final steps!

I am almost finished and i just need to deploy it to the vps and then we are done. What i did: I worked on making the langchain implementation and i was working with different AI frameworks like CrewAI but setteled for langchain.

Tools the agent has access to:

• Fetch porject stats
• fetch user stats
• get shop items
• look at achivements
• see devlogs

Hopefully this is enough to stalk anyone

I played around with the FT API!

So i tried testing out the different features of the falvortown api and I am implementing them as tools right now. So basically to make the agent I have decided to go forward with langchain, because it’s the best agent framework for python afaik

It works!!

It can now respond upon mention and answer sarcastically. Why is the Slack API so complicated
Anyways so now it subscribes to the realtime event and then can respond whenevr you type @mico into slack and the message behind that. My next goal is going to be to give it tools, like a meme database and so on

UHH - lazy

so i just added env’s and made github repo and put all the changes of now there

Changelog

• Uhh first intial commit, setup a simple AI chat request and also made my tast env’s secure!

Initial Setup and first steps

So in this session I worked on fixing my vim a bit up and working on writing a simple chat request to HCAI via the OpenAI Python SDK. Here is an example response for the prompt: What is a good Techstack for the year 2001?

Forget the site - just read the readme

Ok I am too lazy to implement the site so if this project gets a lot of cookies then i am going to make a site and also improve this! For now you can use the readme as documentation. I am very proud of this not gonna lie

Use works now!

Now you can easily switch between packages (e.g from gh 2.87.2 to 2.87.3 and so on) also cleanup wasn’t cleaning up the left over folders properly so I fixed that and also now I am ready to ship the first release, finally write the README and also make the github repository public. After that I wil start writing the website

Definetly not finished

After thourghly testing I noticed that a lot of stuff breaks sometimes. So here is what I have done in the last 3 hours:

1. Added a manifest generator: Simple thing but it makes my life easier then just rawdogging json everytime.
2. Made cleanup remove empty directories: The thing was cleanup removes dead paths but now it also removes empty directories
3. Now “use” kinda works: so basically you can switch package versions now and then but like I am not sure and it hasn’t been tested very well yet

Almost finished?

I implemented tons of features like removing, listing and cleanup. The main problem I was encountering was that the handler was selecting the first version at the top instead of the one for the OS itself. also I somehow broke the SHA256 checker so I also fixed that .
I will finish up the remaining parts and edge cases (e.g Registry corruption) and then I will release the right binaries for the windows and linux systems (idk if it will work on macos). After that I will start making the website for blaze with documentation and troubleshooting and stuff yk

Adding functionality!

It can actually add stuff now! I mean real stuff which get added to the enviroment variables once it’s downloaded. I have also implemented remove and listing and switching which are a bit wonky tbh and I also have to work on edge case.
So the plan for now is to implement better command support and handle dge cases, then package it and work on docs! Below u can see the add implementation

I am switching over to Go because Zig is getting to complicated for me. Also cross compilation Go is very easy, you don’t even need an extra laptop with macos or a vm with linux to get binaries for both platforms. Here is an Image showing how my tests are passing! so this should work now. As of know I am still working on making the add feature actually good, like removing failed installs and so on. so basically i will probably work on getting an mvp like removing, resolving version conflicts and stuff i mentioned in earlier devlogs. Also I wil prob start to like devlog more often

WIP!: Implementing add function!

So right now I am defining the JSON manifest on how it should look like! so basically the add function should parse all the input fields and look at the json for invalid data and so on so it can “sanatize it for the next step. The Manifest itself should contain direct “ZIP” links for diffrerent OSes yk? I think the whole JSON should contain the following:
Name, Version, Decription, Tags, Author, Code Repo (if existent), respective links to different OS packages, SHA256, and other stuff that like tells where the bin is or smthing. Below you will see the blaze add command will look like

Feature Planning and techstack

So I layed out the core concept of the whole thing and also worked on how the system would work (will be changed in the future).
So basically you have the package manager itself. It will be written in Zig for speed and safety and because I don’t trust C. Also most of the C libraries will work fine in Zig (yes zig is c compatible!). So the idea is that the User will type in “blaze add https://coolpkg.com/coolpkg.json” and then the package manager will read the json to like learn about the package and then set it up for the system. After that it will add that to the enviroment variables. lets dive a bit deeper:

Adding packages

you just add it with: “blaze add https://coolpkg.com/coolpkg.json”. The cool thing about this json will be something that points to the portable version of the app and this means that it won’t require admin rights such as MSI’s mostly do. Also we will be prob implementing SHA256 verification and also raise flags for non http downloads.

Dependency resolution

I am thinking more of like an thing where we basically copy nvm-style where you can specify:
“blaze use [email protected]” or something like that so we basically install a copy of the thing yk?

Removing

This is pretty straightforward ngl, you just “blaze remove node” and boom! If you have multiple stuff it says: “You have multiple versions of ! Please specify via appname@version or use –all to remove all!” and problem solved.

Storage.

Nothing fancy. It’s all stored in a json, because i feel and sqlite database is overkill for something like this.

Size

Size matters. Since I am using Zig i am not worried about size

Plans

Right now I will prob setup the Zig project and also get on with creating cli args

Typemusic - music as you type

It now fully works. was a very short project since github copilot was a lot of help. Anyways you input the file (WAV format!) and then it plays the next slice of the music on every keyboard stroke. this was a very fun project to build. Hope you like it

It now automatically deletes the input file and also the unlocked files now have a “_unlocked” text attached to the unlocked files

Finished

I have finally finished my first serious Go Project. What i have accomplished

• AES-256-GCM Encryption: Encryption with GCM mode for integrity verification
• Argon2 Key Derivation: Secure password-based key derivation resistant to GPU/ASIC attacks
• Self-Contained .locker Format: All metadata needed for decryption is stored in the archive
• Streaming Encryption: Efficient file processing with minimal memory footprint
• Cross-Platform: Full support for Windows, macOS (um well not tested but probably), and Linux (may break depending on OS i think)

I have also create a Github releases with the respective binaries so try them out: Github releases page!

