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

28th
Sep 2016
garykidd
@garykidd
Sep 28 2016 01:40
curious if anyone has managed to install new batteries in dex g5 transmitters upon their expiration.
I thought I was going to hate not being able to recharge the dex transmitters when I switched from enlite setup...but it ended up being way better in my estimation.
Tim Street
@tim2000s
Sep 28 2016 05:23
There's a few people in London that I know have done it @garykidd . Once you change them, you have to run with xDrip+ as the transmitter no longer works with the iPhone.
On a different question, while I know Loop runs on the iPhone, has anyone ported the OpenAPS code to Swift?
Dana Lewis
@danamlewis
Sep 28 2016 05:30
No, oref0 (the algorithm) is not yet ported
Tim Street
@tim2000s
Sep 28 2016 05:31
@danamlewis Okay. I was digging around last night on the web and found a Python >> Swift tool, so I thought it might be an interesting challenge!
Dana Lewis
@danamlewis
Sep 28 2016 05:34
There are many who would love for someone like you to take it on :)
Tim Street
@tim2000s
Sep 28 2016 05:39
Ha. I've gone from I'm rubbish at coding to this in two months... I'm still rubbish at coding, but might as well have a hack. Good job there's a MacBook on the way!
Dana Lewis
@danamlewis
Sep 28 2016 05:40
:) this is what I think about my coding: http://xkcd.com/1739/ (but then often it works, so yay :))
Tim Street
@tim2000s
Sep 28 2016 05:42
So very familiar 😂😂
Scott Leibrand
@scottleibrand
Sep 28 2016 06:01
first pass at updating docs for oref0-setup: openaps/docs#393 - please read and comment on the PR if you're interested
Matthias
@ceben80
Sep 28 2016 06:37
Great @scottleibrand
Tim Street
@tim2000s
Sep 28 2016 08:38
Just a quick one - who's the London based person who is under the care of Guy's Hospital and spoke at their Junior Doctor's education session about OpenAPS? Thought it would be interesting to link up as that was mentioned when I raised the topic at clinic recently. There can't be that many hospitals in the U.K. With multiple openaps users 😀
Tim Street
@tim2000s
Sep 28 2016 11:44
Marvellous. G5 transmitter has arrived. Think I need to build a local mongo db. Anyone point me at instructions?
(For local nightscout)
C-Ville
@C-Ville
Sep 28 2016 11:50
@tim2000s if you check the github from @SandraK82 there is a setup skript for NS on a pi. I wrote some lines up as a Gitbook, but it is still in German, however I I stole most of my information from https://gist.github.com/johnmales/1b3c927f2a56aae640b4b2cd0298b1e7. If you are on a pi or Edison you will most likely need to update node and npm as far as I remember. I run Mongodb and NS on my Edison with ubilinux
@tim2000s for mongodb on my Edison I downloaded the 3.2.9 tar.gz from mongo.org and extracted that. As this is not a clean installation you will need to run mondod with several arguments. I start with
screen -dmS mongo [path to mongod]/mongod --storageEngine=mmapv1 --rest --dbpath [path to database]
Tim Street
@tim2000s
Sep 28 2016 12:50
Thanks!
David
@davidkeddydb
Sep 28 2016 13:02
If anyone has it, can you post the link with where I can find information on the Nightscout prediction of bgs that it uses? I know it displays two separate prediction lines which I am guessing are likely outcome depending on next reading into NS in either a +/- of bgs in either direction. I just want to find out more of how it looks at trend, etc to extrapolate that.
David
@davidkeddydb
Sep 28 2016 13:24
Openaps is very clear on the prediction of BGI in the reference, so I understand that. I am just curious as to how/why the NS website gives blue split lines going out from the current bg both up and down.
Gina
@gclyon
Sep 28 2016 14:24
It seems I've fried my G5 receiver. It was just plugged into my raspi and noticed it was super hot. Hit the button and nada. :weary:
Thinking back, I believe I had the code to send more power to usb active in the code. Anyone ever had this happen?
Chris Oattes
@cjo20
Sep 28 2016 14:28
"Sending more power" should only actually increase the allowed current draw, it doesn't force the device to take more
Gina
@gclyon
Sep 28 2016 14:30
Nonetheless, how do I undo that line?
sudo bash -c "echo -e \"#Enable Max USB power\nmax_usb_current=1\" >> /boot/config.txt"
just change that to "Disable"
?
