tl-dr
-Can someone give me step by step instructions (ELI5) on how to get access to my LLM’s on my rig from my phone?
Jan seems the easiest but I’ve tried with Ollama, librechat, etc.
…
I’ve taken steps to secure my data and now I’m going the selfhosting route. I don’t care to become a savant with the technical aspects of this stuff but even the basics are hard to grasp! I’ve been able to install a LLM provider on my rig (Ollama, Librechat, Jan, all of em) and I can successfully get models running on them. BUT what I would LOVE to do is access the LLM’s on my rig from my phone while I’m within proximity. I’ve read that I can do that via wifi or LAN or something like that but I have had absolutely no luck. Jan seems the easiest because all you have to do is something with an API key but I can’t even figure that out.
Any help?
There are 3 lines with the :11434 in them. No brackets or anything like that. -1 has 0.0.0.0 in front -2 has 10.#.#.# in front and has a foreign address that is something other than 0.0.0 -3 is like the 2nd but a slightly different foreign address
The iPhone does have a 10.#.#.# ip number that is slightly different than the PCs.
The subnet mask is 255.255.255.0
Oh yes. I’m on windows 10 as well.
I have taken a pause here while we trouble shoot the subnet mask. We’re getting close!!
Okay, that…should be okay. As long as all of the addresses that it’s listening on are IPv4 — of the format “x.x.x.x”. No colons in them (other than the colon preceeding “11434”). Not IPv6.
Okay, gotcha. In that case, go ahead with the instructions above, just instead of “/8”, do “/24”. So:
Okay. I think that the interface to add the firewall rule there looks the same as the one I Iinked to above. I went searching for screenshots of adding a hole for a port on Windows 10, and the control panel looks identical to me.
So, yeah, should be good to go ahead with the above instructions, just using “/24” instead of “/8” in the two places where I mention “/8”. Hopefully after that it’ll be working; if not, then we’ll need to troubleshoot.
Hate to say it but it didn’t work. I listed below the things I double checked. I really appreciate you helping me troubleshoot this, but it seems like I may have bitten off more than I can chew. I choose Ollama because it was supposed to be one of the easier loca AIs to set up. Do you have any recommendations for alternatives? Or do you think I should incorporate a docker or open web ui as some others have said ?
-when I went to the ollama app and entered the http://10.#.#.#:11434 , it didn’t work. Also tried the enchanted app and that didn’t work as well.
-I double checked the rule I made to make sure that was inputted properly. The 10.0.0.0/24 for the local and remote ip addresses.
-the sanity check went well. The ipv4 said closed. The ipv6 said failed.
-I checked the netstat -abn thing and 0.0.0.0:11434 is still listening.
Well, the ollama bit is up, which is why you can use it on the PC. The problem is network connectivity between the Windows PC and the phone.
Opening a port between two things on the local network is going to be pretty much the same for anything. Some software packages — I dunno about LLM chat stuff — make use of a third, outside system as a point to coordinate, so that software only has to open outbound TCP connections to the Internet. But for local communication, it’s gonna look pretty similar. If you put koboldcpp or llama.cpp or whatever on your machine, you need the same connectivity, though it might default to using a different port number.
I’m happy to keep banging away if you’re also willing, though. I mean, this does kinda narrow it down. If you don’t want to do so though, remove that firewall rule that we added earlier from the Windows PC. If you do:
considers
The next step is seeing where the break in connectivity is.
I’m not familiar with iOS, but let me see if there’s a software package for it that will let it open a TCP connection and preferably ping (and ideally show the ARP cache to see whether Ethernet packets are getting from the phone to the Windows machine at all, though that may not be viable).
Basically, would be nice to see whether packets can currently get from the phone to the PC and back.
kagis
Looks like Windows Firewall blocks ICMP by default, which is traditionally used by ping, the simple protocol to see if one host can reach another on the network. Mmmmf.
And it sounds like ARP isn’t available on a non-jailbroken iPhone, which would be the simplest way to see whether a packet is making it from the iPhone to the PC. I was worried would be the case.
Hmm. This is a little less convenient in that I don’t have tools that I normally would when trying to troubleshoot network problems on a Linux system.
thinks
I guess the simplest thing available, cuts things down as far as possible in terms of connectivity between the two that should be able to reach from the iPhone to the PC should be a TCP connection.
I don’t know the iOS software library well, but lemme search for a telnet client. I’m sure that it’ll have one; every platform does.
searches
Oh, this is even better. It looks like there’s some iOS app, “iSH”, with a tiny Alpine Linux environment for iOS, kinda like Termux on Android. That’ll have telnet and probably other network diagnostic tools, and those I am familiar with, so I don’t have to guess from screenshots how things work. You should be able to try to open a TCP connection from the phone to the PC with the Linux telnet client in that.
goes looking around
Okay. If you’re willing to give this a shot, it sounds like the way this works is:
https://apps.apple.com/us/app/ish-shell/id1436902243
Install that from the iOS store.
When opened, it should show a Linux terminal. If it works like Termux, it’ll have basically nothing from Alpine Linux installed, no telnet client, just a few simple commands. You’ll be looking at a prompt that probably looks something like
iPhone:~#.Then if you run (don’t type the pound sign — it’s just a convention to include it, to show that it’s something to type at a prompt):
That should install the Linux telnet client inside the iSH app using the Alpine Linux package manager.
