View on GitHub

32.22-to-beta

An early warning system that Teslas on 2021.32.22 are receiving Beta

Background

While waiting for Tesla to roll out Full Self Driving (FSD) Beta to my car, I was constantly refreshing web pages from websites like TeslaScope to find out how many cars were on each firmware revision and the prior version that the current version replaced. TeslaScope also shows the model and trim level of the vehicles receiving the updated firmware. I grew tired of refreshing two different web pages and trying to figure out for myself what, if anything, has changed. I knew there was a better way to detect that Tesla was upgrading cars just like mine.

The firmware update path from Production to Beta

On September 25th, Tesla pushed firmware 2021.32.22 to every car with the FSD option purchased or an active FSD subscription. Owners needed to push a button indicating their desire to receive FSD Beta and their acceptance of the ground rules around testing it. Tesla started monitoring how safely the owners drove and started assigning Safety Scores. After a short while, Tesla moved a portion of the vehicles in the “FSD ready” population from 2021.32.22 to 2021.32.25, but most remained on 2021.32.22.

Once drivers surpassed 100 miles with a perfect 100 Safety Score after 7 days, Tesla started delivering FSD Beta to those vehicles by updating them to 2021.36.5. Some vehicles were updated from 2021.32.22 and some from 2021.32.25. My car never left 2021.32.22, so I watched that population fairly closely. Any Tesla that was “FSD ready” could be upgraded to FSD Beta, though, so I also specifically wanted to know each time a vehicle similar to mine was upgraded. FSD Beta went through a few updates within the first couple of weeks, so the current target firmware version changed every few days. Right now, 2021.36.5.3 (aka FSD Beta 10.3.1) is the latest and the version this script is monitoring.

Watching the population on 2021.32.22 change combined with analyzing each new FSD Beta install gave me an early indication of when Tesla was deploying FSD Beta to cars very similar to mine.

Node-RED

Node-RED is a visual programming tool that greatly simplifies the act of assembling small, simple scripts. Node-RED also includes numerous modules (called Nodes) and integrations with external tools and sites. Node-RED enables rapid prototyping of functional code by dragging and dropping nodes onto the design space and allows for customization of the code using a variety of programming or scripting languages. A collection of Nodes that interoperate with one another on a single tab in the editor is called a flow. I have exported the flow that I built and published it here as JSON code.

How to deploy my code

Clone the repository, fire up Node-RED, click the 3-bar “hamburger” menu in the upper right corner, and click on Import. Navigate to the location where you cloned the repo and import the JSON file. You may need to locate a new tab where the flow opened after being imported. Open the debug window, click on the red Deploy button in the upper right corner, and you will see at least the first of the two example lines shown below in the Output section. You will see the number of vehicles running 2021.32.22 and any Model 3 LR AWDs that have recently been upgraded from 2021.32.22 to 2021.36.5.3.

What is it designed to do?

There are two primary purposes of the flow. First is to report the number of vehicles reported by the TeslaScope website as currently running 2021.32.22 – the primary version that is ultimately upgraded to FSD Beta. Second, I wanted to receive a notification whenever a vehicle that is effectively identical to mine is upgraded from 2021.32.22 to 2021.36.5.3.

Vehicles Running 2021.32.22

The first function requires ingesting TeslaScope’s detail page for the 2021.32.22 firmware, found here:

https://teslascope.com/teslapedia/software/2021.32.22

With the page ingested, I used an HTML Parser node to pull in the HTML contents of that page. I then used a Change node to only retain object 76 from the fetched page. That object contains a sentence showing the number of vehicles currently running this version. I then split the contents of that sentence into an array. Next, as I traverse the array, I convert each entry from a text string into a numeric value. If the resulting numeric value is greater than 0 (which only happens once), I convert the value back to a string and append the text “ on 32.22” to the end and print it out. Finally, I pass the payload through a Filter node (previously called Replace By Exception or RBE). This ensures that the resulting message will be passed only if it has not changed. Then the payload is passed to the Debug node, which dumps the payload to the debug window.

Any Model 3 LR AWD moving from 2021.32.22 to 2021.36.5.3

It is important to note that the number of vehicles running 2021.32.22 is not enough to tell the whole story. That could be on any model Tesla. I also needed to see when a car just like mine received the upgrade I was waiting for.

The second function in this flow requires ingesting TeslaScope’s detail page for the 2021.36.5.3 firmware, found here

https://teslascope.com/teslapedia/software/2021.36.5.3

With the page ingested, I used an HTML Parser node to pull in the HTML contents. Then, I passed those contents on to a Change node that only retained object 174 on the page – which is the table showing the detail of all vehicles that received that update. With the table captured, I ran it through a Split node to break it up into lines, a Function node that trims leading and trailing white space from all cells in each row, a Join node that converts the rows of trimmed cells into an array, a Switch node that only retains records where the first field matches 2021.32.22, and another Switch node that only retains records where the second field matches “Model 3 AWD LR”. Once the records not pertaining to cars like mine being upgraded to FSD Beta are filtered out, I send the remaining records to a different Debug node which, again, sends the output to the debug window.

Output

The two functions in this flow dump very simple output. First, if the number of vehicles running 2021.32.22 has changed since the last time it was reported, the first function will dump the new value (as shown in this example)

1336 on 32.22

Second, any time a Model 3 AWD LR is upgraded from 2021.32.22 to 2021.36.5.3, the second function will dump the details of that vehicle as a printed array, for example

2021.32.22,Model 3 AWD LR,3.0,United States,1 day ago

The array fields are Prior Firmware Version, Vehicle Model and Trim Level Description, Autopilot Hardware Version, Country, and time since the update. The output is very basic and is not pretty at all. I could have just printed “Another one!” but I wanted to make sure I saw how long ago the update occured, represented in minutes, hours, or eventually days.

Notes

If your car moves from 2021.32.22 to 2021.32.25 instead of 2021.36.5.3, you may easily modify the first Switch node in the second function so that it looks for that firmware version instead of the one my car was running. Also, once 2021.36.5.3 gets replaced by a newer version of FSD Beta, you will want to modify the end of the URL that the second function ingests to match the newest firmware being rolled out.

Looking at the Timestamp node I use to kick off both functions, I configured the triggering action to repeat at the top of every hour between 4am and 8pm since I knew I would only react to early indications of movement during those hours. There’s no point in querying TeslaScope’s web page more frequently than that and there’s no point in querying their site if you’re not planning to take any action as a result of any notification you may receive.

It is important to be respectful of other peoples’ resources here.

Finally, I terminated the functions with a print to the debug window, but it is up to you to do something different with the output. You may wish to trigger a pop-up on your phone, send yourself a text message, or post a message to or send a private message via Twitter. The debug window is just a dummy output that is useful for functional test but is not very useful beyond that. A friend of mine had already linked his Node-RED instance to Twitter via the API, so it was easy to just give him this code and send the output to me via a private tweet.

If you let your imagination run wild, you could even use MQTT with a switch module from Shelly to turn on a light under a bottle of Tesla Tequila. How awesome would that be as a way to find out the Beta rollout is expanding?

License

This software is published under the Apache 2.0 license. Do whatever you please with it, but please respect TeslaScope’s resources and wishes if they ask you to reduce the refresh frequency.