These are chat archives for nightscout/intend-to-bolus

3rd
Oct 2017
Scott Leibrand
@scottleibrand
Oct 03 2017 00:37
Should be easy enough to do a sed replace of curly quotes if preferences.json fails a jq syntax check.
Or just whine loudly.
We can also add curves to oref0-setup to eliminate manual json editing.
cameronrenwick
@cameronrenwick
Oct 03 2017 00:44
having been the person that brought this up..I'd say that perhaps it be suggested if you are editing/adding to the pref.json to do so in terminal (vs using ms-word etc) or risk the potential for non-ascii characters. Part of me "likes" the little idiosyncrasies of this whole aspect of openAPS as a) it isn't push 1 button and done and b) it involves a fair amount of learning on the user's side. I really fear if it's too easy that the risk to less "fully-involved" users is high. Heaven help us if someone quickly gets a rig hooked up and boluses too much insulin because of some error (on their part without doubt) and all the fantastic work of Dana/Scott is called into question.. High blood sugar will kill you in 20 years, low blood sugar can kill ya today... (sorry for the scary-boo rant)
jaylagorio
@jaylagorio
Oct 03 2017 00:47
Had a 62g carb dinner that has traditionally been impossible to predict - so much so I haven't had it in about a year. Novorapid, did some eating soon temp, another lower temp to 80 as well, and the highest reading was 197 even though I had a few light snacks after. REALLY happy with this, can't wait to see what Fiasp does once I can get into that.
image.png
Seriously, nice work on the 15 minute lookahead as well as everything else
Deweyoxberg
@Deweyoxberg
Oct 03 2017 00:49
Whoa how did you get four predictions?!
and the openaps pill
jaylagorio
@jaylagorio
Oct 03 2017 00:50
Of course the holy grail is eating and not doing anything, but if I have to announce meals I'll still count it as a serious win.
I don't know how I got four predictions, aren't there always four lines after entering carbs?
The actuals generally track where the three most convergent lines lead
Deweyoxberg
@Deweyoxberg
Oct 03 2017 00:52
I've only ever seen two
image.png
jaylagorio
@jaylagorio
Oct 03 2017 00:53
Oh, that's the AR2 line
Click the ... after the 24hr thing and if you have OpenAPS you'll get another option
Show OpenAPS Forebasts
I turned the AR2 off and turned that one on
Deweyoxberg
@Deweyoxberg
Oct 03 2017 00:54
image.png
jaylagorio
@jaylagorio
Oct 03 2017 00:55
image.png
Deweyoxberg
@Deweyoxberg
Oct 03 2017 00:55
O.O
You have more buttons!
Nightscout
version 0.9.0-beta3
yours?
jaylagorio
@jaylagorio
Oct 03 2017 00:55
Do you have the OpenAPS pill enabled in Nightscout?
I upgraded to 0.10
I think the OpenAPS pill might be 0.9.2 or later but I could be wrong
You should check out the new version, it's much, much faster
When I upgraded I made some other changes that ended up screwing the display up but the guys in the public room set me straight
Deweyoxberg
@Deweyoxberg
Oct 03 2017 00:57
hrm used to be a giant purple update button
there it is :D
jaylagorio
@jaylagorio
Oct 03 2017 01:00
Great! Kick those tires and make sure you enable the OpenAPS pill, I get most of my info from that. The purple lines are the algorithm doing the heavy lifting. The blue lines are kind of useless if you're looping.
Dana Lewis
@danamlewis
Oct 03 2017 01:05
@jaylagorio :+1: thanks for being willing to provide feedback as you test! very helpful for us to keep as much moving as possible into dev.
@cameronrenwick great suggestion if you'd PR that into the docs :+1: (edit in terminal)
garykidd
@garykidd
Oct 03 2017 01:06
is there any simple way to visualize autotune basal rate changes over time? Understanding that it should migrate over measured increments...but it'd be neat to see trends in recommendation and distinguish continuing trends from temporary migrations...
Dana Lewis
@danamlewis
Oct 03 2017 01:07
there is likely a simple way, yes :) (but afaik, no one's tackled that yet)
Deweyoxberg
@Deweyoxberg
Oct 03 2017 01:09
Le grr @jaylagorio did you run into any deployment issues? Mine keeps failing
garykidd
@garykidd
Oct 03 2017 01:09
I get that it's so easy to come up with ideas....harder to think through all the ramifications/considerations/etc. and probably harder still to develop. It'd actually be really cool to be able to manually implement autotune recommendations in the actual pump settings.
