fiatrete / OpenDAN-Personal-AI-OS

OpenDAN is an open source Personal AI OS , which consolidates various AI modules in one place for your personal use.
https://opendan.ai
MIT License
1.64k stars 132 forks source link

This project doesn't seem like an operating system #23

Open sharpbai opened 1 year ago

sharpbai commented 1 year ago

Although the project's name includes "OS" I haven't seen any code related to an operating system.

Is it really an operating system project?

DiligentCatCat commented 1 year ago

Indeed, it's only a prototype, nothing about OS yet.

The reason that we release this very start Jarvis, is that want people use Jarvis to feel what an AIOS might be in the future, to feel that how an AI would be used by people.

As for the OS, we think it should be an environment that AIs can be integrated easily, and AIs can interact with each other in a uniform and easy way. Thus developing and distributing AI based applications would be easy.

We are thinking the formal architecture recently.

p6002 commented 1 year ago

I just installed as VM and it looks like normal Ubuntu. What's next?

fiatrete commented 1 year ago

I just installed as VM and it looks like normal Ubuntu. What's next?

You are right, the os_installer is merely a placeholder, signifying our ultimate goal of building an AI operating system. The current code in the repository is for our butler agent, Jarvis. We are in the process of designing and prototyping the code for the OS. We estimate that it will be available within a month. Please stay tuned for updates.

p6002 commented 1 year ago

I'll follow up, but the current readme on Github, along with the description, images and videos, suggests that at least what's in the images is available. So far the iso image has been downloading/creating for an hour, I installed this on a VM in Proxmox, and there is normal Ubuntu :) no IP addresses, ports from the graphics in the Readme work. That's why people are disappointed with this "marketing campaign" in recent days.

fiatrete commented 1 year ago

I'll follow up, but the current readme on Github, along with the description, images and videos, suggests that at least what's in the images is available. So far the iso image has been downloading/creating for an hour, I installed this on a VM in Proxmox, and there is normal Ubuntu :) no IP addresses, ports from the graphics in the Readme work. That's why people are disappointed with this "marketing campaign" in recent days.

Thank you for understanding. It seems like the marketing folks have talked a lot, but that is indeed what we want to do. I will remove the os_installer folder before it can function properly.

The code for the Jarvis agent is executable(I have already used it to generate many images for me using Stable Diffusion.) To run the functions for analyzing YouTube videos and sending tweets, there is some configuration work that is a bit more complex. This is mainly because you need to obtain the relevant secret keys from Google and Twitter. I'm thinking of ways to simplify this process.

Anon2578 commented 1 year ago

what happen to the jarvis core file can we really modify this project ourselves ? i was trying to add configs into this project but it seems you all have it hard coded on what to say like for example it keeps mentioning opendan and cant even edit the configs to change that because the core file was deleted or is no longer available. Is this the purpose of this project ? Because i was going to add some login credentials to give to jarvis to use extend functionality but im not so sure since the core file was deleted im not sure if this project is really able to be modified and used how we want as it states. But i could be wrong can anyone help me with this issue because the jarvis core.py file seems to be gone or deleted and cant further add configs that suits me best.

DiligentCatCat commented 1 year ago

what happen to the jarvis core file can we really modify this project ourselves ? i was trying to add configs into this project but it seems you all have it hard coded on what to say like for example it keeps mentioning opendan and cant even edit the configs to change that because the core file was deleted or is no longer available. Is this the purpose of this project ? Because i was going to add some login credentials to give to jarvis to use extend functionality but im not so sure since the core file was deleted im not sure if this project is really able to be modified and used how we want as it states. But i could be wrong can anyone help me with this issue because the jarvis core.py file seems to be gone or deleted and cant further add configs that suits me best.

