November 8Nov 8 Author @tazman ,eisy-ai does not know anything about network resources neither for programs nor for plugin-dev. And that's on purpose.Regarding session management, that's something we need to finetune. For now, you can change history_token_budget to 500000 in the configuration json. Awesome regarding the occupancy plugin!With kind regards,Michel
November 11Nov 11 I'm just getting back into this, and created a second plugin, of a totally different nature. No hardware or network access, strictly making date related calculations, like if we're in an odd or even week of the year, if we're in the last full Monday to Friday week of the month, etc. Cities have been getting creative on setting up services like garbage collection (pardon the trash talk...) and getting advance notifications for these can be useful. In my city garbage collection is every week, but alternating between recyclables and regular garbage. Compostables is every week, except for winter (December to March) where it's every two weeks. Then there is large item pickup once a month, on your regular garbage day that falls in the last full Monday to Friday week of the month. Automate that... Well now I can! 🙂Other handy values are weekday instance of the month (eg: second Wednesday)I need to get around to setting up github and make this available to others. Edited November 11Nov 11 by Guy Lavoie
Thursday at 03:51 PM5 days I ran into an interesting little issue this morning. In modifying my Calendar plugin, I asked it to remove some text descriptors it was putting in some values, such as "13 days" for the number of days left in the month. I want just the number because I want to be able to do numeric comparisons, like > , etc. It was trying to change the UOM to 56 (raw) to do that and at one point asked me to delete the existing nodes in PG3x. It seems to be thinking that PG3x is being used for plugin management, even though this one was created from scratch as an integrated plugin. The other thing is that unlike with PG3x where you can click on the "Nodes" heading of a plugin and manually delete nodes, there is no such thing available in the integrated plugins, that I can see anyway. When it saw that it couldn't work, it took a more roundabout route to getting done, successfully.
Thursday at 04:38 PM5 days Author @Guy Lavoie ,Thank you very much for the feedback. Definitely remnants in the codebase. We'll fix it.With kind regards,Michel
Thursday at 10:13 PM5 days Author Hello everyone,Some exciting stuff:User guide! [ link ]Migration of Python from 311 to 312Added a tool that manages 90% of github setup so that all you need to do is to a) only once: setup your github account b) create a repo and set it up with the created by eisy-ai, and c) point eisy-ai to that repo. Then just to git stuff directly from eisy-ai such as add, commit, push, fetch ...IMPORTANT WARNING - you need the latest udx and eisy-ui. Upgrading from Staging will upgrade your OS to 14.5 AND your python to 312. Please be patient. It may take up to 2 hours.With kind regards,Michel
Saturday at 02:13 PM3 days I have installed the plugin and it shows connected. I've restarted the plugin after config and rebooted the eisy but I do not see the icon at the top for eisy ai.After some digging around in the logs, log errors suggest some dependency errors for the plugin.Details:eisy firmware: 6.1.2Polyglot: 3.4.18Python: 3.11Plugin: eisy-aiStartup error:ModuleNotFoundError: No module named 'unified'Dependency installation errors:jiter 0.16.0 - Rust/maturin build failurerpds-py - Rust/maturin build failurenucore-ai>=3.6.0 installation incompleteSuggestions how to fix this issue? Edited Saturday at 02:29 PM3 days by donp spelling error
Saturday at 02:47 PM3 days 33 minutes ago, donp said:I have installed the plugin and it shows connected. I've restarted the plugin after config and rebooted the eisy but I do not see the icon at the top for eisy ai.After some digging around in the logs, log errors suggest some dependency errors for the plugin.Details:eisy firmware: 6.1.2Polyglot: 3.4.18Python: 3.11Plugin: eisy-aiStartup error:ModuleNotFoundError: No module named 'unified'Dependency installation errors:jiter 0.16.0 - Rust/maturin build failurerpds-py - Rust/maturin build failurenucore-ai>=3.6.0 installation incompleteSuggestions how to fix this issue?I would suggest running upgrade packages, afterwards reboot, wait until everything is up and then re-install the plugin.(I am not sure if this is still a requirement, but as recently as of two weeks ago, you needed to be on Staging repo to use eisy-ai). Edited Saturday at 02:50 PM3 days by DennisC added note
Saturday at 03:09 PM3 days 17 minutes ago, tazman said:@donp you have to upgrade from the staging repo to run eisy-ai.@tazman Ah. Thanks for the info. I did not see that prereq in the user guide install. I guess I will wait until this is a production release. Too risky to run non-prod release for me.
Saturday at 03:22 PM3 days On 11/12/2026 at 5:13 PM, Michel Kohanim said:Hello everyone,Added a tool that manages 90% of github setup so that all you need to do is to a) only once: setup your github account b) create a repo and set it up with the created by eisy-ai, and c) point eisy-ai to that repo. Then just to git stuff directly from eisy-ai such as add, commit, push, fetch ...With kind regards,MichelHi Michel,Could you expand a bit on any security concerns with storing the github push token inside eisy-ai?The first time I pushed to github before this upgrade, eisy-ai recommended not storing the token within eisy-ai.Thanks.