Dana Lewis
@danamlewis
Oct 03 2017 01:09
another frequent request that no one's tried to finagle is taking autotune and syncing that up with NS profile so that too doesn't have to be manually updated
Dana Lewis
@danamlewis
Oct 03 2017 01:22
@cameronrenwick your green "snake" graph on Twitter is 👌 :smile:
Deweyoxberg
@Deweyoxberg
Oct 03 2017 01:29
Uh oh
Went to update to 0.10 branch but can't get github deployment to work in azure :(
Even went so far as to delete my existing fork and start over, but still failing.
Deweyoxberg
@Deweyoxberg
Oct 03 2017 01:46
@jaylagorio : updated (phew), still no option. Assuming you're adding OpenAPS to the ENABLE field in Azure yeah?
Deweyoxberg
@Deweyoxberg
Oct 03 2017 01:52
nvm helps to read doc... buried... but doc... :D
Scott Leibrand
@scottleibrand
Oct 03 2017 02:18
@garykidd if you want to learn some basic shell scripting, you can build a script (or even a one-liner) to grep through all the autotune log files, and pull out and print each hour. you can then copy and paste that into Excel and graph it.
even in my day job, that kind of semi-automated approach is generally faster than trying to code a full solution
Tim Street
@tim2000s
Oct 03 2017 05:23
@YYCMichael Yup. The whole sequence is in there. Doesn’t look like there are cut paste errors either. Will check again later this morning.
jaylagorio
@jaylagorio
Oct 03 2017 10:22
@YYCMichael Sorry for the delay in response - I don't deploy to the cloud so I'm probably not the best person to help on that part, but yes, OpenAPS is in the ENABLE field and then you check the box to show it in the settings panel
Martin Haeberli
@mhaeberli
Oct 03 2017 16:11
is there a way to set a watchdog timer on the Edison so it reboots if nothing happens / it gets stuck for, say, 20 minutes?
Martin Haeberli
@mhaeberli
Oct 03 2017 16:17
edison/openaps dropped. T1 is elsewhere, with rig. mmtune s now fail. tried rebooting, reflashing explorer radio via cctool, hand mmtune. idea now is power off, disassemble rig, re-seat connector between edison and explorer, power on again, hope it works (or fail over to spare rig, which will need a little configuring). any other suggestions?
Scott Leibrand
@scottleibrand
Oct 03 2017 16:17
watchdog at what level? stuck in what way?
Martin Haeberli
@mhaeberli
Oct 03 2017 16:17
rig is talking to papertrail and cloud
watchdog in the sense … rig was plugged in and charging, but all comms stopped, no papertrail, etc. - power levels about 60%
just stopped looping at 01:20 AM. stopped everything.
Scott Leibrand
@scottleibrand
Oct 03 2017 16:19
was it logging anything to /var/log (syslog or similar)? or was it powered off completely?
Martin Haeberli
@mhaeberli
Oct 03 2017 16:19
i have in the past worked on some systems where you could configure it, if you didn’t poke the watchdog timer every delta-t (say, 5 minutes), then it would reset
no logging.
as far as I can tell from papertrail
Scott Leibrand
@scottleibrand
Oct 03 2017 16:19
worth logging into the rig when you can to check
Martin Haeberli
@mhaeberli
Oct 03 2017 16:19
will have to dissect logs remotely later. acted like hung or powered off completely
Scott Leibrand
@scottleibrand
Oct 03 2017 16:19
I don't know of any hardware watchdog on the Edison
could have been it overheated. I just pushed a branch to stop looping at 85C to prevent a full shutdown...
Martin Haeberli
@mhaeberli
Oct 03 2017 16:20
k thx. we can add timer to next-gen wishlist … maybe … as a safety measure … (ok, I can also look at temps i think in paper trail)
Scott Leibrand
@scottleibrand
Oct 03 2017 16:20
I think we would have to define what exactly you'd want to add
Martin Haeberli
@mhaeberli
Oct 03 2017 16:22
if a reset line is availalable, and IF a watchdog value has been programmed, failure to poke watchdog in delta-t causes reset to be asserted long enough to reboot?