What did you mean by 'core file'? Did you mean the file defining the basic information of Jarvis? If so, the gpt_agent.py might be what you want, and yes, this part is hard coded for now. You can find it in gpt_agent.py. And you can modify the prompt to whatever you want it to say.
As for login credentials, I'm sorry to say that Jarvis does not support permission control for different sessions yet, it's now designed for private usage, in which case it's not necessary. If you want to add login credentials and provide different permissions for different sessions, it will take some effort to support. AFAICS, you will need to add authorization information into CallerContext (or session), and check permissions before invoking functions.

Anon2578 commented 1 year ago

Thank you and can you help me add the llm models I want to Jarvis for example I want to use some huggingface llm instead of openai but I’m not 100% sure how to add the file it link to the code. I see in the env it says small llm and main llm I know adding the huggingface link or finding to that don’t help do I have to download the file and then add it to the code somehow I’m confused on that part and another thing with stable diffusion the web address we’re do I get it from?

On Sat, Jun 24, 2023 at 6:14 AM Diligent Cat Cat @.***> wrote:

what happen to the jarvis core file can we really modify this project ourselves ? i was trying to add configs into this project but it seems you all have it hard coded on what to say like for example it keeps mentioning opendan and cant even edit the configs to change that because the core file was deleted or is no longer available. Is this the purpose of this project ? Because i was going to add some login credentials to give to jarvis to use extend functionality but im not so sure since the core file was deleted im not sure if this project is really able to be modified and used how we want as it states. But i could be wrong can anyone help me with this issue because the jarvis core.py file seems to be gone or deleted and cant further add configs that suits me best.

What did you mean by 'core file'? Did you mean the file defining the basic information of Jarvis? If so, the gpt_agent.py might be what you want, and yes, this part is hard coded for now. You can find it in gpt_agent.py. And you can modify the prompt to whatever you want it to say. As for login credentials, I'm sorry to say that Jarvis does not support permission control for different sessions yet, it's now designed for private usage, in which case it's not necessary. If you want to add login credentials and provide different permissions for different sessions, it will take some effort to support. AFAICS, you will need to add authorization information into CallerContext (or session), and check permissions before invoking functions.

— Reply to this email directly, view it on GitHub https://github.com/fiatrete/OpenDAN-Personal-AI-OS/issues/23#issuecomment-1605358782, or unsubscribe https://github.com/notifications/unsubscribe-auth/A7I36YMVTCIROZKRP33O3S3XM24WPANCNFSM6AAAAAAZHQZN6U . You are receiving this because you commented.Message ID: @.***>

DiligentCatCat commented 1 year ago

No, you don't need to download any model. Jarvis is a pure network program, it does not do any prediction. All AI abilities are accessed through network request.

Thank you and can you help me add the llm models I want to Jarvis for example I want to use some huggingface llm instead of openai but I’m not 100% sure how to add the file it link to the code. I see in the env it says small llm and main llm I know adding the huggingface link or finding to that don’t help do I have to download the file and then add it to the code somehow I’m confused on that part and another thing with stable diffusion the web address we’re do I get it from? On Sat, Jun 24, 2023 at 6:14 AM Diligent Cat Cat @.> wrote: what happen to the jarvis core file can we really modify this project ourselves ? i was trying to add configs into this project but it seems you all have it hard coded on what to say like for example it keeps mentioning opendan and cant even edit the configs to change that because the core file was deleted or is no longer available. Is this the purpose of this project ? Because i was going to add some login credentials to give to jarvis to use extend functionality but im not so sure since the core file was deleted im not sure if this project is really able to be modified and used how we want as it states. But i could be wrong can anyone help me with this issue because the jarvis core.py file seems to be gone or deleted and cant further add configs that suits me best. What did you mean by 'core file'? Did you mean the file defining the basic information of Jarvis? If so, the gpt_agent.py might be what you want, and yes, this part is hard coded for now. You can find it in gpt_agent.py. And you can modify the prompt to whatever you want it to say. As for login credentials, I'm sorry to say that Jarvis does not support permission control for different sessions yet, it's now designed for private usage, in which case it's not necessary. If you want to add login credentials and provide different permissions for different sessions, it will take some effort to support. AFAICS, you will need to add authorization information into CallerContext (or session), and check permissions before invoking functions. — Reply to this email directly, view it on GitHub <#23 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/A7I36YMVTCIROZKRP33O3S3XM24WPANCNFSM6AAAAAAZHQZN6U . You are receiving this because you commented.Message ID: @.>

