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

31st
Aug 2015
Scott Leibrand
@scottleibrand
Aug 31 2015 01:19
@tghoward That is a safety feature. If BG is dropping, you want to wait and see if BG drops more than predicted rather than ending up with a whole bunch of IOB and dropping too fast.
if BG starts to tick back up, it will issue another high temp, but as long as you're dropping, it goes into "wait and see" mode.
In the inverse situation, however, when BG and eventual BG are low, the safer thing to do is to let any low-temps run until eventual BG is in range.
Scott Leibrand
@scottleibrand
Aug 31 2015 02:11
@bewest @loudnate just discovered something weird: openaps use pump read_settings is reporting maxBasal in whole number units/hr, rounded down. if we set it to 2.9U/hr on the pump, it reads 2.
is that something in openaps/decocare, or is it actually reading from the pump that way?
Nathan Racklyeft
@loudnate
Aug 31 2015 02:17
@scottleibrand interesting, I haven't noticed that in the X23 settings, but I know they have different implementations. Mine currently reads "maxBasal": 3.5,
might just be a programming error,
are you on decocare release or dev?
oh I guess there was a release cut 11 days ago. Are you running 0.0.17?
Scott Leibrand
@scottleibrand
Aug 31 2015 02:19
I'll try updating and see if that fixes it
Nathan Racklyeft
@loudnate
Aug 31 2015 02:19
actually, it should :)
Dana Lewis
@danamlewis
Aug 31 2015 02:20
:smile: thanks @loudnate
Nathan Racklyeft
@loudnate
Aug 31 2015 02:20
it was strokes / 40 instead of strokes / 40.0
Scott Leibrand
@scottleibrand
Aug 31 2015 02:21
perfect, reading 3.5 now. :)
Nathan Racklyeft
@loudnate
Aug 31 2015 02:21
hey! good setting.
Scott Leibrand
@scottleibrand
Aug 31 2015 02:21
:)
Nathan Racklyeft
@loudnate
Aug 31 2015 02:22
you'll have to re-round by stroke size when dosing
Scott Leibrand
@scottleibrand
Aug 31 2015 02:23
?
Nathan Racklyeft
@loudnate
Aug 31 2015 02:23
but even then the pump will sometimes program a slightly lower basal than you ask for, based on the length of the basal
Scott Leibrand
@scottleibrand
Aug 31 2015 02:23
I've always noticed the pump rounds my requests to the nearest eighth or something
Nathan Racklyeft
@loudnate
Aug 31 2015 02:23
round(min(max_basal, rate) * strokes) / strokes
right, that's because you set 30 min basals?
Scott Leibrand
@scottleibrand
Aug 31 2015 02:23
yeah
doesn't seem to be an issue
if the current is within 0.1U/hr of requested, I don't try to "correct" it
Nathan Racklyeft
@loudnate
Aug 31 2015 02:24
yeah, there's not really a way around it anyway
at least the history is accurate when pulled :)
Scott Leibrand
@scottleibrand
Aug 31 2015 02:26
yeah, and currenttemp.json is accurate too
Nathan Racklyeft
@loudnate
Aug 31 2015 02:26
right
welcome back, by the way!
Scott Leibrand
@scottleibrand
Aug 31 2015 02:28
thanks! :)
so since we got back, we got one more person up and looping: @esteward :)
Chris Hannemann
@channemann
Aug 31 2015 02:30
:clap:
Any estimate on how many folks are looping now?
Dana Lewis
@danamlewis
Aug 31 2015 02:33
4+ regularly. Probably a few more who are testing and haven't shouted from the rooftops yet :)
Chris Hannemann
@channemann
Aug 31 2015 02:35
Nice
Might be worth a roll-call to see who is there and who is on the verge but needs a little push to get over the hump
Dana Lewis
@danamlewis
Aug 31 2015 02:37
Yea, we just had a few folks join openAPS-dev so plan to do that this week so people would intro and also update or input to the spreadsheet to re-inspire anyone who's stuck
Chris Hannemann
@channemann
Aug 31 2015 02:39
I think that's an excellent idea
diabeticgonewild
@diabeticgonewild
Aug 31 2015 03:00
Yeah I haven't been motivated to do much. Just want to go back to school and feel better. It's really hard being me. Seeing my GI doctor on Thursday for GI pain (also calling them tomorrow morning) and have intravenous immunoglobulin on Wednesday. Frankly sick of being sick....
Scott Leibrand
@scottleibrand
Aug 31 2015 04:35
@amazaheri I think I'm going to update getprofile.js to read max_iob from a config file, and default to 0 if it's not present. that way if anyone installs openaps-js and hasn't yet read through the code to see how the algorithm works, it will operate in "safe mode" and low-temp as needed, but only high-temp to bring basal IOB back up to zero (if you're rebounding after being low-temped).
I think that's the only per-user variable that's in the code and not read directly from the pump
Ali Mazaheri
@amazaheri
Aug 31 2015 04:43
go for it.