Chris Oattes
@cjo20
Sep 28 2016 14:32
=0 I guess
But, as I said, that shouldn't cause anything to break. It just means it'll provide more power if the device asks for it, which it shoukdn't do if it can't handle it
Gina
@gclyon
Sep 28 2016 14:39
So, how honest am I to Dexcom when explaining?
:flushed: Just had it charging... ...Openaps who? :grimacing:
C-Ville
@C-Ville
Sep 28 2016 14:45
yepp, of course you did charge with the original charger provided by Dexcom if I understood you correctly ;-)
Gina
@gclyon
Sep 28 2016 14:51
Uuummmmm, yeaahhhh.
Actually though, as I recall, it only came with a cable. (no adapter)
@SandraK82 I have a static ip address on my raspi. Does the code default to the loopback address 127.0.0.1:1337? If so, where can I modify this
Dana Lewis
@danamlewis
Sep 28 2016 14:57
@tim2000s you're probably thinking of @oskarpearson :)
@davidkeddydb ah if you are talking about the blue lines, that's the AR2 forecast in NS
Chris Oattes
@cjo20
Sep 28 2016 14:59
@tim2000s I'm London-based and I go to Guys, but I haven't given any talks. What Dana said about Oskar is probably right :)
Sandra Keßler
@SandraK82
Sep 28 2016 15:17
@gclyon the Setup uses LocalLoopback for ns2mongo this works Even if your ip is changed. 127.0.0.1 is always the address of a system
Scott Leibrand
@scottleibrand
Sep 28 2016 16:06
@davidkeddydb AR2 is basically a 95% confidence range for future BG based only on the last couple readings (no knowledge of IOB, COB, or anything else). if you're instead talking about the purple prediction lines, those are from the predBG arrays generated by oref0.
garykidd
@garykidd
Sep 28 2016 16:35
@scottleibrand aI'm figuring you didnt get a chance to try chging that default variable in mmtune.py on one of oyur rigs..right?
Scott Leibrand
@scottleibrand
Sep 28 2016 16:46
no, I got distracted with end-to-end testing of the oref0-setup script and updating the openaps/docs for it
h/t @jbwittmer for first spotting
katie disimone
@Kdisimone
Sep 28 2016 17:26
Wow. What's the word on the guardian 3 sensors? Accurate? Reliable?
Dana Lewis
@danamlewis
Sep 28 2016 17:27
I have a friend who did their clinical trial and she said they’re a lot better and usable for looping
Scott Leibrand
@scottleibrand
Sep 28 2016 17:28
approval letter says they can be used for the 670G automatic basal adjustments, but not directly for manual (bolus) dosing
Dana Lewis
@danamlewis
Sep 28 2016 17:29
but humans do what they want to do
Scott Leibrand
@scottleibrand
Sep 28 2016 17:30
off-label use?!? no way! ;-)
Pete Schwamb
@ps2
Sep 28 2016 17:31
Wow, this is ahead of schedule! At the beginning of the year, medtronic was saying they wanted to have it on the market by April 2017.
Dana Lewis
@danamlewis
Sep 28 2016 17:31
:+1:
Scott Leibrand
@scottleibrand
Sep 28 2016 17:32
might still take them a few months to get production ramped up. will be interesting to see when they say they plan to start selling it.
but yes, approval was crazy fast. :-)
live4sw
@live4sw
Sep 28 2016 17:39
Unfortunately they are still saying Spring 2017 launch
The FDA approval came earlier than expected and they weren't ready for it
Dana Lewis
@danamlewis
Sep 28 2016 17:39
I am wondering if their 630G will enable them to scale faster than they would have without it
live4sw
@live4sw
Sep 28 2016 17:40
"While we’ve been working closely with the FDA on this product submission for the last several months, the approval did come much earlier than expected. Given this and the novel nature of the technology, we won’t be ready to ship the MiniMed 670G system until spring of 2017."
Dana Lewis
@danamlewis
Sep 28 2016 17:40
boo
live4sw
@live4sw
Sep 28 2016 17:42
I was kind of expecting the Spring 2017 release to drag a bit anyway, so I take this as slightly good news.
Dana Lewis
@danamlewis
Sep 28 2016 17:42
:+1:
Scott Leibrand
@scottleibrand
Sep 28 2016 17:42
now they're under huge pressure to launch ASAP, as they can't hide behind FDA holding them up
so I bet it'll be more likely to be early spring than late.
live4sw
@live4sw
Sep 28 2016 17:45
I agree. I almost wonder if it is partly intentional if they built a bunch of 630G inventory and want to get through that. The whole 630G launch made very little sense to me.