Scott Leibrand
@scottleibrand
Oct 03 2017 16:22
I have no reason to think that is (or isn't) possible on the Edison
anything that shuts down the OS is likely to kill any software watchdog we tried to create
Martin Haeberli
@mhaeberli
Oct 03 2017 16:23
indeed - concur re software watchdog
about to try to remote in to logs, but - what keyword if any for logged temperature?
Scott Leibrand
@scottleibrand
Oct 03 2017 16:26
I'm not sure if temperature is logged.
If so it might have °C
Martin Haeberli
@mhaeberli
Oct 03 2017 16:34
@scottleibrand is there a command to look at edison temperature? if so, what?
Chris Oattes
@cjo20
Oct 03 2017 16:38
I think there's a hardware watchdog on an edison
But, as usual for their hardware, documentation is sparse
Scott Leibrand
@scottleibrand
Oct 03 2017 16:39
@mhaeberli sensors (from the lm-sensors package)
Chris Oattes
@cjo20
Oct 03 2017 16:40
mhaeberli: sensors returns the temperature
sensors | grep Core | awk '{print $3}' | head -1
To just get the temperature of the first core on its own
Martin Haeberli
@mhaeberli
Oct 03 2017 16:42
@cjo20 k thx - btw, did you ever get that temp into the openaps pill in NS? (I just found an earlier post from you related to this).
@cjo20 ‘we don’t need no stinkin’ documentation'
looking at logs now
Chris Oattes
@cjo20
Oct 03 2017 16:42
WRT watchdogs, there is a /dev/watchdog and /etc/watchdog.conf, which does include options related to temperature
Martin Haeberli
@mhaeberli
Oct 03 2017 16:43
k i’ll also look there
Chris Oattes
@cjo20
Oct 03 2017 16:44
#*/3 * * * * /usr/local/sbin/voltage json batteryVoltage battery | json -e "this.temperature='`sensors | grep Core | awk '{print $3}' | head -1`'" > /home/edison/myAps/uploader.json
That's how I used to generate my uploader.json, which then included the temperature in the pump pill
Martin Haeberli
@mhaeberli
Oct 03 2017 16:46
@scottleibrand re crash - syslog just dies at 01:19 this morning; no indication as far as I can tell for crash reason. then resumes 08:10 or so.
Chris Oattes
@cjo20
Oct 03 2017 16:48

You might be able to change battery-status in openaps.ini to something like

! bash -c "sudo ~/src/EdisonVoltage/voltage json batteryVoltage battery | json -e "this.temperature='`sensors | grep Core | awk '{print $3}' | head -1`'" > monitor/edison-battery.json"

That's untested and off the top of my head, so you'd have to test it yourself

There's going to be a problem in that with conflicting " I think
Martin Haeberli
@mhaeberli
Oct 03 2017 16:51
K thx I’ll play ...
David
@davidkeddydb
Oct 03 2017 18:43
Screenshot 2017-10-03 14.35.28.png
I have been trying to get @danamlewis like results using maxSMBBasalMinutes and I have mine set at 105. I am not manually bolusing at all and just using an eating soon and letting the rig handle it. I could be bolusing earlier for sure as I am not using FIASP and using Humalog but also considering going up to 120 minutes. Anyone else up at 120 minutes?
Diabetic_Dude
@diabetic_dude_twitter
Oct 03 2017 19:20
I am at 90 mins now with maxSMBBasalMinutes and seeing closer to 250+ with high carb meals (77+) Only when I do not bolus, I am using FIASP. I am also inputting carbs into NS and using the eating soon button about 15 minutes before and again about 1.5 hours after the meal. I need to update however, my dev rig is about 5+ days since the last update. I need to change my settings in a big way going to research it and considering upping another 15 just like you did.
WOAH.PNG
jaylagorio
@jaylagorio
Oct 03 2017 19:21
What's the default right now?
cameronrenwick
@cameronrenwick
Oct 03 2017 19:22
ok crazy question about installing an Anaren 66089-093 antennae on my EB. They are 30mm 915MHZ but the thing is actually about 4 inches long... is it supposed to be trimmed or..?
Diabetic_Dude
@diabetic_dude_twitter
Oct 03 2017 19:23
@jaylagorio default is 30
jaylagorio
@jaylagorio
Oct 03 2017 19:23
Thank you!