Ali Mazaheri
@amazaheri
Aug 31 2015 04:48
this makes it with no hard coded value %100 which is always good :)
Scott Leibrand
@scottleibrand
Aug 31 2015 04:48
yeah. :)
Ali Mazaheri
@amazaheri
Aug 31 2015 04:49
we have come a long way since June when we first started :)
Scott Leibrand
@scottleibrand
Aug 31 2015 04:52
also thinking about looking at @TC2013's roundtrip code and seeing if we can replace the algorithm with a direct function call to determine-basal.js
Ali Mazaheri
@amazaheri
Aug 31 2015 04:52
I am hoping to start seeing some PR from folks using it already to get it more polished day by day.
Scott Leibrand
@scottleibrand
Aug 31 2015 04:52
probably more complicated than that unfortunately
Ali Mazaheri
@amazaheri
Aug 31 2015 04:52
i am working on his app to port to Windows using our bridge toolkit
but main priority is to have the Intel Compute Stick fully functional with BT
for a 24/7 run
Scott Leibrand
@scottleibrand
Aug 31 2015 04:53
I would love to be able to have a single codebase for the core algorithm that we can update once and will run on rPi, windows, android, etc.
Ali Mazaheri
@amazaheri
Aug 31 2015 04:53
Amen to that
specially with IOB and Determine_basal alot of important stuff happening
and helps us with data analysis regardless of device being used
Scott Leibrand
@scottleibrand
Aug 31 2015 04:55
@esteward mentioned that he has now tried both the roundtrip and openaps-js algorithms, and feels the openaps-js one is doing the right thing for him far more often. but he (and we) would much prefer to be able to use the roundtrip UI with android when mobile, using the determine-basal.js algorithm.
Ali Mazaheri
@amazaheri
Aug 31 2015 04:55
I see a merge effort
Scott Leibrand
@scottleibrand
Aug 31 2015 04:55
yeah, the other big thing we need to do is standardize on a way to upload data to funnel cake using the nightscout api
yeah, sorta merge, sorta "make the algorithm portable"
anyway, lots to do, but I think I'm at a stopping place for tonight. :)
Ali Mazaheri
@amazaheri
Aug 31 2015 04:57
so @jasoncalabrese documenting it with Swagger and I have an API app in order. This way we have a end to end orchestration for updates and notification, its working fine and code free and reliable
go rest
Scott Leibrand
@scottleibrand
Aug 31 2015 04:57
looking forward to using that
Ali Mazaheri
@amazaheri
Aug 31 2015 04:58
almost there I will setup a time to set it up for u, @oschumac is setup with that as well
now to give @bewest a hard time, Ben where is my RileyLink lol
Scott Leibrand
@scottleibrand
Aug 31 2015 05:00
cool
yeah, we need to schedule some working sessions
Ali Mazaheri
@amazaheri
Aug 31 2015 05:00
i am game let me know when
Scott Leibrand
@scottleibrand
Aug 31 2015 05:00
compare notes and get each other set up with all the best stuff each of us are running. :)
Ali Mazaheri
@amazaheri
Aug 31 2015 05:01
have an easy one this week and available whenever u can
Scott Leibrand
@scottleibrand
Aug 31 2015 05:01
we're meeting up with @esteward tomorrow...
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:01
Google Hangouts on Air works really well.
I'm just saying...
Ali Mazaheri
@amazaheri
Aug 31 2015 05:02
ok let me know when u r ready , BTW @sholehf has access to my data and is warming up with ML
Agree @diabeticgonewild no one like skype 4 business lol
Dana Lewis
@danamlewis
Aug 31 2015 05:04
@amazaheri Can't wait to see what @SholehF is up to :)
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:04
@amazaheri @scottleibrand, not sure if you saw in the other room, but we're planning to follow up on the release next week, with another release 2 or 3 weeks later
Scott Leibrand
@scottleibrand
Aug 31 2015 05:05
@amazaheri we want to get a few things done and the blogs updated, so let's shoot for later in the week
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:05
Dana Lewis
@danamlewis
Aug 31 2015 05:05
What's in the second release @jasoncalabrese ?
Scott Leibrand
@scottleibrand
Aug 31 2015 05:05
@jasoncalabrese cool. no, not keeping up on the public room
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:06
more from @MilosKozak
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:06
I can set up the meeting no problems. You can meet on your mobile phone even if you want.
Scott Leibrand
@scottleibrand
Aug 31 2015 05:06
once we get the API bit figured out, I also need to see about getting my wip/iob-cob temp basal display stuff merged/ported over to funnel cake
Ali Mazaheri
@amazaheri
Aug 31 2015 05:06
deal!
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:06
he ported all of the reports from the chrome tools
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:06
It just drains battery.
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:06
and added some more
Dana Lewis
@danamlewis
Aug 31 2015 05:07
@jasoncalabrese cool. @scottleibrand Or to Gluten free Griddlecakes depending on how long it takes
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:07
so if we need api changes to work better with openaps, that might be a good time to do it
Ali Mazaheri
@amazaheri
Aug 31 2015 05:07
YES!
