Poolside Laguna M.1 API - Free AI Coding Model

本站添加: 2026-05-17

[00:00:00]So, Kilocode basically runs some of the free models.

[00:00:05]Those models appear in their leaderboard.

[00:00:08]But, they are usually offered by the model makers themselves. Right?

[00:00:14]And they have listing on Kilocode like this. So, in Inclusion AI, they have like two models.

[00:00:25]Then comes Poolside, Tencent, Step one, and so on. So, I wanted to check the Poolside, which has this Laguna M1.

[00:00:36]And there is also one more model, which you may find in their official page showing this XY X S2. Right? So, these two models are out from Poolside. So, Laguna M1 is what I am attempting to test right now.

[00:01:00]To do that, I can open Kilocode from command line.

[00:01:05]Alternatively, I can also use VS code.

[00:01:08]So, whichever option you prefer, choose that. Right?

[00:01:14]There is also the option to sign up to their official site.

[00:01:19]You can see this is Poolside AI, and they do have API that you can access. Right? So, if you go to models, and if you scroll to the bottom, here you'll see the get API key option.

[00:01:35]You may require to basically click on their sign up, and then choose between GitHub or Google. Right? So, here you have to authorize.

[00:01:50]Right?

[00:01:50]I will go back to our Kilocode, and here you can see I can basically click on the model drop down, and here I have to select the pool side Laguna.

[00:02:05]Now, what I can do is I can pick a project, and then we can make it go through certain prompts to build or maybe even get a code review.

[00:02:19]So, one of the project that I launched into is basically an Android app, and it's built with Expo framework. So, I will let run this particular prompt to get the code review, right?

[00:02:34]So, to get reviews, you can either use planning mode or code or ask mode. I don't think orchestrate or debug mode should be used. So, I will go with the code mode.

[00:02:47]So, we can explain to the model that I want to create a code review for this app, and check if it's ready to be And also, you have to explain the checks that you want to do with the application, right?

[00:03:09]So, let's see.

[00:03:12]Also, check if the user workflow for the app is good enough for Google Play.

[00:03:25]Right?

[00:03:27]So, we will basically get the code review, but for that, it has to go through entire files and do the check. So, the model is free.

[00:03:38]You can see it has 131.1 K as its context.

[00:03:45]It will read entire application, and then it will generate a code review. We can save the code review in let's say HTML file or maybe markdown depending on the type of output you wish, right?

[00:04:04]So, let's wait for it to initial summary.

[00:04:08]Right. So, at the end you will notice that it will come up with the summary on things that are good and things [clears throat] that need improvement, >> [laughter] >> right?

[00:04:20]So, we can see save that data stored somewhere and next iteration we can actually use it to improve the app.

[00:04:31]Right.

[00:04:32]So, this is the Laguna M dot one model from Woodside and it's specifically designed for code.

[00:04:42]So, you can actually use this for improving any of your existing projects. So, I just used this on Expo app.

[00:04:52]I consider using it to update this existing app. Also, you can see the context is also decent.

[00:05:03]Save the file and I may even show you within our file explorer. And there as you can see it has stored this particular file and it has also given warning that this is not ready for Google Play launch due to missing privacy policy, which is fine.

[00:05:23]We can check the review. Here you can see code check and it has explained the things we lack and also it has mentioned the fix.

[00:05:36]Now, the context is with respect to Google Play Store, right?

[00:05:41]Now, following fixes are there like privacy policy, setting screen, valid support channel, improved navigation flow. So, these are some of the updates we can make below. So, the Laguna model seems to be decent in terms of getting the HD code to work.