I'm considering moving it ahead to 45 minutes tonight and doing another 60g carb dinner to see if I can get the peak reduced
PieterGit
@PieterGit
Oct 03 2017 19:24
not a good day with 0.6.0 dev checkout today (754WW, Dex G4, DIA=5, rapid-acting, SMB).Quite some oscillation of bg's, eating extra sugar was necessary to prevent hypo's decreasing max_iob, disabling autotune and setting "SMBBasalMinutes": 20. Other options/things I should consider?
Diabetic_Dude
@diabetic_dude_twitter
Oct 03 2017 19:25
WOAH @PieterGit we need to merge your build + mine and things should be perfect!
PieterGit
@PieterGit
Oct 03 2017 19:26
double check: disabling autotune with cd myopenaps; mv autotune autotune0 ; mkdir autotune and disabling crontab. is that ok?
@diabetic_dude_twitter : are you sure your insulin and infustion set are ok, with that many boluses one would expect bg to go down
Diabetic_Dude
@diabetic_dude_twitter
Oct 03 2017 19:27
@PieterGit yes, I only get the result when eating high carb meals and announcing and inputting carbs into NS
if I bolus I get near perfect results
tested this morning with a smoothie and did not go over 200
actually did not go over 180
it was like SMB was hitting constant but not hard enough, it hit me over and over again for hours until I gave up and manual bloused 4 units
PieterGit
@PieterGit
Oct 03 2017 19:30
on the other hand: when still using boluses the new adapted 0.6.0 algorithms should not cause hypo's. i upgraded a good working 0.6.0 dev rig to latests (thought it was my test rig :worried: )
Diabetic_Dude
@diabetic_dude_twitter
Oct 03 2017 19:36
damn that sux... @PieterGit it is hard for me to confuse my rigs. they look different, have different batteries, different names and different passwords. Because I tend to do the same type of thing you did when I get in a rush or get distracted.
Glenn Primack
@primags
Oct 03 2017 19:42
I'm not having any luck with my WPA2-Enterprise setup at work to get my rig online. I'm wondering if there is a way to do TCP/IP through the USB/Serial connection. Or some other way of me getting me on the network.
The wifi logs don't tell me enough info to be able to discern what is happening
Scott Leibrand
@scottleibrand
Oct 03 2017 19:59
@primags the Edison does IP over usb0 by default. you could set up your laptop to do Internet Connection Sharing to its USB interface.
if you do that you'd need to modify the Edison's usb0 network config to make it request an IP from the PC, or statically configure the PC to use one in the same subnet as the Edison uses
@PieterGit could i go over your NS to review where you think it's being too aggressive? or could you post screenshots here?
Glenn Primack
@primags
Oct 03 2017 20:02
@scottleibrand Would that be through the OTG side?
Scott Leibrand
@scottleibrand
Oct 03 2017 20:04
@diabetic_dude_twitter @davidkeddydb @PieterGit we have not tested no-bolus eSMB with rapid-acting insulin, just with Fiasp. But one other thing we've noticed with Fiasp is that if you want to reduce the meal spike you can pre-enter the carbs and get the SMBs started earlier. We haven't seen that cause any pre-meal lows, but obviously you'll want to try gradually earlier announcements to test it safely.
@primags yeah
Martin Haeberli
@mhaeberli
Oct 03 2017 20:04
just found this pointer to Intel watchdog stuff - this may all be known already - https://www.intel.com/content/www/us/en/embedded/solutions/iot-gateway/enabling-and-configuring-watchdog-timer-app-note.html
Scott Leibrand
@scottleibrand
Oct 03 2017 20:05
@mhaeberli one challenge with watchdog is knowing whether it's helping: we had some watchdog stuff enabled for the Pi, but no one could ever reproduce the failure scenarios in order to tell if it helped or not
Diabetic_Dude
@diabetic_dude_twitter
Oct 03 2017 20:06
@scottleibrand pre enter carbs and set for eating soon or just pre enter the carbs?
David
@davidkeddydb
Oct 03 2017 20:07
@scottleibrand I am not shy. Will test that out more aggressively than I have been. I did up my minutes also to 120
Glenn Primack
@primags
Oct 03 2017 20:11
I'll give that a try, thanks @scottleibrand
Scott Leibrand
@scottleibrand
Oct 03 2017 20:13
@diabetic_dude_twitter your call. entering carbs is a way stronger signal than just doing an ES, but they would be additive if you wanted to use both
Diabetic_Dude
@diabetic_dude_twitter
Oct 03 2017 20:14
thnx