Scott Leibrand
@scottleibrand
Aug 31 2015 05:07
yeah, we should think about planning some hangouts with more people from the openaps-dev community too, once we have our updates posted etc.
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:07
would like to get temp basals modeled some way
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:08
As long as everyone has a pair of headphones/earbuds we can have a good live meeting.
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:08
something you'd create from the pump history
Dana Lewis
@danamlewis
Aug 31 2015 05:08
Probably after Labor Day for a bigger sessiOn w/ folks
Scott Leibrand
@scottleibrand
Aug 31 2015 05:08
the important thing I think is to scope topics for whatever meetings we do schedule
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:08
I will try to get my stuff commented tomorrow.zz
Yeah cause we need to value people's time.
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:09
anyone going to CWD besides @amazaheri and @bewest?
Scott Leibrand
@scottleibrand
Aug 31 2015 05:09
@jasoncalabrese you doing anything with pump history events for detecting insertion set change?
Dana Lewis
@danamlewis
Aug 31 2015 05:09
@jasoncalabrese Yea temp modeling would be good. I like @amazaheri's view for now but not sure how it would work for bigger increments
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:09
It needs to stay on task too. Take it from someone who has ADHD (did I just admit to that?! Hahahaha)
Scott Leibrand
@scottleibrand
Aug 31 2015 05:09
@danamlewis pointed out that would be excellent for resetting CAGE automatically
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:10
I've been waiting on the RileyLink
Scott Leibrand
@scottleibrand
Aug 31 2015 05:10
seems a lot of people are. does anyone have the executive summary of where we are there?
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:10
would like to skip the Rpi and carelink
Ali Mazaheri
@amazaheri
Aug 31 2015 05:10
mine is a hack for now, had to remove IOB so @TC2013 wont give me a hard time on twitter ;)
Scott Leibrand
@scottleibrand
Aug 31 2015 05:10
I've been watching ps2/rileylink, and still see lots of "what does this byte do" kinds of stuff
Dana Lewis
@danamlewis
Aug 31 2015 05:11
@jasoncalabrese no to CWD plans, it's pretty expensive unless we at some point make specific plans for doing stuff there.
Scott Leibrand
@scottleibrand
Aug 31 2015 05:11
so not sure how close we are to having rileylink up and running as a carelink replacement
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:11
would be nice to keep using js
doing everything in java doesn't seem fun
Dana Lewis
@danamlewis
Aug 31 2015 05:11
They used to have CWD focus on tech in seattle sometimes, maybe they'll do that here next year and give us a break
@amazaheri Everything's a hack ;)
Ali Mazaheri
@amazaheri
Aug 31 2015 05:12
lol
Scott Leibrand
@scottleibrand
Aug 31 2015 05:12
yeah, I haven't looked into what it takes to call javascript from android
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:14
Rpi without carelink isn't so bad
think I could make that work with school
an Rpi, phone and RL might fit in the small backpack
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:15
Flipbelt might handle all of that, IDK about comfortably...
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:16
so I guess the main issue for me is the range
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:16
and CGM...
Indeed
Scott Leibrand
@scottleibrand
Aug 31 2015 05:17
Heh, we already discovered flipbelt. ;-)
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:17
brought an RL and 522 to endo appt just for fun, to see the reaction
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:18
Probably a long time ago
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:18
wasn't any concearn
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:18
How was that? (mine was cool about it)
My endo said she adored me!! <3
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:18
gave some different mdt pump sites
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:18
That's badass
Jason Calabrese
@jasoncalabrese
Aug 31 2015 05:18
would have been different last year I'm sure
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:19
Hahahaha!
Paul Andrel
@stavlor
Aug 31 2015 05:19
heh
Pete Schwamb
@ps2
Aug 31 2015 05:22
RL can currently be used as a CGM data uploader (to NS) for Enlite users on iOS. Work is ongoing for using it to dump full pump history, and to use it in an openaps setup, and for Android use.
(my attempt at an executive summary)
Scott Leibrand
@scottleibrand
Aug 31 2015 05:23
What other data is in those mysentry packets? Does it have insulin remaining in reservoir?
excellent summary thx. :-)
Pete Schwamb
@ps2
Aug 31 2015 05:23
Yes
Scott Leibrand
@scottleibrand
Aug 31 2015 05:24
so we could keep track of insulin delivery completely passively right?
That should be all the info we need to calculate IOB (if we've been running for DIA).
Pete Schwamb
@ps2
Aug 31 2015 05:25
It also has IOB (as calculated by the pump).
reported every 5 min
Scott Leibrand
@scottleibrand
Aug 31 2015 05:26
So we can determine-basal and then only engage Carelink when we determine that we need to administer a basal.
Ah good.
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:26
Nice!
Scott Leibrand
@scottleibrand
Aug 31 2015 05:26
That would let us bootstrap if we don't have reservoir history.
