@canadaduane Glad I can help

thanks for your return; as you expected; yes the issue is located somewhere else than Cloudron or Open-WebUI it's at a proxy level; which in my case CloudFlare

@firmansi I will have to take a look at mcp and mcpo and OpenWebUI to understand this tool and then make an informed decision.

@JOduMonT is the reranker using ollama or with Jina api?

Question: What should be the correct directory ownership and permissions setting for the ollama-home folder on a mounted Volume?

@BrutalBirdie Hey, thanks for the recovery mode tip for weekend tinkering! I gave it a shot: cloned the app, used recovery mode, and manually ran pip install google-genai in the terminal. That worked, and the Gemini function requiring it did become active in Open WebUI. Unfortunately, it turns out Google removed the "thoughts" feature via the API recently (see discussion here: https://github.com/googleapis/python-genai/issues/226), so that specific goal is moot now, it seems. However, this experiment confirms the underlying need. It would still be very valuable if Open WebUI's own mechanism for installing function dependencies (defined in frontmatter) could work correctly on Cloudron by targeting /app/data. This would allow other current or future functions needing packages (beyond just this specific Gemini one) to be installed seamlessly as intended by Open WebUI, without needing manual recovery mode workarounds. So, the original request to fix that installer path for the Frontmatter mechanism remains relevant for general function extensibility. Thanks again for the suggestion!

It would be awesome if openwebui could use the Cloudron Tika server.

@nebulon understood; it's just in my testing environment (home) I'm use to connect my container via container_name:port without exposing the port. such as OpenWebUI connect to my Searxng because --- services: open-webui: searxng: container_name: searxng Now I understand this is not viable into a multi-domains, multi-services environment Do you think something similar is possible and viable ?

Perfect, thank you!

@ekevu123 said in Transcribe YouTube videos: this videos has captions, right? yes @ekevu123 said in Transcribe YouTube videos: Does it work for you? It doesn't work for me I don't currently have an instance running so can't check, sorry

@nebulon thanks for the tips and do you mind looking at this one ? https://forum.cloudron.io/topic/13320/accessing-the-openwebui-api/5?_=1740050615827 I'm unable to use the APP ID to connect to it.

@timconsidine no, but you can get GPU servers with AI niceties from Koyeb. I think They have a Ollama installer.

@girish Any chance of adding popular python packages to this Cloudron App? Quite a few functions here needing them: https://openwebui.com/functions Eg: https://openwebui.com/f/fovendor/flux_inpainting https://openwebui.com/f/fovendor/flux_gen

This would certainly be a very valuable addition to Cloudron. I have upvoted. That said, I guess that the configuration process could be challenging since there doesn't appear to be any point-and-click UI for that.

I forgot an important advantage: You're supporting open source.

I have found some answers about the Channels feature on Github: Channels Discussion Q: "I've been trying the new Channels feature. I can create Channels, assign groups and use it like a chat between group members. For the life of me, I have not found how to send a message to the AI model." A: "Has to be externally hosted with https://github.com/open-webui/bot for now, but built-in integration will be added soon." Integrating AI bots to Channels This repository provides an experimental boilerplate for building bots compatible with the Open WebUI "Channels" feature (introduced in version 0.5.0). It serves as a proof of concept to demonstrate bot-building capabilities while highlighting the potential of asynchronous communication enabled by Channels. Key Highlights Highly Experimental: This is an early-stage project showcasing basic bot-building functionality. Expect major API changes in the future. Extensible Framework: Designed as a foundation for further development, with plans to enhance APIs, developer tooling, and usability. Asynchronous Communication: Leverages Open WebUI Channels for event-driven workflows. ️ Getting Started with Examples This repository includes an /examples folder containing runnable example bots that demonstrate basic functionality. To run an example, execute the corresponding module using the -m flag in Python. For example, to run the ai example: python -m examples.ai Note: Ensure that your current working directory (PWD) is the root of this repository when running examples, as this is required for proper execution. Replace ai in the command above with the specific example you’d like to execute from the /examples folder. Disclaimer This project is an early-stage proof of concept. APIs will break and existing functionality may change as Open WebUI evolves to include native bot support. This repository is not production-ready and primarily serves experimental and exploratory purposes. Future Vision We aim to introduce improved APIs, enhanced developer tooling, and seamless native support for bots directly within Open WebUI. The ultimate goal is to make building bots easier, faster, and more intuitive.

BTW it seems to automatically recognize any audio file you throw at it. Either via Microphone or file upload and process TTS locally before passing it on to the selected model. Logs and graphs are pretty clear about this. Documentation however is not lol.

Are you referring to the Cloudron package? It is disabled by default in the Cloudron package in both Cloudron auth and non-Cloudron auth mode.

thanks for the info ...have awesome wonderful job