johner

Ok great, maybe github had a brainfart? i uninstalled and reinstalled - all fine to. Let me know if it happens again, maybe I can add some better logging

Thx again for the analysis! fixed and released.

good news! thx for the update.

hi, can i check, are you installing from the CA App Store, or from the link in the first post in this thread?

Hi, so you can configure each agents as per their feature set as defined in their documentation. Most can use a model behind an openAI compatible end point, they are typically configured with environment variables. You can set these for each agent (e.g. Claude, Codex, OpenCode etc.) in the settings for all workspace/sessions, or set them for a specific workspace on the main tab. Is this what you meant?

Hi, thanks for the detailed analysis, let me check it out and get back to you.

Ok so Claude's reasoning :-) Can it be added? Technically yes; good fit, no. The main implications: 1. Category mismatch — it's a general chat-assistant platform, not a workspace coding agent. Different value prop from the fleet. 2. Heavy install — a curl script that pulls Python 3.11 + uv + Node + ffmpeg, partly system-wide. That fights the plugin's isolated-per-agent, RAM-first/SquashFS-on-flash storage model — likely hundreds of MB and containment work needed. This is the biggest problem. 3. Session model mismatch — sessions are conversations, not workspace-scoped, so the per-workspace resume model (#146/#147) doesn't map; only its global -c/latest really works. 4. New wiring required — a new YAML mcp_servers MCP projector (subclass the Goose pattern), plus instruction/skill Config-Hub decisions, a curl version source, and relay delivery + readiness verification, to satisfy the mandatory onboarding guards. So it's point 2 that worries me as Python is not deployed by default with Unraid. I have it installed via a plugin, but I wouldn't want my plugin dependant on it if that makes sense. If you deploy one of the Hermes dockers, the new version of my plugin as an MCP server end point Hermes could call to send messages to the agents you have set up in workspaces in this plugin. They can call Hermes via it's MCP. Thoughts?

Hi! I have no objection in principle, let me see if it complicates things from an agent store point of view etc. A new feature I am testing but not yet released is the ability for agents in different workspaces to DM each other, or subscribe to a channel for broadcast messages, say one agent is the owner of the GPU responsibilities (e.g. local inference) and wants to take it offline for model training, it can broadcast this to all agents listening. I was using monitors for claude, but they are dependant on the model creating them, but I'm now keeping it simple and using TMUX injection to 'type' the announcement, so the user (you) can see it. I'm testing the terminal awareness though, e.g. the session is asking a question so the injected message one be ignored (or worse select an answer the user didn't choose!) It might be a better way than relying on -p if Anthropic ever proceed with making that path pay per use like they previously threatened and dropped. I can have a test tonight, but it might take a few days to release as i need to finish testing the relay/DM solution.

Ok will check it out. I run under root. Did you switch from root to non root and now trying back to root? Any context on what you were doing would be appreciated.

Hi both - it should be fixed now, would you mind giving it a test please?

ok should be fixed now, thanks for your feedback. Appreciate.

great, the latest version have them both integrated. Let me know if anything doesn;t work right, or too constrained in the framework. If you need to run these agents directly on the command line, use the little terminal icon on the drawer tab to launch bash directly in that workspace, you can then run kimi --login or whatever IF needed.

I have a fix prepared, let me see if I can before I release it. I won’t be able to test them well as I don’t have accounts.

thanks, I've fixed the plugin, I'm just testing before publishing. Thanks for your feedback!

oh wow! ok, sorry about that, let me check and fix. Have you been able to remove what ever is left behind manually? If not, point a local instance of one of your coding agents at the server via ssh, it'll figure it out and remove it. Out of interest, and feedback on what would have made it useful for you? Thanks John

Hi, thanks for reporting. This should be fixed now, I'd appreciate it if you could test and let me know as I wasn't able to reproduce on my servers, but claude found a reason why this would happen for some. Thanks John

Hi! So I assume the opencode agent works fine on its own? How are you launching ollama etc.? In the standard unraid terminal, the little terminal button on the workspace drawer tab or another way? Once I know I’ll see if I can get Claude to design a way to support this launch method. In the interim use kilo (or opencode) free models to config the local models in opencode so you can launch it as blower the current design :-)

I went with Orange as Unraid uses 'awesome fonts' for icons and I didn't find a suitable one, and the theme engine doesn't allow for multiple image files.

Hey! These should be fixed now, thx for the continued feedback.

I did think about this but went looking for a theme option and couldn’t find one, clearly I should have googled it! Let me explore the options.

Thanks for letting me know, I’ll get on it.

ok i went with one of yours in the end. let me know if it shows ok.

Thanks Ricardo777, to be honest, because i'm not very good with paint! I think I created that one with gemini originally and AI can't create transparent backgrounds. If you have time and want to tweak it or send another, I'd be happy to consider replacing it? Else, I'll probably have a play when i get a spare hour (which isn't looking likely for a while, I have a 3rd plugin in the creation!)

Antigravity Cli added, along with a number of fixes and a few other enhancements.

Hi @resin1980 @Joly0 I have almost finished a new version that should address both of these. @Joly0 To be sure, if your system still has the debug log, it would be helpful so I can double check not an unrelated issue to what I have fixed. Thanks John