If we are just starting up the loop, pump IOB should = net IOB.
Carelink also has much better range for small stuff like administering temps, vs. pulling a full pump history.
Pete Schwamb
@ps2
Aug 31 2015 05:29
So this insulin remaining is from the mysentry packets; I think many of the older pumps (those with temp basal) do not support this style of comms, unfortunately.
Scott Leibrand
@scottleibrand
Aug 31 2015 05:29
And we could more aggressively retry enacting the temp
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:29
yeah true. I already made equations for the 522 pumps though and you can use that. They're polynomials. I can always redo it and use piecewise hermetian cubic polynomials so the derivative of that works out well so "activity" turns out right.
Scott Leibrand
@scottleibrand
Aug 31 2015 05:29
Ah, bummer. Yeah, I think we did look into that once upon a time and found the 522 doesn't do mysentry. :-(
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:30
Yeah it doesn't
Pete Schwamb
@ps2
Aug 31 2015 05:30
However, I see ReadRemainingInsulin commands in decocare. do those also not work on 522?
Scott Leibrand
@scottleibrand
Aug 31 2015 05:30
Bummer.
That should be fine.
diabeticgonewild
@diabeticgonewild
Aug 31 2015 05:31
I need to rework those equations, whenever I feel up to t.
Scott Leibrand
@scottleibrand
Aug 31 2015 05:31
I think it's just the broadcasts that the older pumps don't do.
Pete Schwamb
@ps2
Aug 31 2015 05:31
If the ReadRemainingInsulin command works on 522, then the RL can poll for that.
Scott Leibrand
@scottleibrand
Aug 31 2015 05:32
Are you suggesting we just ReadRemainingInsulin frequently, and only read full pump history occasionally?
that might help
Pete Schwamb
@ps2
Aug 31 2015 05:32
Yeah, something like that.
Scott Leibrand
@scottleibrand
Aug 31 2015 05:32
we can implement that with Carelink first too.
Pete Schwamb
@ps2
Aug 31 2015 05:32
yep
iirc, there might be something about the resolution of the ReadRemainingInsulin command?
Scott Leibrand
@scottleibrand
Aug 31 2015 05:34
I think so, but IIRC it was good enough for my purposes.
Looks like it's in 0.1U increments.
Which is perfect for me.
Scott Leibrand
@scottleibrand
Aug 31 2015 05:43
Biggest challenge is that to do this we'd need to switch openaps from mostly stateless to stateful operation.
Ben West
@bewest
Aug 31 2015 07:08
the reports are stateful
so is session ;-)
still catching up, but recommend the seeming conclusion of using readresvoir to get high-fidelity understanding of doses given
@scottleibrand the key to tracking state I'm recommending for now is more interstitial reports, if possible
Tim Howard
@tghoward
Aug 31 2015 11:26
@scottleibrand Thanks for the reply and explanation. That makes sense but I may need to be a bit more aggressive -- when we are 100 or 200 points too high and start falling just a bit I still want the high basal on. I'll add a check for that kind of discrepancy.
Ali Mazaheri
@amazaheri
Aug 31 2015 13:11
in those situations you probably better off with a manual bolus and let APS to control post correction.
Tim Howard
@tghoward
Aug 31 2015 13:22
yes, good point. Thanks.
Toby Canning
@TC2013
Aug 31 2015 13:43
We have varied a bit from the reference design and let the APS manage the whole correction. For us, it works very well.
However, I would like to built in a feature to remote bolus
Or rather, entirely control the APS remotely
Oliver Schumacher
@oschumac
Aug 31 2015 14:58
@jasoncalabrese Hi as i understood the cgm-remote-monitor wip/iob-cob branch can visu temp basals?
Jason Calabrese
@jasoncalabrese
Aug 31 2015 14:59
@TC2013 yes, thats what I'd like, remote bolus
@oschumac, @scottleibrand added something to his iob-cob branch, but that never made it back to the main repo. Also the iob-cob branch is really behind. I thought we'd have everything ported over now, but with openaps ignoring carbs I'm not sure it makes sense
Oliver Schumacher
@oschumac
Aug 31 2015 15:05
I used the regular branch sending treatments from pi. Also worked for me but maybe there are better solutions. I love to get a propper visu for openaps.
Chris Hannemann
@channemann
Aug 31 2015 15:11
@jasoncalabrese the version of openaps that @loudnate and I are running uses carbs, so having the IOB-COB branch brought in would still be worth it
"worth it" being relative to other features requiring resources, of course; if there isn't great demand, might be worth shelving
Dana Lewis
@danamlewis
Aug 31 2015 15:40
When I am not running APS, and just using DIYPS, the visualization is still used, so carbs are still good. Scott and I were talking about bringing iob-cob back up to date in the next few weeks.
Ali Mazaheri
@amazaheri
Aug 31 2015 15:50
As @scottleibrand and I were discussing last night we need to work on a centralized repo for anything OpenAPS and apply changes centrally. Currently we have so many variations and it makes it hard to plan design and implement plug ins for things like NS C-R-M, reporting, etc. Also one of our goals is to eventually use the data for some predictive analysis and using a shared repo would be beneficial.