Dana Lewis
@danamlewis
Sep 28 2016 17:46
it’s the same physical pump body as the 670G, so I was assuming that would speed things up because they’ve already got it in production for 630
fiberfan
@fiberfan
Sep 28 2016 17:49
OpenAPS capable pump and a carelink stick will be headed my way by the end of the week!
Dana Lewis
@danamlewis
Sep 28 2016 17:49
woot @fiberfan!
Tim Street
@tim2000s
Sep 28 2016 17:54
I figure that it's just a firmware flip to turn a 630 or 640 into a 670. I suspect inventory is unlikely to be a problem. Amazing early release though! Glad I wrote the blog about ignoring the 630!
Also, Medtronic presentation at EASD talked about a meal assist type function incorporated into 670 algorithm.
Scott Leibrand
@scottleibrand
Sep 28 2016 17:58
:+1:
garykidd
@garykidd
Sep 28 2016 18:07
You think the community will be hack the 670 to make it work for openaps....I can't imagine MM freely opening the door to this possibility anytime soon
@scottleibrand no worries...do let me know if you may be able to test this evening..understand if you dont have time though...always appreciated. ANd let em know, if you'd like me to test anything with your script or otherwise--would love to be able to try and repay the favor.
garykidd
@garykidd
Sep 28 2016 18:24
@scottleibrand Was thinking about a tweak to oref algorithm that would compensate for initial jumpiness of sensor values seen immediately following sensor startup. Might make it less reactionary during this period.
Dana Lewis
@danamlewis
Sep 28 2016 18:25
have you tried soaking your sensors, yet?
Scott Leibrand
@scottleibrand
Sep 28 2016 18:26
@garykidd yes, that. also, what was your tweak?
garykidd
@garykidd
Sep 28 2016 18:30
Still thinking about it but.... on the surface was wondering what the impact would be if you modified the way the rate of change is calcualted to be based off of a period further back in history....maybe more of an average...
of course would still need a lot more analysis but it seems that often times you get an erroneous blip in the readings..that probably suggest the cgm was wrong at that time....but it seems oref0 makes an immediate determination based on the data still........
of course...it would seem there is an inherent trade-off.....an d a bit of delay in reposnsiveness...but maybe it would also mute reponses that may have been made on in accurate data.....I guess this presumes that the data proves out over time...which,i think it does
Scott Leibrand
@scottleibrand
Sep 28 2016 18:33
we already compare against an average of the ~15m ago and ~30m ago times. I think you're suggesting we also average the "right now" datapoint?
the way we do it now biases toward low-temping quickly on a drop, but only high-temping if the 5m delta, short avgdelta, and long avgdelta all agree that it's appropriate
(we compare against minDelta, the min() of all three)
garykidd
@garykidd
Sep 28 2016 18:40
and maybe I'm misunderstanding or just not understanding the inner working of oref0...but I had thought that you are comparing those different period -15 and -30 against the latest BG......in that instance maybe its not comparing against the avg. of say last 2 BGs.....but somehow delaying the enact to confirm what it had predicted---was accurate (within a tolerance) before it is enacted.
Scott Leibrand
@scottleibrand
Sep 28 2016 18:42
I find it's generally safer to stop insulin delivery, and then high-temp if it turns out that was unnecessary
a 5-10m delay on insulin delivery won't hurt anything, but you can't take back insulin if you give it
garykidd
@garykidd
Sep 28 2016 18:43
right---that makes total sense
so anyway--the idea was just that maybe a little bit of extra delay in high temping while using a new sensor...might make whatever gets enacted....a little more reliable. Again...just an idea
Sandra Keßler
@SandraK82
Sep 28 2016 18:48
lost link to the autoinstall script :/
garykidd
@garykidd
Sep 28 2016 18:49
I thought this was really good for me......
blob
Sandra Keßler
@SandraK82
Sep 28 2016 18:49
thank you @danamlewis
Dana Lewis
@danamlewis
Sep 28 2016 18:51
:+1:
@garykidd awesome!
garykidd
@garykidd
Sep 28 2016 18:52
@danamlewis thank you
Scott Leibrand
@scottleibrand
Sep 28 2016 18:52
@garykidd how exactly would you detect the new or jumpy sensor and change behavior, or how would you change the algorithm to get the smoothing without making it less responsible normally?
garykidd
@garykidd
Sep 28 2016 18:54
I was thinking that if you entered new sensor information on NS that that would trigger the different behavior...I wouldnt --I dont think---want it to try and self detect based on readings
and the idea would be that it would be that it wold be more responsive after a defined period of time...maybe a full switch-over at a certain point...or maybe a gradual switch.