Then to try to open a TCP connection from the phone to the Windows PC, you want the private IP of the Windows PC, the thing you see in ipconfig (which I’ll type as 10.1.1.2 here, but replace with yours):
That’ll try to open a TCP connection from the phone to port 11434 on the PC.
Now, what would happen if everything were working correctly, is that the phone would send an ARP request saying “what is the MAC address — the Ethernet address — of the machine with IP address 10.1.1.2 on the local network?” The wireless access point would hand this to the PC. The PC would respond. The phone would then send a series of packets to that IP address to open a TCP connection on port 11434.
My guess is that you’ll see one of several things at this point.
First, it might be that the wireless access point is refusing to let packets from the phone reach the PC at all — they only let the phone talk to the Internet, not to the PC. Some wireless access points can be configured to do this or have a “guest” wireless network that impose this constraint. Then the phone won’t get an ARP response, since the PC will never see the ARP query. That’ll look like this (using a network I’m on at the moment to demonstrate):
Second, it might fail because I dicked up in some way and Windows Firewall is still blocking the phone from connecting to the PC. The ARP request is going out, the response comes back from the PC, the phone tries to open a TCP connection to the IP address on the host with the specified MAC address, and never gets a response. If that’s the case, it’ll probably look like this:
If that’s what you get, the problem is likely the Windows Firewall configuration (well or theoretically the wireless access point could be configured to do that, but I doubt it).
Third, it might succeed. That’ll look like this:
If you see that, you can open a TCP connection from the phone to the PC, and whatever issue you’re hitting with Reins isn’t a network problem. Maybe I gave the endpoint syntax wrong, for example. But the issue will be at the application level, not the network level.
Hell ya! I would definitely appreciate some hand holding thru this process! This self hosting stuff is going to take a bit longer and more learning than I anticipated.
-the opening the port process makes sense. It seems like if I have a backend on my rig, I’m going to need to open a port to access that backend from a front end of a phone device. Or possibly even access that same backend on the phone device via a mirror?
-it seems like it would be easier if I could connect to the rig via an android phone instead of an iPhone. My end goal is to use Linux but I’m not ready for that step. Seems like android would be an adequate stepping stone to move to, especially if we have to go thru all this trouble with iPhone. Shall we try on the android instead? If not I’ll follow the directions you put above and report back on Saturday.
-the opening the port process makes sense. It seems like if I have a backend on my rig, I’m going to need to open a port to access that backend from a front end of a phone device.
Yes. Or even if you run a Web-accessible front-end on the LLM PC — the Web browser on the phone needs to reach the Web frontend on the PC.
Well, the term wouldn’t be a mirror. In your shoes, it’s not what I would do, because introducing some third host not on your network to the equation is another thing to break. But, okay, hypothetically, I guess that doing that would be an option. thinks. There might be some service out there that permits two devices to connect to each other, though I’m not personally aware of one. And, say you got a virtual private server for $10 a month or whatever the going rate is, yeah, that could be set up to do this – you could use it as an intermediate host, do SSH tunneling from both the PC and the phone of the sort that another user in this thread mentioned. I guess that that’d let you reach the PC from other places, if that’s something that you want to do, though it’s not the only way to accomplish that. But…I think that that’s most-likely going to add more complexity. The only scenario where that would truly be necessary is if the wireless access point — which I assume your ISP has provided — absolutely does not permit the LLM PC and the phone to communicate at all on the WiFi network, which I think is very unlikely, and even then, I’d probably just get a second wireless access point in that scenario, put the PC and the phone on it.
In general, I don’t think that trying to connect the two machines on your home network via a machine out on the Internet somewhere is a great idea. More moving parts, more things to break, and if you lose Internet connectivity, you lose the ability to have them talk to each other.
If you have an Android phone available, that would probably be easier from my standpoint, because I can replicate the environment; I have an Android phone available here. But it’s not really the phone where setup is the issue. Like, it’s going to be the LLM PC and potentially wireless access point that require any configuration changes to make ollama reachable from the phone; the phone doesn’t need anything other than Reins installed and having an endpoint set or just using a Web browser and using the correct URL there. I’m just mostly-interested in that the phone has to be able to talk to the PC, has to be able to open a TCP connection to the PC, and so having diagnostic tools on a phone is helpful. I don’t have to guess how the diagnostic tools work in Termux on an Android, because I can use them myself locally.
I wouldn’t suggest going out and buying an Android phone to do just that, though. I mean…this is a one-off diagnostic task, just trying to understand why the phone isn’t able to reach the LLM PC. If you can open a connection from the Android phone to the LLM PC, then you should also be able to open a connection from the iOS phone to the LLM PC. If you do have one already available, though, then yeah, my preference would be if you could install Termux on it for the diagnostic tools rather than install iSH on the iOS device. It should still be possible to get the LLM PC reachable on the iOS device either way.
I don’t mind trying to diagnose connectivity on the iOS device. Just keep in mind that I may have to guess a bit as to what the behavior is, because I can’t actually try the device here, so we may potentially have a few extra rounds of back-and-forth.
If you do want to use an Android phone, then just put the phone on the WiFi network, install Termux, open Termux, run the apk command to install telnet (
apk install telnet) and then try the telnet command I mentioned and just report back what error, if anything, you get when trying to open a connection to the LLM PC — hopefully it’ll be one of the above three outcomes.