Scott Leibrand
@scottleibrand
Aug 31 2015 17:01
@tghoward Yeah, in that situation we often do what @amazaheri said and just do a half unit bolus to start bringing it down. But one non-obvious thing is that if you're actually running openaps, it will be high-temping you the whole way up to 200, so you'll have quite a bit of IOB by that point, and you're unlikely to just ride high and continually keep needing more. That said, I have considered adding some sort of logic to not kill the high-temp until the eventual prediction is within 40 mg/dL of the high target, or something like that. But it seems like most of the cases where that would be necessary, there's something else going on (like yesterday it was a bad infusion site) that it might be better to just let the CGMs alarm and make the human decide what to do.
Chris Hannemann
@channemann
Aug 31 2015 17:05
@scottleibrand how do you calculate the trend for determining whether or not to administer a high temp? Is it simply dBG over the last two points (5 minutes), or are you smoothing?
Or using the trend arrow, for that matter.
Scott Leibrand
@scottleibrand
Aug 31 2015 17:06
at the moment it's just the 5m tick. I'm thinking of maybe switching to 10m (or reading raw as well) to avoid some of the effects of dexcom filtering, but haven't decided it needs to change just yet.
the other thing we do is take the 15m change into effect when calculating the eventual BG
we calculate the 15m delta expected from the current BGI (IOB activity), and compare that to the actual 15m change
and adjust the eventualBG by the discrepancy
Chris Hannemann
@channemann
Aug 31 2015 17:07
Do you see chatter when BG is high and you would temp basal but readings are jumping a bit?
Scott Leibrand
@scottleibrand
Aug 31 2015 17:08
so basically, we're predicting the current deviation from BGI will continue for 15m
which does a good job of picking up carb rises and other unexpected excursions
yeah, we do see that some. but like I mentioned, once reading start fluctuating +-, we've usually run high-temps long enough that there's enough IOB and it starts coming back down soon
Chris Hannemann
@channemann
Aug 31 2015 17:09
True, and the net effect is minimal.
Scott Leibrand
@scottleibrand
Aug 31 2015 17:09
if we don't yet have enough IOB, we'll end up running a high-temp for 10m, canceling it for 5, running it for another 5, etc.
and there's no real harm in doing that
Chris Hannemann
@channemann
Aug 31 2015 17:09
Precisely.
No harm either way, really. We don't have any check on BG direction for high temping, for example, and haven't seen problems.
But I do understand why you like to have that in place.
Scott Leibrand
@scottleibrand
Aug 31 2015 17:30
We've seen lots of cases where BG drops below 100 when the prediction was over 140 at the peak. That may not be common for everyone, but I definitely wouldn't want to disable that safety feature on any code that's being used by anyone other than the authors.
Tim Howard
@tghoward
Aug 31 2015 17:46
Do you have any problems with the loop taking more than 5 minutes to run? I've actually backed up to ten minutes in crontab because of this and sometimes it doesn't even finish within this time and so the loop is run after 20.
Scott Leibrand
@scottleibrand
Aug 31 2015 17:52
I have my loop running every 3m, but with a lock file so it bails if previous run is going long.
See openaps-js/bin/loop.sh
almost always finishes within 5m though.
Chris Hannemann
@channemann
Aug 31 2015 17:56
Every five here, takes about a minute to complete
Jason Calabrese
@jasoncalabrese
Aug 31 2015 17:57
@scottleibrand, I'd be happy to give you a tour of the new NS plugins, think we can add what you need in a clean way now, just a few things to figure out
Tim Howard
@tghoward
Aug 31 2015 17:57
Yes, it bails if already running, so no problems with taking longer, I was just expecting it to finish more quickly. Good to hear from both of you it should be shorter.
@scottleibrand Yes, I'm used that (loop.sh) as a starting point and have filled in the gaps as best as possible.
Oliver Schumacher
@oschumac
Aug 31 2015 18:28
My try outs on openaps giving me 30s runtime of the script due to don't grap every time the configuration of the pump. I am almost using Scotts script. I did a flow chart of the funktion i have done. See atached file.
If we find a implementation everbody agree to i'll put this to the docs.
Tim Howard
@tghoward
Aug 31 2015 18:33
NIce, clean flow charts!
Dana Lewis
@danamlewis
Aug 31 2015 19:03
Need to check and see how this compares to ours, but love laying it out like this @oschumac !
Oliver Schumacher
@oschumac
Aug 31 2015 19:18
Thanks, if there are some ideas i'll rework the charts.
