Repository navigation
Swift version implemented for Mac, ios, and ipad os #7049
Replies: 5 comments 8 replies
|
i can make it available for free, i have an ios account and can help with this if I like your source code... i was just going to open a more general discussion |
|
the other efforts i've seen add wayyyy too much crap to odysseus and don't agree with its original design i'm a dad and don't need 50 api endpoints... i love that odysseus is meant to be local, personal, and SIMPLE i was thinking more that the ios app should ideally be just a client for the odysseus server though... so maybe that conflicts with your vision for what you've designed i was going to ask about the roadmap around this as well ... but I think thats a bigger discussion, so i've made an issue specifically to talk about where the current roadmap is here: #7107 contributing docs don't highlight a current roadmap: |
|
last bit, and then I'll wait for your or another's reply does your app have built in ai-chat model... thats one feature i was wanting to add was a totally local voice interface that was integrated into the app (basically a tts translation engine, that then interacts with odysseus just like normal chatting would) |
|
Yes it has a built in access to apples model. You can easily download a couple small hugging face llama models. You can use lm studio, openai, x.ai etc. I can give you beta access if you just want to straight up try it or can can just check in as a new repo. I can also demo it for you which is the easiest. |
|
lol i was hoping to see the code before i used it I do like the features you're talking about though |
|
It has that.. |
|
So, this is mostly AI coded. I have provisioning profiles. so I build locally.This is a sanitized version of what I have in a private repo. Let me know what you think. If you wanted we could get on a call I can demo it for you so you can see if it is worth your time. My main workflow has been openai for llms and elevenlabs for voice. Eleven labs has a $6 membership. I use an API key and a membership for openai. |
|
oh man, you can do local voice stuff without eleven labs :) i need to get my ios setup legit, then I'll reachout, might be a bit, but i'm excited i got a trashy version that did all the builds via github... but now I have a real mac and want to make it all local my biggest issue with how you've designed the app is that its totally independent of odysseus... since Ideally what I think we'd want is for the app to just be a front end, for the same backend toolset that odysseus exposes to an ai agnet... the thing I need to figure out is if you can separate odysseus' tooling around a persona ai assistant from the harness (since i'd much rather use pi as the main harness, and just kinda want the odysseus UI + feature set) so could be i'm off in lala land and missing where the project is designed/headed thats why iasked about the roadmap here: #7107 |
|
I was looking for exactly what Odysseus was but in app form. I wanted to be able to open it on my Mac, my iPad, my phone, have all the data sync, but still be private. I achieved that if a little buggy. As far as using a local voice provider, you can but I like a quality voice. I had apples voices working before I plugged in eleven labs, but openai's voice model worked last time I used it. Realistically you could use your own hosted one too. PI has api access according to AI and if it supports openai style commands it should integrate neatly. The main thing to think about is that I wanted voice to be a first class interface and it is. I wanted everything to be searchable and it is. I wanted an assistant that could summarize or draft emails for me. It can, i wanted calendar and task access it has that. I could tell it to research or write me a doc. It has memory, it will have context compaction, not sure if it made it into this commit. You really should vet it with whatever tools you want then try it. I have used it with open AI's models, local models, models hosted on my network, Apple's model. Some models have tool access, some do not. If you want it to be able to search and do all that stuff, it needs to be a tool utilizing model. Unfortunately, llama from Meta, the small models don't support tools. For the record, I can also publish via the App Store, but I was hoping to connect with someone on the Odysseus dev project to see if they cared. It would be free, of course |
|
yup, and i think with a few tweaks i'm happy to make this work :) i think just in its current state its not what i'd want. since ideally the app interface is just a client for the odysseus server sorry if that bums you out, overall i do feel like soon we're going to be having evolutions in current ways computers are used, so i'm wanting all of this to be a totally local AI setup that I can use and like so if you're cool with me changing it to match the above workflow, i'll go ahead and try and take that for a spin |
|
You are welcome to copy it and do whatever you please within the licensing. Feel free. I am not bummed. I have done what I intended. The store is secondary. I would prefer you check it into a different repo though as it is substantially different. Please feel free to rename it as well. Just keep the license and an attribution wouldn't hurt if it becomes something. Beyond that... I am happy you found something that might work for you. |
|
I also have a separate app, an IDE for coding that allows me to rsync source from my mac mini to my ipad, then use an AI to make code edits, view them, change them, then rsync back to the mac to build and test. It features a voice driven interface and keyboard of course. I drive a fair amount of development work from grokbot or from that app. The nice thing about my app I call “deep edit” which is probably an oxymoron, all things considered, is that the AI has full control of everything in that space because it is just a directory. It can read the logs, it can run builds (on the mac mini) all very fluidly. It understands cmake and xcode projects at this point, though you can use whatever you want if you change the code. It also has a terminal integrated (thanks to the swift packages that are imported) and operates overssh. This is another vibecoded app with a bit of manual work thrown in. |
Uh oh!
There was an error while loading. Please reload this page.
Sometime back when the project first went live, I tried it and loved it, but disliked the fact it was not native. I have created a fully functional iOS, ipad os, and Mac version of the app. It has been my pet project for the last few months. I am interested in sharing the source, but also interested in ensuring someone makes it available for free through the app stores (swift based). Who can I speak to about it?
It has access to the on device ai, download some smaller ollama models, or remote models. It has a voice mode, and more it is pretty complete. It is my go-to platform on my iPhone. It also has import and export capability of the db, import from a notion export, and buggy synchronization across your iCloud account. It can read and access your Gmail, calendar, has tool use and access to chat, research, docs, etc. I have also added search across the lot of it.
All reactions