[0:00] One of the editors I use is Zed editor,
[0:02] and most of the time I use it together
[0:04] with local models. So today I want to
[0:06] show you how I connect it to local
[0:08] LLM's, both models running on the same
[0:10] computer as the editor, and models
[0:12] running on another computer in the local
[0:14] network that can be very convenient when
[0:16] you want to run models on a more
[0:17] powerful machine with a stronger GPU.
[0:20] And since I usually use Ollama and LM
[0:21] Studio to run models, I'll use those as
[0:23] examples today. But before that, I want
[0:26] to congratulate the Zed editor team on
[0:28] the release of version 1.0. I've been
[0:30] using this editor since the closed beta
[0:31] days, and the progress over the last few
[0:33] years has been truly impressive. So
[0:35] let's start with Ollama. First, I'll
[0:37] check in the console that Ollama is
[0:39] working properly, so if anything goes
[0:41] wrong later, I know the issue is not
[0:42] there. First, let's look at the list of
[0:44] models available on this computer. I
[0:46] already have several downloaded. Let me
[0:48] choose Gemma 4. Now I'll launch it and
[0:50] ask it to answer something, just to make
[0:51] sure it works correctly. Yes, the model
[0:54] loaded into memory successfully, thought
[0:55] for a moment, and then gave me a
[0:57] response. So everything looks good. Now
[0:59] I know that both Ollama and the selected
[1:01] model are working. Now let's go to Zed
[1:03] editor and open the agent panel. In the
[1:05] settings for this panel, there is a
[1:06] section for LLM providers. That's
[1:09] exactly what we need. There we can find
[1:11] an option called Ollama. By default, the
[1:13] local URL with the correct port is
[1:15] already filled in here. That is the
[1:17] default Ollama address, and if you
[1:19] didn't change anything when setting up
[1:20] Ollama, then you don't need to change
[1:22] anything here either. I will only
[1:23] increase the context window size because
[1:25] when the model uses tools, 8,000 tokens
[1:28] may be too little. I'll increase it to
[1:30] 60,000 for now. That should be enough to
[1:32] start with. Now we press the connect
[1:33] button so Zed editor can verify that
[1:35] Ollama is available at that URL and load
[1:38] the list of available models. As you can
[1:40] see, a green check mark appeared along
[1:42] with the word connected, which means Zed
[1:44] successfully connected to Ollama, and
[1:46] everything is working properly. All
[1:47] right, now let's close the settings
[1:49] window and look at the list of available
[1:51] models. We can see a new section called
[1:53] Ollama, and inside it are the models
[1:55] available to us. As you can see, they
[1:57] are the same models that were shown in
[1:58] the console earlier when we launched
[2:00] Ollama. Let me also choose the same
[2:02] model here that I used in the console,
[2:04] Gemma 4. And again, just to verify that
[2:06] everything is definitely working, I'll
[2:08] type something to it. It doesn't really
[2:09] matter what. The important thing is to
[2:11] confirm that I can interact with my
[2:13] local model directly from the editor.
[2:15] Okay, the model thought for a moment and
[2:17] then replied. Great, that means
[2:19] everything is configured and working
[2:20] exactly as it should. Let me also show
[2:22] you a couple of important things you
[2:24] should pay attention to. The first one
[2:25] is the agent profile. To simplify it a
[2:27] lot, this determines what the agent is
[2:29] allowed to do. There are three built-in
[2:31] profiles, and if you want the model to
[2:33] be able to edit and create files inside
[2:35] your project, choose the right profile.
[2:37] The second thing I want to point out is
[2:39] the thinking mode switch. If the LLM you
[2:42] selected supports reasoning before
[2:43] answering, you can enable or disable it
[2:45] here. It can be useful to experiment
[2:47] with turning it on or off, especially if
[2:49] the model responds very slowly and you
[2:51] want to speed things up a bit. All
[2:53] right, now that we've covered Ollama,
[2:55] let's do the same thing with LM Studio.