AES-256-GCM encryption with Argon2 key derivation

• Added password-based key derivation using Argon2
• Implemented AES-256-GCM encryption for data and streams
• Included salt generation and serialization
• Supports both in-memory and streaming encryption modes
  Well i can’t show an image yet of it working because have too intergrate the functionality into the tool so here is an image of me running the code (it win’t work because package command-line-arguments is not a main package)

I worked on learning Encryption in Go. So basically as of now i have the basic skeleton of the cli and I am working on the file encryption and decryption so it can lock and unlock files. Below is an image of the basic skeleton and yeah I know that I need to fix the tabbing

So i fixed some bugs (it only worked on my User directory) and now I am going to ask a friend to review it and yeah.

Finished!

I have finished the project with one last tweak: It now asks the path of the music at the beginning! I wasn’t intressted in making copy-paste work but who cares? Anyways it looks really awesome i think and I just need to test it out on an other maschine to see if it works!

Particles!

I added particles and color bars to the programm. depneding one the frequency the bars will change their color to red, green or blue. The particles will show the peak points of the music. Below is a short demo video

It now renders the bars of the music

It can now output the frequency data. I am occuring a lot of problems because the docs and tutorials are tailored to sfml 2 and prior sub-versions but since SFML 3 (which i am using) the only reliable documetnation is the official one and that has a a very bad experience. Below it outputs the data of the 1st sample (music only started so it’s 0)

I have finally succedded in creating a window! I was struggling with compiler issues and cmak ebut now it finally created a window. This is my 1st c++ project so i am still figuring out things

I had messed up auto injection so fixed that and also turned the disable button blue and added icons

I wrote a simple html UI to change the terms and disable the functionality. will soon publish it to chrome webstore

It now auto injects the images of the terms you provide

Removed the dashboard and am ready to ship the first version. PS: it now has dynamic routes!!!

The domain and stuff relating to the domain like ratelimits and more are now in the sqlite database instead of a config.toml. This will make it easier for the planned dashboard to manage stuff. I think maybe I will run a seperate API for managing this stuff instead of directly putting this into main.rs

Made a stats api to get information about the current configuration and details about the proxy. I will use this to control the proxy via an SvelteKit Frontend

I added rate limiting. Seems awfully lot of time but I kept encountering a bug so add the end I decided to just use AI assistance. Sorry fellow anti-vibecoders.

I implemented caching. This will reduce load on the origin servers

I finally got SQL to work. It now successfully stores the logs in proxynet.db. Here is a preview of how it looks (using SQLite DB Viewer)

I am working on storing the logs and the config in an SQLite Database. My verdict is that I like SQL, but SQL doesn’t like me. Same goes for Rust

I added logging. It now checks how long the response time was, which IP and which User-Agent type requested the information. Now I have to put all of the collected information in an SQLite DB and see how to check how many bytes I sent. Below is an example Log when i run:
curl.exe -H “Host: test.local” http://localhost:8080

Ok added support to read of config files. so basically at the root of the project you have to create an config.toml and write stuff like the domain you want to proxy and which port and host the server should run on

Ok I build a minimal working version of the proxy. It now maps the domain to the origin.

I have decided to stick with rust and use svelte for the frontend dashboard since python has poor async support and will consume more CPU. I have to lock in and get better at rust ig.

I explored different stuff cloudflare provides and also decided to use python. I intially planned on using Rust, but I feel more comfortable with Python

I soldered everything an finshed writing the code. It displays and HTML page while it is running by connecting to the local internet.

I finished the App! I implemented the backend and worked on proper redirecting. Also analytics now shows where the app came from, which browser was used and which OS. also how many times the link was visited!

created a simple analytics UI. See the image below

now I have got a good looking and minimal UI. It’s dark mode forever (light mode users go delete yourself). Now I will focus on other pages like the analytics page, then I will work on the backend!

Built an basic landing page after struggling with NextJS. Looks ugly rn, but I promise it will get better

After fooling around a ton with UI and breaking my app about 20 times, I have finally decided to let clanker ai cook a bit. Also I improved the prompt that’s give to the AI, so that it doesn’t ONLY spam insults. I think I am finished and start deploying to prod!

I got a basic UI ready! I spent like 20mins on this thing and then it wasn’t working…I see the logs. Turns out I was having an server error and nothing was wrong with frontend. Now I have to make it more prettierrrrrr

I started a new project! As of now I finished writing the endpoint for the API and now I need to work on the frontend! I decided on using Grok 4.1 fast, because well…it’s kinda unfiltered. Also I am using a tiny mini function provided by Appwrite, which allows me to fetch screenshots of sites!

Ok I have finished this project!!! Now the AI can use the context of given images to edit them! Please check it out at: https://imago.ultimatedev.space

I finally got the thing to work! My project was also delayed, because Hackclub AI wasn’t working, but now it’s back up and running! I have got multiple styles and you can now also download the image by clicking on download, after generating the image! Now I need to get “Attach reference” to work and actually send the image with the user prompt.

I made the theme to a catpuccin theme (it just looks too good). What I think is that the I should fix up the buttons, change the UI to reflect a yk more un-ai like vibe

Started this awesome project to learn Svelte and SvelteKit. I got bit of the starting point ready. It’s not much but I have yet to cook. So let me cook.