He is running in a venv, not docker, so this doesn’t affect @Juggler
Try using
pip show tensorflow
pip show opencv-python
and see what is running now 
The following is currently installed:
tensorflow. 1.13.2
opencv-python 3.2.0.6
It looks like a new version of tensorflow was installed that is incompatible with my platform.
Moved from 0.97.2 to 0.98.1 on Hassio Raspberry Pi 3. I am seeing very slow response using GUI under 0.98.1. 2-5 seconds to refresh views. No errors in log. Anyone else seeing this? It happens the worst on Hassio page.
EDIT: Moved now to 0.98.2. Still the same issue with Hassio page reloading very slowly. Typically takes 5 seconds to load. Sometimes more.
EDIT2: after a few hours, the system seems to have settled down and the delay when opening the Hassio tab is only a couple seconds now.
Try
pip3 install --upgrade tensorflow==1.5.0Tried this, however I still get the same error. I think I’m going to hold off on doing anything else custom until after the next release. There looks to be a cool integration (New Custom Compontent - Image Processing - Object Detection - DOODS) that might be included in the next release that will accomplish the same thing.
It’s a custom component so won’t be in any release untill the developer submits and gets approved to add it to HA
I’m getting same envisalink errors after updating to 0.98 and above. Did you figure this out. Nothing else has changed in my config.
Did you ever get this working. I’m running Tensorflow in a venv as well, and dont want to lose that.
My tensorflow is working and I have a venv, this is really going to come down to specific platforms I think. In this case the user needs to install 1.15 however the manifest.json updates to 1.13.2 so I suspect if he were to run a custom component with the expected tensorflow version it should be fine. Me for example I was previously on 1.11 so 1.13.2 works. They probably added support for that platform in a later version than 1.13.2 I am not sure why we are sticking with this version. Check what version you are running and if its later than 1.13.2 then you may have some issues you will need to correct. Probably just need to ensure the correct version is loaded. Previously this integration did not require a specific version but called out 1.11 in the docs.
No luck. I tried @nickrout suggestion, but still saw the same errors. I haven’t delved deeper as I think the DOODS custom component is on track to be officially integrated.
Good deal, when I ran pip show tensorflow from my venv it returned
Name: tensorflow
Version: 1.11.0
Summary: TensorFlow is an open source machine learning framework for everyone.
So it sounds like I’ll be good as well
System Options
Enable newly added entities.
If disabled, newly discovered entities will not be automatically added to Home Assistant.
this doesn’t actually work for me, new entities are still being added inside the integration page… anyone else having this issue?
I’ve noticed an icon missing after upgrading from 0.97.x to 0.98.1 ,2 &3 …icon still missing (icon: 'mdi:weather-partlycloudy'). Did your’s come back?
EDIT: nevermind, I found a missing’-’ in the code… not sure how that happened unless MDI changed it and it wasn’t there previously…? anyway, fixed now
No. It did not. I was using mdi:coin for some of my prices, and they just stopped. I just had to change them to something else. There were two others, but I can’t recall what they were.
mdi develop new icons, and drop some old ones, from time to time. When ha syncs a new version into the ha source, there are drop outs accordingly.
1 Like
The partly cloudy icon was renamed by material design. It is now:
mdi:weather-partly-cloudy
note the extra dash between partly and cloudy.
This has happened before, they renamed their raspberry pi icon a couple of releases ago.
It seems philips hue support is broken on 0.98.3:
2019-09-06 09:41:54 ERROR (MainThread) [homeassistant.components.hue.light] Unable to reach bridge 192.168.0.202 ()
2019-09-06 09:41:54 ERROR (MainThread) [homeassistant.components.hue.light] Unable to reach bridge 192.168.0.201 ()
2019-09-06 09:41:58 WARNING (MainThread) [homeassistant.components.light] Setup of platform hue is taking over 10 seconds.
2019-09-06 09:42:00 WARNING (MainThread) [homeassistant.components.light] Setup of platform hue is taking over 10 seconds.
2019-09-06 09:42:06 WARNING (MainThread) [homeassistant.components.binary_sensor] Setup of platform hue is taking over 10 seconds.
2019-09-06 09:42:06 WARNING (MainThread) [homeassistant.components.binary_sensor] Setup of platform hue is taking over 10 seconds.
Keep getting “entity not available” errors from all lights.
I used to resolve this by restarting home assistant, as it would reload the lights, but lately that fix is not working at all.
Funny tho, all the temperature, motion and light sensors work correctly its just the light that are missing. Like this from the integrations page:
Hue motion sensor
by Philips
Connected via Philips hue 1st
Firmware: 6.1.1.27575
Kitchen Sensor light level
sensor.kitchen_sensor_light_level
Kitchen Sensor motion
binary_sensor.kitchen_sensor_motion
Kitchen Sensor temperature
sensor.kitchen_sensor_temperature
Hue ambiance lamp
by Philips
Kitchen
Connected via Philips hue 1st
Firmware: 5.127.1.26420
(entity unavailable)
light.kitchen_cool
1 Like
Noticed a small problem with the UI in this release (on 0.98.1 at the moment) that I don’t believe has been mentioned yet: The state of a script is not updated in the UI. That is, if a script is running, the little “switch” on the UI card always stays off.
This was not the case in prior releases but unfortunately I don’t remember which version I came from to 0.98.1 (cleaned up docker images).
Easy to reproduce. Just launch two browsers and toggle some script that has delay or otherwise runs for a while. It will look active in the browser where you activate it, but off in the other.
It used to be that when my automations activated scripts, I would see the script states toggle in the UI. Not the UI doesn’t change even though the scripts to get turned on and run as expected.
Just me? Known break?
No one has seen any of this on Chrome or Firefox using GUI edit mode?
Card previews in the editor are flickering.
Pop up notifications are occasionally looping.
After updating a card the screen occasionally goes blank with “reload lovelace” visible in the middle of the page.
No relevant errors in the log.
It’s like the browser is constantly refreshing during editing but fails to do so once editing is complete. Ctrl + F5 is required on all views after editing a card in one view.