Anon2578 commented 1 year ago

Thanks for the reply. How do I specify the huggingface model I want Jarvis to use ? Can you provide me an example on how to set this up ? And what happens if I want to switch to a different model in the future? And is their any resources or documentation you can recommend for learning more about this because I seen the opendan videos with multiple models in one OS system it was interesting and I been trying to do that for the longest I’m excited to get started in this project.

On Mon, Jul 3, 2023 at 3:14 AM Diligent Cat Cat @.***> wrote:

No, you don't need to download any model. Jarvis is a pure network program, it does not do any prediction. All AI abilities are accessed through network request.

Thank you and can you help me add the llm models I want to Jarvis for example I want to use some huggingface llm instead of openai but I’m not 100% sure how to add the file it link to the code. I see in the env it says small llm and main llm I know adding the huggingface link or finding to that don’t help do I have to download the file and then add it to the code somehow I’m confused on that part and another thing with stable diffusion the web address we’re do I get it from? … <#m8743851918811510524> On Sat, Jun 24, 2023 at 6:14 AM Diligent Cat Cat @.> wrote: what happen to the jarvis core file can we really modify this project ourselves ? i was trying to add configs into this project but it seems you all have it hard coded on what to say like for example it keeps mentioning opendan and cant even edit the configs to change that because the core file was deleted or is no longer available. Is this the purpose of this project ? Because i was going to add some login credentials to give to jarvis to use extend functionality but im not so sure since the core file was deleted im not sure if this project is really able to be modified and used how we want as it states. But i could be wrong can anyone help me with this issue because the jarvis core.py file seems to be gone or deleted and cant further add configs that suits me best. What did you mean by 'core file'? Did you mean the file defining the basic information of Jarvis? If so, the gpt_agent.py might be what you want, and yes, this part is hard coded for now. You can find it in gpt_agent.py. And you can modify the prompt to whatever you want it to say. As for login credentials, I'm sorry to say that Jarvis does not support permission control for different sessions yet, it's now designed for private usage, in which case it's not necessary. If you want to add login credentials and provide different permissions for different sessions, it will take some effort to support. AFAICS, you will need to add authorization information into CallerContext (or session), and check permissions before invoking functions. — Reply to this email directly, view it on GitHub <#23 (comment) https://github.com/fiatrete/OpenDAN-Personal-AI-OS/issues/23#issuecomment-1605358782>, or unsubscribe https://github.com/notifications/unsubscribe-auth/A7I36YMVTCIROZKRP33O3S3XM24WPANCNFSM6AAAAAAZHQZN6U https://github.com/notifications/unsubscribe-auth/A7I36YMVTCIROZKRP33O3S3XM24WPANCNFSM6AAAAAAZHQZN6U . You are receiving this because you commented.Message ID: @.>

— Reply to this email directly, view it on GitHub https://github.com/fiatrete/OpenDAN-Personal-AI-OS/issues/23#issuecomment-1617506428, or unsubscribe https://github.com/notifications/unsubscribe-auth/A7I36YIUFXLKPNI7WY5OAY3XOJWPDANCNFSM6AAAAAAZHQZN6U . You are receiving this because you commented.Message ID: @.***>

DiligentCatCat commented 1 year ago

You can specify JARVIS_OPENAI_URL_BASE if the API is compatible with OpenAI API (but the output may be bad since the prompts are not tested).

If not, you'll have to write another agent, see agent_jarvis/jarvis/ai_agent/webui_agent.py, this is an unfinished code for text-generation-webui. You can write your own, and make agent_facotry return your own agent. You can also provide an OpenAI API for your own API, text-generation-webui has a built plugin for that.

Whatever the method you choose, you may need to take some effort to make it work. At least the prompts only suits ChatGPT now.

@Anon2578