Cloudron Forum

@humptydumpty I finished the upgrade !
Apologies again that you got caught up in the middle of it

Attempting to be concise, I have completed changes to the CCAI environment.

CCAI

1. original "hosted-as-a-service" ccai.appx.uk is now disabled and now a placeholder to avoid broken forum links
2. new "entry point" is customappgateway.appx.uk

CustomAppGateway

1. previously just an installer for CCAI-P, via browser or binary
2. deliberate web UI change to create distance from old system/concept

3. no more browser-based installations involving submission of creds in a browser
4. only supported deploy route is an installer script using a binary (Mac, Linux, Windows as needed), which cleans up by removing the binary on completion
5. now can install not just CCAI-P but other custom apps
6. custom apps master catalogue now lives in customappgateway
7. recommended to install CCAI-P and then install apps from there to enable updates
8. but there is also an option to install a single custom app from customappgateway, for those who don't want CCAI-P and don't care about updates
9. I re-factored the installer binary to be available for use by others - driven by run-time arguments, no custom app meta-data baked into the installer : see customappgateway.appx.uk Advanced section

CCAI-P

1. CCAI-P now supports updates to installed custom apps ... providing they were installed using the new CCAI-P version (uses local 'installed.json' to know what to check)
2. Master App Catalogue now maintained (only) by me - I will attempt to keep populated & updated, but let me know if something needs adding or changing - happy to do so
3. CCAI-P now supports a Local Catalogue, maintainable by you ! Independence, privacy, experimentation - whatever !

4. CCAI-P app update checking and processing is not fully checked (I haven't got any apps to update and I haven't invested time in faking a test). Let me know if any issues. Will try to do some real or fake testing.

Given Cloudron 9.1 upcoming support for custom apps, I am not planning any further major changes, just fixing any reported issues.

@humptydumpty sorry about your experience - heavy updates and and changed locations - mid-process of going live - almost done now - you might be running into the ecosystem update process.

I will update you when finished. Hopefully not long now

@crazybrad yep - in my cautious approach I set up a dedicated address e.g. my-company-finance@app.domain.tld and forward manually/automatically emails to it, bot monitors email inbox and works only on those - not full primary email access

@robi good approach, but I suspect your neuro-plasticity far exceeds mine !

@crazybrad I have very similar thoughts & cautions.

• packaging it because it looks cool and fun to learn what the fuss is about
• I would NEVER deploy this on a primary working computer like desktop/laptop
• deploying it in a container seems like it could be viable

I have a feeling that Agent Zero is much better than Openclaw, but I can't really say that until I have test Openclaw fairly. So continuing on the package "just in case" it can help me or others.

Click-bait heading but explanation :

Suffering from a tantrum (I blame it on low blood sugar), I decided that I have had enough of AI chat engines forcing me to compose longer messages using <shift>+<enter> because plain <enter> submits the message to the AI dev assistant.

Usually that causes a mal-formed request to the agent, before I have given all current instructions and relevant details. It's especially bad when I am asking for strategy advice, and it only gets half the relevant story.

While some chat UI's have handled this by allowing the user to choose what is the submit key, most haven't and most frustratingly, ALL the cores ones I use for dev haven't, and I get the impression never will.

It really slows down my flow.
So I decided there's only answer - disable the enter key !

I used a free utility https://karabiner-elements.pqrs.org to re-map the <enter> key to <space>. I could have totally disabled it, but I thought <space> is the safer option.

Now I will never send a message to an agent by mistake, and I no longer care whether they fix this issue in future or not.

So how do I actually implement <enter> when I need to ?
I also re-mapped <enter> to a seldom used key (in my case <RightCommand> on Mac).

Is it a perfect answer ? Absolutely not.

• I still have to exercise some restraint not to press <RightCommand> when I don't mean to, but as it is non-standard, it is much easier to do that
• I still have to press <shift>+<RightCommand> for longer multi-line messages. But that's not significantly different to previously.

Dies it prevent malformed submits ? Absolutely

• I am protected from sending <enter> when I don't mean to, using <RightCommand>is a more conscious choice.

I haven't yet bothered to configure key usage on a per-application basis, (in email I might want to keep normal <enter> usage).

Posting this in case you clever people have better solutions.

• I tried dictating prompts and only reaching for the keyboard when necessary. But it's not a flow I am used to.
• I tried 'gush mode'. No line breaks at all - one long paragraph. But this doesn't cut it when I need a list.

How have you solved it ?
It would probably be a good idea to have conventional use of <enter> key

@girish thank you - very helpful - still digesting it.

Hoping to get my final "end of life" changes done today, which can allow CCAI-P to limp on until Cloudron in-platform functionality is tested, then turn the CCAI and CAG websites into digital tombstones.

No need for any rush - always happy for you to take your time to do a proper job (unlike my hashed together PoC).

House-keeping :

• Cloudron have announced they will introduce support "in-platform" for custom app deployments
• consequently CCAI (the open hosted-as-a-service version) can no longer be installed from CCAI. (It can be installed manually if really needed but I don't really see why anyone would want to).
• CCAI-P (the personal edition which stores auth creds locally in app) can no longer be installed from CCAI : if you want CCAI-P, use the new https://customappgateway.appx.uk distribution point
• CCAI-P has been cleaned of legacy auth timeouts (inherited from original CCAI) so runtime gremlins should not occur. There is still a manual logout and connect option, in case the user wants to do so manually. Otherwise it should remain 'ready to go' without timeouts.

Improvements :

• both CCAI and CCAI-P still focus on new installations, and there is no support for updating an existing custom app. I am working on introducing support for updates.