Ben West
@bewest
Aug 31 2015 19:19
I think rather than erasing json files
some people are using some checks to ensure that the data being worked with is recent
Scott Leibrand
@scottleibrand
Aug 31 2015 19:19
yeah, that's what we do in openaps-js
Ben West
@bewest
Aug 31 2015 19:20
and that step actually could reset the recommendations/interstitial values
Scott Leibrand
@scottleibrand
Aug 31 2015 19:20
makes it much more reliable when you're at the edge of range and some pump communications are failing
I do erase the recommendedtemp.json every run
Ben West
@bewest
Aug 31 2015 19:20
right
there's an inconsistency with return 0, return 1
in linux, the normal thing is to return 0 for no errors
and something else 1, 255 or something in between for errors
Scott Leibrand
@scottleibrand
Aug 31 2015 19:22
@tghoward interested what gaps you've identified. loop.sh should work if you have all the required reports defined correctly in your openaps.ini. interested to hear your experience so far.
Ben West
@bewest
Aug 31 2015 19:23
there's a nifty tool some of us use to maintain/share these kinds of diagrams, might come in handy, or not, http://blockdiag.com/en/seqdiag/demo.html, http://blockdiag.com/en/actdiag/demo.html
Scott Leibrand
@scottleibrand
Aug 31 2015 19:23
@jasoncalabrese definitely. probably after I do a working session with @amazaheri and we figure out what we want to do for data upload.
Ali Mazaheri
@amazaheri
Aug 31 2015 19:27
looks good @oschumac one minor but important correction, we need to get the profile info and pass it for IOB calculation
Oliver Schumacher
@oschumac
Aug 31 2015 19:29
@amazaheri you mean profile.json ?
Ali Mazaheri
@amazaheri
Aug 31 2015 19:30
right
Oliver Schumacher
@oschumac
Aug 31 2015 19:30
oh looks like wrong sequence in the doku.
Tim Howard
@tghoward
Aug 31 2015 19:31
@scottleibrand Sorry that wasn't really the right wording - apologize that it came out wrong. I don't have it in front of me but some of the calls in loop.sh were (are?) for alias commands to openaps and I had to stumble through it to figure out what they were (at least I think that's what they are for ... I'm not sure). What you've provided at openaps-js is truly fantastic and has been so helpful and I really appreciate it getting put out there.
Oliver Schumacher
@oschumac
Aug 31 2015 19:32
fixed
Ali Mazaheri
@amazaheri
Aug 31 2015 19:32
:+1:
your webapi on azure works fine?
Scott Leibrand
@scottleibrand
Aug 31 2015 19:33
@tghoward yeah, the best starting place so far is the openaps.ini in the openaps-example repo, but that's not complete
Ben West
@bewest
Aug 31 2015 19:33
@channemann has a great write up
Scott Leibrand
@scottleibrand
Aug 31 2015 19:33
not sure if I've seen that: have a link?
Ben West
@bewest
Aug 31 2015 19:33
@channemann I think it'd be interesting to turn that write-up into an interactive worksheet
I'm pressuring him to share it now ;-)
Scott Leibrand
@scottleibrand
Aug 31 2015 19:34
ah
Ali Mazaheri
@amazaheri
Aug 31 2015 19:34
SHARING IS CARING GUYS! :)
Ben West
@bewest
Aug 31 2015 19:34
something where you can type the preferred name in a form at the top, maybe
and then if fills in the names so it's all right
we can eliminate the <foobar> typographical confusions
Rachel Sandlain
@audiefile
Aug 31 2015 19:38
interactive worksheets are great. Ages ago one was created for setting up nightscout and it helped with support a lot. Helps users organize and gives them something to refer to
Oliver Schumacher
@oschumac
Aug 31 2015 19:38
@amazaheri Yes upload is working fine. But i did not work with the Power BI yet. I have copied the sendAzure script. To send APS outout to Nightscout as Treatments.
Chris Hannemann
@channemann
Aug 31 2015 19:45
Thanks for the pressure, @bewest. It's a private gist at the moment; I'll get the link and post it here. It was essentially me trying to make sure and document all the steps I went through with @loudnate and @bewest to get up and running.
I'm going to do a fresh install following it, perhaps tomorrow, to see what's missing.
Tim Howard
@tghoward
Aug 31 2015 19:45
this might have been the alias: "openaps pumpquery" that I had to turn into five report requests:
openaps report invoke pump_settings.json.new
openaps report invoke bg_targets.json.new
openaps report invoke isf.json.new
openaps report invoke current_basal_profile.json.new
openaps report invoke carb_ratio.json.new
because I don't know how to link them into one alias. This string of five report requests is also what sometimes slows to over 5 minutes!
FYI, this doesn't use openaps-js at all
Dana Lewis
@danamlewis
Aug 31 2015 19:47
Love all the write-ups and docs, we have a couple of people who didn't have a ton of deep coding knowledge but can use these to help us kick tires more. Thx @channemann
Chris Hannemann
@channemann
Aug 31 2015 19:47
(Skip down to the plugins section for the interesting stuff)
Scott Leibrand
@scottleibrand
Aug 31 2015 19:47
yeah, we need to merge this approach with openaps-js
Tim Howard
@tghoward
Aug 31 2015 19:48
@channemann very nice. Thanks!