Kelly Watson
@kellyisworking
Sep 28 2016 19:05
Out of curiosity - does anyone know if enlite 3 is coming out same time as 670? No use of dex w/ 670 correct?
Dana Lewis
@danamlewis
Sep 28 2016 19:06
correct. Enlite 3 is what will be used in 670G
Kelly Watson
@kellyisworking
Sep 28 2016 19:07
Gotcha
Martin Haeberli
@mhaeberli
Sep 28 2016 19:10
looks like an article of general interest to this group : https://agilemeddev.com/2016/09/25/medical-devices-software-testing-overview/
Scott Leibrand
@scottleibrand
Sep 28 2016 19:11
@garykidd so you would need to add some new mechanism for entering when a sensor is really new (vs. restarted). What benefit would the user get for that additional work?
David
@davidkeddydb
Sep 28 2016 19:20
Screenshot 2016-09-28 15.20.03.png
@danamlewis @scottleibrand Thanks. Scott, I don't have any purple prediction lines. I only have blue lines from the Ar2 I think
garykidd
@garykidd
Sep 28 2016 19:20
@scottleibrand good point new vs. restarted...(I'm sure there are a lot of other things to think through obviously)...I think this part though may be easy enough with the addition of the appropriate field/s in careportal....new sensor...and restart sensor......
Dana Lewis
@danamlewis
Sep 28 2016 19:21
you have to turn them on :) @davidkeddydb and make sure you’re passing the right stuff
David
@davidkeddydb
Sep 28 2016 19:22
I gather that is just in the NS variables? I somehow missed turning them on
garykidd
@garykidd
Sep 28 2016 19:25
@scottleibrand the idea wol dbe that the benefit would be smoother, more reliable response albeit less responsive responses...but again...this is more conversational/conceptual.
benhamou
@benhamou
Sep 28 2016 19:31
This message was deleted
Tim Street
@tim2000s
Sep 28 2016 19:41
So anyone else notice the 670 has an unmovable target glucose when looping of 120 mg/dl? Feels high to me.
Scott Leibrand
@scottleibrand
Sep 28 2016 19:48
@tim2000s that's what has been reported, yes. seems the problem is not so much the target as the fact that the pump often fails to get BG down to target. we'll have to try it out and see if there are any tricks that work well to counteract that. I've heard of lots of people entering fake carbs to make it be a bit more aggressive...
Sandra Keßler
@SandraK82
Sep 28 2016 19:49
i need some assistance please, i have bought a 522 WW serial, i have a ti stick flashed with ww subg_rfspy firmware (make -f Makefile.usb_ep0 RADIO_LOCALE=WW install)
how can i verpfiff that it works
C-Ville
@C-Ville
Sep 28 2016 19:49
@scottleibrand faking cgm data could help to cheat on the target :-)
Sandra Keßler
@SandraK82
Sep 28 2016 19:50
used setupscript and did choose mmeowlink (installed) and gave serial
Scott Leibrand
@scottleibrand
Sep 28 2016 19:50
@garykidd yeah, reducing temp basal variation has been a goal of mine, but not at the expense of reduced responsiveness when needed. there are a few other things I need to do around AMA to reduce oscillation. but at the end of the day, bursty insulin delivery doesn't really reduce the effectiveness of the loop, so it's a lower priority than some other things, and not worth adding additional usability burden
Sandra Keßler
@SandraK82
Sep 28 2016 19:52
mmtune gives 0,-99 for each frequency
Scott Leibrand
@scottleibrand
Sep 28 2016 19:52
is it scanning US or WW freqs?
C-Ville
@C-Ville
Sep 28 2016 19:53
(installed) means that you installed mmeowlink seperately?
Sandra Keßler
@SandraK82
Sep 28 2016 19:56
ah no i did not
oki how can i verify that it works
C-Ville
@C-Ville
Sep 28 2016 20:01
openaps use pump mmtune
Matthias
@ceben80
Sep 28 2016 20:01
This message was deleted
C-Ville
@C-Ville
Sep 28 2016 20:02
does someone know if the animas ping can do bolus calculation wothout use of the handheld?
without
Mitchell Slep
@moomoobloo
Sep 28 2016 20:02
@C-Ville Yes, you can do everything on the pump. The remote only supports a subset of the functionality of the pump.