[2:57] Just like with Ollama, I'll first check
[2:58] in the console that LM Studio starts
[3:00] without any problems. I'll also launch a
[3:02] model to make sure everything is working
[3:04] properly there as well. Since I
[3:05] currently have only one model loaded in
[3:07] LM Studio, that's the one I'll use. Yes,
[3:10] everything works here too. The model
[3:11] responds to my requests. Okay, now let's
[3:14] go back to Zed editor and once again
[3:16] open the agent panel settings section
[3:18] with LLM providers. This time we look
[3:20] for the option called LM Studio. Here
[3:22] again, just like with Ollama, the
[3:24] correct default URL is already entered,
[3:26] so there is nothing to change. We simply
[3:28] click connect, and Zed editor checks
[3:30] that everything is working, connects to
[3:31] LM Studio, and retrieves the list of
[3:33] available models. As you can see,
[3:35] everything went successfully. The editor
[3:37] connected to LM Studio. Now let me open
[3:39] a new chat and then choose a new model
[3:41] for that chat provided by LM Studio. As
[3:44] you can see, a new section dedicated to
[3:45] LM Studio has appeared, and since I only
[3:48] have one model loaded in LM Studio,
[3:49] there is not much to choose from. So we
[3:51] simply select that one. Okay, this time
[3:53] I'll ask the model to do something
[3:55] useful, and as you can see, it
[3:56] successfully processed my request and
[3:58] wrote "Hello world" in several
[4:00] languages. Excellent. The combination of
[4:02] Zed editor and LM Studio works as well.
[4:05] So, as you can see, connecting local
[4:07] models to Zed editor is quite simple. It
[4:09] only takes a few clicks, and by the way,
[4:12] you can switch between them at any time.
[4:14] Now let me show you a slightly more
[4:15] advanced case, when the LLM's are
[4:17] running on another computer inside your
[4:19] local network. To do that, we go into
[4:21] the model settings. The simplest way
[4:23] would be to just change the URL for the
[4:25] existing Ollama and LM Studio entries.
[4:28] If you only plan to use models running
[4:30] on another computer, then that is
[4:31] exactly what you should do. But I want
[4:33] to be able to switch dynamically between
[4:35] smaller models running on my laptop and
[4:37] larger models running on my desktop
[4:39] machine. So I'll use a different
[4:40] approach. In the header of the LLM
[4:42] providers section, there is a button for
[4:44] adding a new provider. We click it and
[4:46] choose the option OpenAI compatible.
[4:49] Then we enter a name. In my case, I'll
[4:51] call it Ollama desktop, so I always know
[4:53] what these models are. And then I enter
[4:55] the address of the other computer where
[4:57] Ollama is running.
[4:58] In the API key field, you can enter any
[5:01] text. By default, Ollama does not
[5:03] require a key. After that, we save it.
[5:05] Oh, wait, it won't let me save without
[5:07] specifying the default model name. All
[5:09] right, let me enter one of the models I
[5:11] already have loaded on my desktop. Okay,
[5:13] now that everything is filled in and
[5:14] saved, we can finally test it. I'll type
[5:17] a test message in the chat.
[5:19] Oops, it doesn't work. Looks like I made
[5:21] a mistake somewhere. But that's fine,
[5:23] we'll figure it out. Let's close the
[5:25] agent panel for a moment and open the
[5:26] settings file so it's easier to see
[5:28] where I went wrong. By the way, this is
[5:30] a very convenient feature in Zed editor.
[5:32] Every change you make through the UI is
[5:34] reflected directly in the config files.
[5:37] All right, here is the section for the
[5:38] provider I just added. So what's wrong
[5:40] here? Ah, I see it. In the API URL
[5:43] field, when adding your own LLM
[5:45] provider, you need to specify the full
[5:47] path including the port number and the
[5:49] protocol version. All right, I'll save
[5:51] the settings and try again. Success. Now
[5:54] it works. I was able to access an LLM
[5:56] running on my desktop from the editor
[5:58] running on my laptop. Excellent. I'm not
[6:00] going to demonstrate connecting to LM
[6:02] Studio on the desktop because the
[6:04] process is identical to Ollama, only the
[6:06] URL is slightly different. So as you can
[6:08] see, using local models in Zed editor is
[6:11] very simple. And on top of that, you can
[6:13] configure everything in whatever way
[6:15] works best for you. That's exactly why I
[6:17] like Zed so much. That's all for today.
[6:19] If you enjoyed this video, give it a
[6:21] like and subscribe so you don't miss the
[6:23] next episodes. See you soon. Take care.