. Check the car's lock state in the room after the reply, using the cloud read. ### Prompt to paste into your agent ```text Using the CarWatch repo at https://github.com/ThinkOffApp/CarWatch and the setup in docs/remote-access.md, send a lock request to the car from and report the lock state read back afterwards. Verify that the reported state is locked and that the request was made from outside the home network. ``` ### Caveat The README says unlock, open and start are deliberately not implemented, so this use case only covers locking and closing a window. Risk, credentials: This setup needs a Home Assistant login and a Tailscale mesh, so protect both accounts and the Pi's access to them. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/thinkoffapp-carwatch/lock-car-from-phone-after-alert/