C-Ville
@C-Ville
Sep 28 2016 20:05
@moomoobloo thanks so the Ping remote does not hold a diary of BG and Carbs and Dose?
Sandra Keßler
@SandraK82
Sep 28 2016 20:05
Mmtune gives me some frequency
garykidd
@garykidd
Sep 28 2016 20:06
@scottleibrand keep in mind it would only be enacted during periods of initial jumpiness and then fall away (in an ideal world). and please do understand that I am only conceptualizing and not actually suggesting that you work to incorporate this or anything like it into your awesome work---which it trul is (and thank you, sincerely) ---just theorizing more or less. but thinking a bit more about this...if you did anything that could smooth basal delivery---no matter what the course---I think a lot of approaches may very well lead to less repressiveness...the question may be is the trade-off worth it. In any case, maybe a simple user toggle may be warranted (especialyl in the case where a trade-off is not avoidable) for a feature like this.
C-Ville
@C-Ville
Sep 28 2016 20:06
in the 800 oder 900 mhz range?
Sandra Keßler
@SandraK82
Sep 28 2016 20:06
800
C-Ville
@C-Ville
Sep 28 2016 20:06
868 EP
:-)
gogo @SandraK82 you are almost there
Scott Leibrand
@scottleibrand
Sep 28 2016 20:08
@garykidd I really appreciate people other than me thinking about ways to improve the algorithm. please keep it up. :)
Russell Gardner
@RussGG
Sep 28 2016 21:14
I'm still going to finish my openAPS!!!!!
Dana Lewis
@danamlewis
Sep 28 2016 21:14
670G won’t ship til Spring, so not surprised to hear that from lots of folks :)
C-Ville
@C-Ville
Sep 28 2016 21:16
@danamlewis lucky people in US, in germany it will take even longer
Dana Lewis
@danamlewis
Sep 28 2016 21:17
@C-Ville we still have a lot of work to do to improve access, cost, etc. worldwide for sure!
Tim Street
@tim2000s
Sep 28 2016 21:25
That's really the biggest issue. While the pump will likely be available in the U.K. in 2017, to get supported APS function will be stuck till 2020.
C-Ville
@C-Ville
Sep 28 2016 21:29
in Germany the Health Insurances rarely pay for CGM. For Libre they seem less reluctant
Scott Leibrand
@scottleibrand
Sep 28 2016 21:57
hopefully that'll be easier with APS, where the benefits have been demonstrated in lots of studies
R. Dennis Steed
@rdsteed
Sep 28 2016 21:59
On another FDA note, the Freestyle Libre is making its first foray in the US market with the Libre Pro, a version of the Libre that can only be read by taking it back to the Doctor's office. Wonder if @SandraK82 's BlueReader will be able to work with these US Libres.
C-Ville
@C-Ville
Sep 28 2016 22:00
for now we need to push the bluereader from @SandraK82 over the goal, so people can loop with the libre https://www.startnext.com/en/bluereader :-)
R. Dennis Steed
@rdsteed
Sep 28 2016 22:06
@C-Ville I made a contribution to the project. I think BlueReader is a great idea, even if it isn't something that is applicable for the US market yet.
C-Ville
@C-Ville
Sep 28 2016 22:09
@rdsteed cool, is libre not so common in US yet? in Germany it seems that Health insurances use that as a cheap excuse for not having to pay for a real CGM
R. Dennis Steed
@rdsteed
Sep 28 2016 22:16
@C-Ville In the US Libre is unavailable. The FDA only approved the "Libre Pro" for the US this week. Unfortunately, that device is applied at the doctor's office and can only be read when the device is returned after collecting readings. No readings are available in real time to the person with diabetes. The consumer version, like the one available in Europe, has not yet been approved by the FDA.
C-Ville
@C-Ville
Sep 28 2016 22:25
okay, I was not aware that the libre is not approved in US
hmm.. looks like might be similar to nutshell?
garykidd
@garykidd
Sep 28 2016 23:57
can someone direct me to how/what to enable to reflect the purple predictive BG in NS...much appreciated.
Scott Leibrand
@scottleibrand
Sep 28 2016 23:59
are you running oref0 dev with meal assist turned on?