Scott Leibrand
@scottleibrand
Aug 31 2015 19:48
but it does give you a clue for openaps pumpquery. add an alias like chris did:
$ openaps alias add predict "report invoke predict_glucose.json predict_glucose_without_future_basal.json"
Chris Hannemann
@channemann
Aug 31 2015 19:49
Yes and yes, @scottleibrand and @danamlewis. At that point, I'd like to merge the stuff from that gist into the docs.
Scott Leibrand
@scottleibrand
Aug 31 2015 19:49
to be clear, these instructions are using @loudnate's stuff, not openaps-js. we should merge the two into a common framework that supports both.
Chris Hannemann
@channemann
Aug 31 2015 19:49
But agreed we need to look at -js and these plugins and see what we can do to align
Tim Howard
@tghoward
Aug 31 2015 19:50
Yes, understood from here.
Dana Lewis
@danamlewis
Aug 31 2015 19:50
Ditto. I also want to add some language about testing & proceeding with caution, esp. because all of these docs will make it a LOT easier to get started. Also see last night where @scottleibrand & I were talking about some safety features like net IOB only being 0 as default, to help the next round of folks who get started
Chris Hannemann
@channemann
Aug 31 2015 19:52
Yeah, I saw that and though it was a great idea. I think you could extend that methodology to a few other aspects as well, such as whether to temp high with a falling BG
Default to no, but enable the "informed" user to change this.
Dana Lewis
@danamlewis
Aug 31 2015 19:52
Yea, would love to draft a list of what all we could default as safety features.
Scott Leibrand
@scottleibrand
Aug 31 2015 19:52
is anyone running loop overnight and has seen a situation where openaps didn't high-temp enough on the rise to need to keep high-temping after it started falling?
or is that just a suspicion that you'll need to?
Ali Mazaheri
@amazaheri
Aug 31 2015 19:53
Well said @danamlewis this is not a turn key solution and while most folks can use the shared IP and code as is it requires vigorous testing
Chris Hannemann
@channemann
Aug 31 2015 19:53
No, that happens to me with some regularity
Scott Leibrand
@scottleibrand
Aug 31 2015 19:54
how much IOB do you usually have when riding high and flat?
Chris Hannemann
@channemann
Aug 31 2015 19:54
would have to look back, don't have it at my fingertips right now
Ali Mazaheri
@amazaheri
Aug 31 2015 19:54
it happened to me last night
Chris Hannemann
@channemann
Aug 31 2015 19:54
same here
I think it was around 1 - 2 U
Scott Leibrand
@scottleibrand
Aug 31 2015 19:55
k
Chris Hannemann
@channemann
Aug 31 2015 19:55
ended up needing another 1 U or so
Ali Mazaheri
@amazaheri
Aug 31 2015 19:55
again either site failure, high carb requires manula bolus
Scott Leibrand
@scottleibrand
Aug 31 2015 19:55
one other approach is to only cut off high-temp when BG delta is falling more than BGI
Tim Howard
@tghoward
Aug 31 2015 19:56
I'm building this for our 11 year old and we regularly have to get up over and over throughout the night to keep bolusing the high and keep the temp basal up. Often after a birthday party or something!
Chris Hannemann
@channemann
Aug 31 2015 19:56
@danamlewis this same framework would enable a default prediction algorithm but have several others embedded that simply aren't turned on; again, the "informed" user could swap between them
Scott Leibrand
@scottleibrand
Aug 31 2015 19:56
if you have 1U+ on board, you "should" be falling ~1 mg/dL/min or so depending on the timing
so if you're -2 over 5 minutes, no need to cancel high-temp
but if you go -10, probably should
Dana Lewis
@danamlewis
Aug 31 2015 19:57
@channemann yep and would also enable some of us to cross-test others algorithms to see what ends up working well.
Chris Hannemann
@channemann
Aug 31 2015 19:57
Bingo
Scott Leibrand
@scottleibrand
Aug 31 2015 19:57
@channemann good point. we might want to disable our 15m "deviation extrapolation" feature by default
Ali Mazaheri
@amazaheri
Aug 31 2015 19:58
Capture.PNG additional 1.5 u (0.6, 0.6 and 0.3) helped - remote bolus
Scott Leibrand
@scottleibrand
Aug 31 2015 19:58
or maybe enable it only on the downside to more quickly low-temp when needed, but disable it by default on uptick
not sure that's necessary if we have max IOB set to 0 though
Dana Lewis
@danamlewis
Aug 31 2015 19:58
@amazaheri do you ever think about pooling? or just assume 'bad site'?
Ali Mazaheri
@amazaheri
Aug 31 2015 19:59
I don't think last night was bad site
Ben West
@bewest
Aug 31 2015 19:59
I see that kind of activity, but don't think it's bad site/pooling
Scott Leibrand
@scottleibrand
Aug 31 2015 19:59
ok. I think the trick is to compare delta to BGI instead of to 0. we should try and test that.
