I've been building the most advanced automotive diagnostic app on the market
I used to be a mechanic, and spend a lot of my free time working on project vehicles, motorcycles, tractors, trucks etc. I wanted a tool to track the tasks I needed done, then a way to easily access VIN info for parts search, then I DMV records, and finally built an iphone app to scan the OBD2 diagnostic codes and hook them up to a custom ai diagnostic prompt.
It's been a blast to work on, I need to focus on marketing it and get some users that aren't close friends, but it's been serving my purposes this summer so far.
I've seen nothing with anywhere close to this depth of diagnostic reporting.
From your description I hoped that you would be analyzing audio to solve problems that don't raise codes.
Years ago my daily driver had a noisy wheel hub bearing, so when I heard a similar noise in a different vehicle this year, I assumed it was the same problem. However, when a mechanic investigated, it turned out to be an issue with the breaks instead. I have now learned to discriminate between a bearing and a brake issue, but there are a ton of other conditions I've never heard before that would be obvious to an experienced repair technician.
I bet an app could become as good a seasoned mechanic at identifying mechanical issues from audio samples.
I agree, its an idea I've been playing with for a while also. The difficult part would be isolating the sound, but perhaps with new ai audio models that has become trivial, its definitely worth revisiting
Don't worry about isolating the sound too much. If your human ear can identify a bad bearing even if the radio is on, you can train a model to do the same. Some of this will come just from spectral analysis, which makes it easy to distinguish the rumble of a V8 a belt squeak.
You can also probably guide a user through a couple of system changes to isolate sounds. For instance, a sound related to the alternator may change with both RPMs and electrical load. A sound related to the serpentine belt might be affected by engaging/disengaging a mechanical load like the A/C. A wheel bearing or brake noise will track with the rolling speed, not the RPMs, and may be affected by steering (a hard turn to the right may unload the left side of the vehicle.)
I'm blown away by how much diagnostic info a good repair tech can get just from listening to a car; I would pay to have this kind of guidance in a phone app.
Very cool indeed. Make it really easy also for non mech people to diagnose and get an idea on a problem while having only an OBD2 adapter.
How are the reparations prices calculated? Is there some kind of dataset you builder or is it just a guide prompt to an LLM?
Thanks, yes its been empowering my less handy friends to fix things on their own, and building their confidence. I figure, we can't have right to repair without a more informed/confident public to demand better, repairable, and surveillance free cars.
For Pricing calculations it's a mix of the fetched tool search prices data and guided prompt. There are lots of published records on what hours jobs _should_ take and average labor hours per region.
At the wire level, no: it speaks generic OBD2 (ELM327), so it sees powertrain codes and standard PIDs. It won't talk to every module on a VAG car the way VCDS does — no ABS/airbag/body modules, no adaptations or coding. If you need that, VCDS stays on your bench.
Where it is make-specific is the interpretation layer. The code database includes manufacturer-specific DTCs for the 10 major makes, and there's per-model data down to the generation level — known failure patterns, which codes are common on which engines. So a P0171 on an E46 gets read in the context of "this generation eats intake boots and DISA valves," not a generic lean-code writeup. The AI report is conditioned on the exact year/make/model/engine, which is where most of the diagnostic depth comes from.
That being said, I'm continually expanding support, it really just depends on what sort of barriers I run into with these bluetooth scanners. For certain types of diagnostics (large truck / Buses etc) I'll likely need to make a separate desktop app that connects directly wired.
the Vgate works good, there is a spec for the OBD2 Devices over bluetooth, so any of the cheap ($20-$30 range) should work fine, I haven't found one that hasn't connected to the app yet https://www.amazon.com/dp/B06XGB4873?th=1&tag=crewchief-20"
Describe your symptoms in as much detail as detail as you like and get a full diagnostic report with parts links, tutorials, price estimates, guides for diy
Includes car sales tool, generates all the documents you need for the dmv in your county/state
iPhone all connects to Bluetooth obd2 sensors for check engine lights and live driving data
I'm in talks with some factories that manufacture the scanners, I'd like to produce branded versions and ship them out automatically to users who signup for the paid plan, but I need to get some initial users first to verify the idea. Eager to move onto the next steps and build all of that out
Crew Chief (https://crewchief.cc) — a Vehicle diagnostic and management tool. Plug in your OBD2 codes (or just describe symptoms) and get a structured diagnosis in under 30 seconds: ranked probable causes, DIY vs. shop cost estimates, severity rating, and matched parts/repair videos.
I have too many project cars and bikes, I wanted one place to store vin numbers for searching parts, and then just kept adding useful features.
Supports 16 vehicle types (cars, trucks, motorcycles, boats, tractors, ATVs, RVs, etc.), not just cars. Also includes maintenance tracking, a browser extension that auto-fills your vehicle info on parts sites like RockAuto and AutoZone, a community-vouched trusted shops map, and a vehicle selling wizard with state-specific bill of sale generation.
Free tier gives you 1 vehicle with a full diagnostic.
Mikeal was extremely welcoming to me when I met him over a decade ago, It was either JSConf or NodeCamp, He was always super insightful and encouraging to my personal growth as an engineer and open source contributor. NodeJS wouldn't be what it is without him, especially the community around it in the early days. I got to work with him for a bit at protocol labs. This is a major loss to open source
I live at a _very_ loud and busy intersection in downtown manhattan, it's usually gridlocked every day from 4pm - 7pm at least. That hasn't happened once since congestion pricing kicked in (I'm enjoying the relative quiet, but I do sort of miss the chaos). The first two days, everything was covered in snow down here. I wonder how much this will hold up into summertime.
As an unrelated aside a new bmw sedan driving the wrong way on mulberry street intentionally ran into me at a crosswalk today. So don't let your guard down lol