Saturday at 05:37 PM3 days What do I need to do to install the plugin? It just times out when I try to install it.Logs:2026-11-14 12:32:27 [error] web: Error installing plugin eisy-ai in slot 1: timeout of 360000ms exceeded2026-11-14 12:32:27 [error] web: Error with API request to POST /api/plugins/store/prod/install: timeout of 360000ms exceeded2026-11-14 12:32:27 [http] web: POST /api/plugins/store/prod/install : 500 : 360290ms Versions:OS Version: 14.4-RELEASE-p6eisy-ui: 1.4.8IoX: 6.1.2_10UDX: 5.1.0PG3x: 3.4.28iox-matter service: 1.3.24mosquitto: 2.0.18mDNSResponder: 2881.100.56.0.1libwebsockets: 4.3.3
Saturday at 05:45 PM3 days 6 minutes ago, dinglis said:What do I need to do to install the plugin? It just times out when I try to install it.You need to have the latest packages from Staging, per Michel's post above:"IMPORTANT WARNING - you need the latest udx and eisy-ui. Upgrading from Staging will upgrade your OS to 14.5 AND your python to 312. Please be patient. It may take up to 2 hours."
Saturday at 11:06 PM3 days Lots of eiay-ai sucess!Yesterday ai helped me create a solaredge cloud/API based plugin that works around many of the new limitations of the new v2 API.We also created a plugin for my Hyundai Ionic 5 which show its charging status, climate status and windows/doors status and it accepts commands.Today we completed a plugin for the solaredge which connects locally via ethernet using the modbusTCP protocol and avoids all the new API limits and can be polled every few seconds rather than every 15 minutes. I never got the modbusTCP plugin in the store to work so I had eisy-ai build its own.The issues with the context window overflow are now gone!A suggestion:ai needs to be able to read and modify notification templates. With the new plugins I am faced with finding and replacing lots of ugly old node and variable definitions with new ones. seems like the perfect job for AI.
Monday at 02:56 AM2 days I am just starting with the ai plugin. It is configured and working. However, just trying to get it to design a simple program adding days of rain over a week it says"sorry, something went wrong processing that".
Monday at 03:05 AM2 days Same here, I couldn't get anything done on it this evening either. Seems disconnected in some way.
Monday at 06:08 AM1 day Look at the plugin log. I tried it earlier today and it said no more tokens available until 12/1.
Monday at 05:25 PM1 day Michel said I had used up my API allocation so I’m investigating the API options available.Are there any statistics on how large the context windows are on typical eisy-ai tasks? Are there favorite models for cost conscious users? Are there examples of prompts that minimize token usage? I’ve been reading about the most productive ways to use LLM APIs since I’ve only been a casual chat type user up until now. I was surprised to learn that the LLM sends all the input and output information from a full session in each subsequent turn (each command you give). I’m wondering if EISY-AI uses any techniques to minimize token usage. such as model routing, more frequent compaction, caching, flex service tier.
Monday at 06:34 PM1 day Author On 11/14/2026 at 7:22 AM, DennisC said:Hi Michel,Could you expand a bit on any security concerns with storing the github push token inside eisy-ai?The first time I pushed to github before this upgrade, eisy-ai recommended not storing the token within eisy-ai.Thanks.There are many security holes with this approach, specifically, a malicious injected prompt can get the security keys. I think we can make it a little more secure by making the keys scoped to the repo, but at the end of the day, the most secure approach is set things up manually.48 minutes ago, photogeek54 said:Michel said I had used up my API allocation so I’m investigating the API options available.There were only 2 cases of usage of over $100, and you were one of them (around $140). I suspect the issue is the 500K context history, but I am not certain. We'll review and report.12 hours ago, Jimbo.Automates said:Look at the plugin log. I tried it earlier today and it said no more tokens available until 12/1.Please try again, please make sure your context history is not more than 50K.With kind regards,Michel
Monday at 06:39 PM1 day @Michel,Thanks for your reply.I received the following this morning from Anthrop:Your prompt cache hit rate dropped Individual Org's prompt cache hit rate dropped 39% compared with the previous week. Check recent changes to your system prompt or cache_control placement.
Monday at 06:41 PM1 day I still get the same "sorry, something went wrong processing that" error and have not used any of my allowed $20.00.
Monday at 06:45 PM1 day Author @DennisC ,Our cache hit rate, across all the keys, is 80% (used to be 93% before increasing the history token budget). You might want to decrease the history token budget.@SHM ,Are you at the latest version of eisy-ai, OS, and software?With kind regards,Michel
Monday at 07:06 PM1 day The json says history_token_budget is 20000. You said mine was 500k? Should I change something? I’m thinking of using haiku 5.5.
Create an account or sign in to comment