Commit Graph
1780 Commits
Author SHA1 Message Date
me 8a2a723d4b llm, continue dropping keep alive since tools which use LLMs correctly manage this themselves. 2026-10-17 12:01:57 -07:00
me 1e8aa6956f llm, always update open-webui. 2026-10-17 11:46:30 -07:00
me caf60d1ccd i3, fix comment since i3 config only allows comments at the start of the line. 2026-10-17 11:46:12 -07:00
me f17b6a96b2 swap-stopper, only restart ollama. 2026-10-17 09:11:08 -07:00
me c8f32b917f llm, try mounting user's Code directory in the Open WebUI container. 2026-10-17 08:47:16 -07:00
me e5891a72c6 system, shorten the runit success sleep. 2026-10-17 08:42:39 -07:00
me b1ebe20874 llm, leave the rm of the docker container to the explicit commands, rather than also expecting run to take care of it. 2026-10-17 08:42:24 -07:00
me 1746d7f11a system, adjust finish script to sleep on success too, just for a shorter time. 2026-10-17 08:19:46 -07:00
me 90ecb842d3 system, add runit sleep script. 2026-10-17 08:18:29 -07:00
me be53012c48 llm, adjust some more parameters and try reducing complexity of qwen config a bit to see if that's why it's stopped working decently, move runit script to system.. 2026-10-17 08:18:18 -07:00
me c709930bf0 llm, add model section back to qwen settings in case that's why it's freaking out. 2026-10-16 17:47:50 -07:00
me 7e0df2ee13 llm, remove model section from qwen settings. 2026-10-16 11:52:09 -07:00
me 6a446a8383 llm, handle max_tokens a little more dynamically, likely only affects the smaller models. 2026-10-16 09:46:46 -07:00
me bedfbe566d rc, add shortcut to restart a service then immediately follow its log. 2026-10-16 09:37:54 -07:00
me 629f7c424d llm, add the actual context values rather than rounding down. 2026-10-16 09:34:08 -07:00
me 0a0b85e66d llm, increase context as to not hinder models from achieving their intended abilities. Keep max_tokens limited where applicable. 2026-10-16 09:23:58 -07:00
me 2706011fa0 llm, add moe tag to a few more model provider names. 2026-10-16 09:10:06 -07:00
me 669095e7cb llm, add more memorable aliases for the fancy ollama show loop. 2026-10-16 09:08:47 -07:00
me 13f20a4e98 packages, add a shell_check, AUR tasks need to be changed similar to android.yml where it echoes what was truly done so we know if it is OK, CHANGED, FAILED, etc. 2026-10-16 09:01:29 -07:00
me cab2013a97 packages, add English and Spanish dictionaries for workstations. 2026-10-16 08:53:16 -07:00
me 8ed0aca797 llm, add a temperature for qwen code providers. 2026-10-16 07:54:34 -07:00
me 79d6f53424 llm, add additional qwen code config to help tame the models. 2026-10-15 18:01:02 -07:00
me 79a8c32584 llm, add longer tineout for qwen, add a few more documentation links. 2026-10-15 17:28:13 -07:00
me 0f4fe291b2 llm, edit provider suffixes. 2026-10-15 12:07:56 -07:00
me 490df1e24c llm, use underscores for model names in qwen, not spaces. 2026-10-14 11:27:42 -07:00
me b26ead5577 llm, fix extra space in provider name. 2026-10-14 11:05:34 -07:00
me bcf2a5aee6 llm, make the qwen drop-down more readable. 2026-10-14 10:32:28 -07:00
me 98b9691e1e llm, add comment for nemo. 2026-10-14 10:28:09 -07:00
me 62306f886e llm, re-reverse the qwen prefix logic. 2026-10-14 10:24:41 -07:00
me 03d0978a64 llm, add sleepy finishes to each service, further edit facts, fix qwen config file and add validator task, move non-ollama web services to non-critical check, start refactoring shell functions, make all context 3200 so programs don;t have to restart the models. 2026-10-14 10:23:03 -07:00
me f3632fa13a llm, add a few more aliases for model purge closer to pull-models (the one used to download them). 2026-10-14 07:29:55 -07:00
me f828e0a9f2 llm, fix the port checking for open-webui. 2026-10-14 07:27:58 -07:00
me 852b904dd4 llm, large refactor on the choices, add a lot of Mistral family models too. 2026-10-14 07:25:12 -07:00
me b7738a1b78 lm, add a few more functions, one just a TBD. 2026-10-14 07:24:45 -07:00
me 13801d9d86 llm, add missing provisioned var to the print. 2026-10-14 05:35:55 -07:00
me 0dc7f258a3 llm, reduce and simplify concurrecy and parallel as to not waste system resources or time. 2026-10-14 05:34:41 -07:00
me d20b1b8114 llm, experimented with adding providers to both directories but was not needed, JSON is strict and does not allow trailing commas. 2026-10-14 04:52:50 -07:00
me f070896ad3 llm, create the qwen config in both the home and projects directory. 2026-10-14 04:44:31 -07:00
me 822140efe6 llm, ensure open-webui check hits the right port. 2026-10-14 04:04:46 -07:00
me 231aa8a9a2 llm, add config file for qwen code. 2026-10-14 03:58:32 -07:00
me 5e3d162481 llm, more changes to open-webui setup to hopefully get it happy. 2026-10-14 03:55:23 -07:00
me eab790c2cc llm, ensure open-webui can see ollama. 2026-10-13 10:51:14 -07:00
me 3c04476217 llm, ensure user and root and in docker group in case that's why the port was not being forwarded? Also may allow user to run docker commands without sudo. 2026-10-13 10:38:33 -07:00
me e762e1231a llm, ensure the OLLAMA_BASE_URL still gets set. 2026-10-13 10:22:24 -07:00
me 44ae9967f1 llm, move links to ollama env documentation. 2026-10-13 10:21:48 -07:00
me fe4f23fdb0 i3, don't worry about Telegram. 2026-10-13 10:11:13 -07:00
me c213397550 llm, add comment as to why the -p and related commands are not being used. 2026-10-13 09:58:19 -07:00
me 951959a00b llm, finalize the open-webui setup, network=host works great, still really unsure why the port mapping was not forwarding correctly. 2026-10-13 09:54:29 -07:00
me 9d7a6414b3 swap-stopper, go back to continuing after the restarts are called. 2026-10-12 12:12:59 -07:00
me 18b73f7665 swap-stopper, add another process which needs stopped so that it does not keep starting models. 2026-10-12 12:10:44 -07:00