Ali Mazaheri
@amazaheri
Aug 31 2015 19:59
its an evil called Pizza :)
Ben West
@bewest
Aug 31 2015 19:59
I think it's just the food continuing to have effect
with pizza in particular I see steady rises for 12 hours
Ali Mazaheri
@amazaheri
Aug 31 2015 20:00
when we make pizza with almond flour things are smooth but
Dana Lewis
@danamlewis
Aug 31 2015 20:00
Right, just wondering if sometimes you observe if he gets up and goes whoosh, I have pooling a lot depending on where sites are and which way I lay, but not sure if other people see that a lot. It's not something you can really deal with for the loop, but it's another mental 'a-ha' for me as 'this is what could be causing it)
Tim Howard
@tghoward
Aug 31 2015 20:00
Yes, we have the pizza problem too
Dana Lewis
@danamlewis
Aug 31 2015 20:00
Pizza is not evil, ya'll.
Ali Mazaheri
@amazaheri
Aug 31 2015 20:00
lol
Ben West
@bewest
Aug 31 2015 20:00
I see the whoosh effect sometimes
Chris Hannemann
@channemann
Aug 31 2015 20:00
Pizza is delicious
Ben West
@bewest
Aug 31 2015 20:00
but similar for indian food, etc...
thick rich curry with rice, etc..
Ali Mazaheri
@amazaheri
Aug 31 2015 20:01
@danamlewis he hates the site on his belly so constant fight not to put it on his back only
Ben West
@bewest
Aug 31 2015 20:01
many sympathies... belly is not good place for devices
Chris Hannemann
@channemann
Aug 31 2015 20:01
I'm about to check out here (work and all that), but wanted to say that the project seems to be moving more and more towards a conservative plug-and-play version with many options available for advanced use, which probably warrants a discussion on how to go forward.
Ali Mazaheri
@amazaheri
Aug 31 2015 20:01
took us a while to convince having DEX on his arms
Ben West
@bewest
Aug 31 2015 20:02
for me, I'd like to have very nice technical docs
Dana Lewis
@danamlewis
Aug 31 2015 20:02
Yep. Probably next week or so after Scott & I do some updating work, would love to schedule a chat so ppl who happen to be working don't get left out :)
Ben West
@bewest
Aug 31 2015 20:02
such that reading through them is enough to get you started if you undersatnd and scare you off if you don't
Chris Hannemann
@channemann
Aug 31 2015 20:02
All for it, @danamlewis!
Scott Leibrand
@scottleibrand
Aug 31 2015 20:02
@bewest heh, that's an interesting aspect I hadn't considered. :)
Ben West
@bewest
Aug 31 2015 20:02
part of that means continuing with the design of lots of simple composable pieces
Dana Lewis
@danamlewis
Aug 31 2015 20:02
@bewest yep see above about language adds
Ben West
@bewest
Aug 31 2015 20:03
then describing the system/behavior/architecture how pieces might work together in complete exhaustive detail
Tim Howard
@tghoward
Aug 31 2015 20:03
@amazaheri we've had dex on arm for many years and insertion site on butt
Ben West
@bewest
Aug 31 2015 20:03
and then list, ok here are the "blocks" that behave by these rules
Oliver Schumacher
@oschumac
Aug 31 2015 20:03
@amazaheri my son hates site in his belly to. Then we tricket him telling the new teflon side is designed only to be in the belly. Since then no prob. :-)
Scott Leibrand
@scottleibrand
Aug 31 2015 20:03
I want to be careful not to give people too many open-ended options without some "here's the safest baseline approach" type of guidance
Ben West
@bewest
Aug 31 2015 20:04
hmm, not sure we know what that is
Scott Leibrand
@scottleibrand
Aug 31 2015 20:04
or we'll have people putting them together in random aggressive configurations
Ben West
@bewest
Aug 31 2015 20:04
yeah, good questions
Scott Leibrand
@scottleibrand
Aug 31 2015 20:04
things like the max IOB of 0
Ben West
@bewest
Aug 31 2015 20:04
polictically, I'm thinking about FDA/industry as well
Dana Lewis
@danamlewis
Aug 31 2015 20:04
Yep, I want a default.
Ben West
@bewest
Aug 31 2015 20:05
right, so a relatively easy thing would be turning channeman's write-up into interactive worksheet using reveal.js or something
but not sure if that'd be too easy
but it would at least provide sane defaults
could just set up for monitoring
and say here's how everything works, here's the rest of the tools
Ben West
@bewest
Aug 31 2015 22:46
blob
that's quick hack feeding that write up through reveal.js
diabeticgonewild
@diabeticgonewild
Aug 31 2015 22:47
I'm not really up to doing a writeup unless somebody wants me to do something specific. Kind of have limited energy recently...
So what do we have write ups for so far?
I mean we should get a list going cause OpenAPS is very complicated.
diabeticgonewild
@diabeticgonewild
Aug 31 2015 22:57
That's nice