The Future :

• with Cloudron's announcement that their v9.1 release will include some support for in-platform deployment of custom apps, "the writing is clearly on the wall" and CCAI / CCAI-P / CustomAppGateway will die and git repos / docker images will be removed at some point.
• I doubt anyone is more pleased than me !
• however we do not know the timescales for Cloudron 9.1 (other big stuff in it too), so I will continue to make available CCAI / CCAI-P until some point after Cloudron 9.1 is stable.

I'm calling this a v1.0.0 package
I may add/tweak something later, but I'm now using this as a production tool.

Currently I like it more than openclaw (clawdebot, moltbot).
But returning to work on my package of openclaw, and will reconsider which is better when openclaw package is complete.

I really like Agent Zero support for Telegram, and a relatively easy addition of new skills and tasks using the /run xxx structure.

My testing and live usage has mostly been with my selfhosted Ollama instance which is configured with ollama cloud models (so large models run on ollama servers). Seems low cost and large power).
If you test with other providers and issues found, let me know.

It's a big image (3.9Gb), so make sure you have disk space. But it installs quite quickly.
Make you set memory for your installation quite high - I have 4Gb and maybe will bump a little higher for production use.

v0.1.12 : fixed searx issue, now using Brave Search API key

now pushed to git.cloudron.io/cloudron-agent0 and docker hub image tcmbp132021/cloudron-agent0:latest (aka v0.1.12)

@miednr I build all my apps using local device docker, push to a cloudron private registry and use cloudron install --image xxx.
I feel guilty in some way for not using Cloudron Builder, but local gets the job done fine, so good that you got your build out and deployed.

Nice, but I know nothing (yet) about Ansible

@robi thanks for that link
still working on my openclaw package, so that will be useful (incorporated if I can)

@humptydumpty apologies for the rough ride you're experiencing - CCAI-P was spawned from CCAI and while I removed some things for the personal scenario, I think I mistakenly left in the auth timeout.

I have a total rework of CCAI and CCAI-P in production. It might be redundant in terms of Cloudron's announcement of 9.1. But they have a number of big items in that release so maybe I can get the re-worked versions out for a some some service lifetime until 9.1 is available.

Usually logout + connect clears it, but maybe mileage varies.

@luckow strange ... and apologies

line 1 and column1 is unusual
let me look into it.

do you have any more logs (sanitised if need be) ?

problems I hav e seen in CCAI-P are all due to the timeout (inherited from CCAI and due to be removed) on auth - usually solved by logging out (button above list on right of screen, then pressing the connect button which appears in the middle of the screen. I guess restarting does the same thing in effect.

v0.1.6 : ensured MCP servers work, resolved searx dependency

now pushed to git.cloudron.io/cloudron-agent0 and docker hub image tcmbp132021/cloudron-agent0:latest (aka v0.1.6)

v0.0.8 : **extended telegram bot functionality **
now pushed to git.cloudron.io/cloudron-agent0 and docker hub image tcmbp132021/cloudron-agent0:latest

• /run : launch custom user scripts (no container rebuild) from your telegram app - open extensibility :
  • invoice / document downloader (send it a link)
  • task manager (add/remind/complete) ??
  • <whatever you would like it to do>
• /chat : chat with AgentZero from inside your telegram app
• /learn : get AgentZero to memorise new info
• /health : get AgentZero core system report

Bonus cool : some of this was designed/built by me & my TRAE dev assistant, but some was AgentZero itself !
Suggesting and writing changes "to itself" (well, to how the telegram bot could be better deployed inside AgentZero).
Self-healing, self-extending apps around the corner !

@robi you're right.

I have set the task to be run only once per week, it checks for an incoming email received with a subject like "your invoice is ready" from a specific email address (the supplier).

It only executes anything if there is a waiting email like that.

So limited email box polling and very constrained execution scope, and no AI credits used at this point (at this stage it's not parsing or processing the downloaded invoice - just a python script).

AI credits used in building the script, but none currently in execution.

The script is not in the package, it needs to be built by AgentZero specifically to the Cloudron's users needs. I'm still tweaking the script, but don't mind sharing it afterwards as a generic script if anyone wants it (upload to /app/data/instruments and then tel AgentZero to build a scheduled task invoking it).

Manual admin tasks like invoice download and processing are the bane of my life, so I was excited by what openclaw offered, but I think it can be done better in AgentZero.

Specifically about notifying the agent, I'm just building a telegram invoked task - go get this invoice. That might be relevant to your comment.

@robi well, there could be 101 answers to that, email being what it is, but the way I have done it is :

• AgentZero is deployed on Cloudron, e.g. a0.domain.tld, so I have set up a mailbox a0bot@domain.tld. Created an extra Cloudron user xxx_bot and made it the owner of the mailbox (i.e. it has it's own password, I'm not sharing my main password with AgentZero)
• I tell AgentZero the necessary IMAP details, username a0bot@domain.tld, password, mail server address used by your Cloudron, port 993
• AgentZero builds a task (script) using the mailbox config which it is given to poll for emails received, processes them and (optionally) deletes them, using the body content and attachments in the script it is building
• I forward emails to a0bot@domain.tld from other Cloudron email boxes or my primary email (ProtonMail). The email from my Proton has a forwarding rule (e.g. send Hetzner invoices to a0bot@domain.tld). It could be manual forward or auto-forward.

This could be much slicker, but at this stage (a) I'm experimenting, (b) I'm controlling what gets sent to AgentZero.

Depending how ambitious one wants to get, you can build a whole array of email inbox processing w/o forwarding, e.g. just give it my primary Proton mail creds. Maybe I will be brave on that later