| 7 Aug 2024 |
 hexa | local LLM control in today's release | 13:41:54 |
| hexa | looks like our ollama package is too far behind for the recommended model (llama3.1:8b) | 13:55:48 |
| hexa | should' | 16:09:00 |
| hexa | * should've played the lottery today | 16:09:04 |
| hexa | https://github.com/home-assistant/intents-package/issues/3#issuecomment-2273720952
https://github.com/kk7ds/pynx584/commit/d1dc263c9c29fc429921846bf07ee09cf4304c84#commitcomment-145129705
https://github.com/laundrify/laundrify-pypi/issues/1#event-13791042500 | 16:09:19 |
| hexa | got three upstreams to fix missing tags in under 5 hours | 16:09:46 |
| hexa | meh, they changed the way they deliver static resources | 19:51:36 |
| hexa | which we patch downstream to follow symlinks | 19:51:43 |
| hexa | https://github.com/home-assistant/core/pull/123166 | 19:53:05 |
| hexa | fortunately we have a test for that scenario | 19:55:13 |
| hexa | will have to pass through staging-next for aiohttp 3.10 unfortunately | 20:40:11 |
| 8 Aug 2024 |
|  mrene joined the room. | 01:05:42 |
|  @galaxyyy:matrix.org left the room. | 05:37:22 |
| 9 Aug 2024 |
|  muhlinux joined the room. | 12:47:38 |
 mfed3 | Hey hexa I saw your change to the wiki the other day, so if I just include isal in my extraComponents declarative it will just speed up some operations? Do I need to do anything else? | 18:03:31 |
| hexa | that's all | 18:06:03 |
| mfed3 | Sweet, good deal man | 18:06:19 |
| hexa | I suspect if your machine is fast you won't notice a difference anyway, but it might be cheaper? | 18:06:55 |
 Mindavi | I think I understood it was related to websockets compression so likely it'll help a bunch there | 18:06:56 |
| Mindavi | But not completely sure about it | 18:07:03 |
| hexa |
zlib is a bottleneck for aiohttp, especially for WebSocket connections.
| 18:07:29 |
| hexa | apparently | 18:07:34 |
| mfed3 | Rebuilt fine in 2 seconds...all good | 18:07:40 |
| hexa | [nix-shell:/tmp/python-isal/benchmark_scripts]# python benchmark.py --functions
zlib compression
name isal zlib ratio
0b 1.62 1.64 0.99
8b 1.9 1.83 1.04
128b 2.96 3.55 0.83
1kb 5.23 9.49 0.55
8kb 18.31 49.75 0.37
16kb 32.83 135.72 0.24
32kb 61.92 295.53 0.21
64kb 124.69 621.94 0.2
zlib decompression
name isal zlib ratio
0b 0.56 0.19 2.91
8b 0.6 0.24 2.47
128b 1.53 1.27 1.21
1kb 3.06 3.26 0.94
8kb 11.14 14.23 0.78
16kb 19.79 26.02 0.76
32kb 36.05 81.77 0.44
64kb 77.51 180.41 0.43
| 18:45:18 |
| hexa | and gzip | 18:45:45 |
| hexa | [nix-shell:/tmp/python-isal/benchmark_scripts]# python benchmark.py --gzip
gzip compression
name isal zlib ratio
0b 2.15 2.34 0.92
8b 2.45 2.54 0.97
128b 3.56 4.38 0.81
1kb 5.85 10.49 0.56
8kb 19.03 51.33 0.37
16kb 34.09 138.19 0.25
32kb 63.83 299.16 0.21
64kb 127.24 627.7 0.2
gzip decompression
name isal zlib ratio
0b 1.1 1.55 0.71
8b 1.14 1.62 0.71
128b 2.01 2.72 0.74
1kb 3.69 4.85 0.76
8kb 11.9 15.76 0.75
16kb 20.53 27.56 0.75
32kb 36.91 71.95 0.51
64kb 78.15 179.79 0.43
| 18:45:50 |
| hexa | on a Ryzen 5600X | 18:46:35 |
| 10 Aug 2024 |
|  lxsameer joined the room. | 20:39:06 |
| 11 Aug 2024 |
| hexa | no pretty graph card for sections view type | 00:04:32 |
| hexa | 
Download image.png | 00:05:36 |
