Please sign in to access this page
July 02, 2025
Okay went on with the frontend to test how it is working. and it is terrible the performance is terrible. when i was making the api i was just testing if it is wroking or not. not the performance and now that i am testing the performance, I have realized this is terribel right now. i am really frustrated since i spent so much time on the backend and it is still so bad.
Started serious work in frontedn
DOne with the auth setup wroking
now maybe i will add the chat setup workin
focus on ui or appearance later.
Amost all the routes that i need now are made and tested, there was an issue with first message of each session but i am pretty sure it is resolved now.
Added protection to my chat routes using dependency so only logged in user can use these routes. Also added schemas to solidify my api .
Done with the generate response and search l ogic just need to finish the chat route now, then i will secure these routes using a middleware, and then i will uhh,,, do something else i will think about it.
Done with the two authentication routes to enable login and logout using bcrypt and jwt, took much longer than usual since it is my first tiem working with postgres or any type of sql database.
Done addin more more vector embedding of law pdf-json, still one left but skipping for now since this is getting boring. I want a working chat bot fist then I will expand on teh data.
I am quite frustrated right now, I have to vote 88 more times and writing such long votes is taking more than 5 minutes each for me. I am thinking of quitting som, I will not get to the laptop anyway, and i need laptop more than anything. Anyways, I have added chromadb and sentence transformer model and i have completed the embedding for two most impotant jsons the bns and the constitution. i ahave also queried the chroma db to test and got quite good results. I will do the two more jsons and proceed to make the api.
Added two more act pdfs. Now i have come to a decision i will move on to next step skipping the rest of the pdfs for now since this is getting boring and tiring now. It is better to move to the next step now. Now I will make vector embeddings using sentence transformers model. Right now i am waiting for torch to download.
Yeah finally i did it , i gave up. I downloaded a json from github which has all constitution articles but it did not have the parts so i addded the parts to each articles and prepared tehm like teh bns json for embedding. Many more pdfs left but they are easy. None is as complex as the constituion.
This is 10 x harder than I though it would be the structure of the constiution pdf is so wierd i have just been able to extract the parts so far, now i am working on the individual articles. Hopefully it will take less time in this part since this is similar to section extraction in bhartiya nyaya sanhita.
I have now started working on this, i prepared data for bhartiya nyaya sanhita. Split the pdf into sections using regex. I am getting one extra section i don't know where but is ok i guess. ITs just one and anyway later it will all become embeddings. Maybe it is an issue but for now iam going on with it. I have to prepare many more pdfs now. This will take a while.
Ready to ship again now, i will not repeat the same mistake again this time. Last time, I did not add the project screenshots, or demo video. And it is kinda hard to test my project. So, this time I am attaching a demo video with this last devlog so it shows at the top and voters can see what i made.
I added batch processing to it now you can apply some changes to all images in one folder if you want. also now u can pick output dir from the explorer like ui just like for input file, also can mention path for output directory. And there is an menu in the tui to avoid cluttering. Added help/about part in the menu.
Just got the payout on first ship. Massive disappointment, but i believe this project deserve more and i will keep working on it. Worked on few bugs that were there, also added ui to pick output directory and add output path like input path. I think there was no screenshot of the file picker ui that is why i got less votes. So i added one now. hopefully, next one will be better for my project.
A website i will make where you can ask your queries regarding indian laws, and it will answer very accurately from the pdfs i have given to it like bhartiya nyaya sanhita, the constitution, etc.
Added two features
One pallete generation from image
And one to strip metadata from image
Also fixed jpeg not supported issue
jpg and jpeg is same so now treated as saem
Added to new search routes
to search regrets by message and title
one protected route to search in you own regrets
and one unprotected for all public regrets
Ran into some routing issues but fixed it
The regrets/:id was interfering with regrets/search
So i just defined regrets/search earlier
And yeah wrote a nice readme for the project all by myself
Deployed api on render now
Configured cors to allows docs to make requests and be interactive
Docs are now available, i have added the demo links
I have also added two more routes now to search for specific regrets
Now all that is left is to right documentation for those routes and write the readme
Added delete regrets route and the documentation for it. I am done with the eight requirement routes for endpointer but i have 1 and a half hour left for 8 hours so i will add some more maybe 2 more routes, write a nice readme deploy the api somewhere for free and submit.
Added a new put request to allow users to edit their regrets.
Only authenticated users can edit, and event they can edit only their own regrets obviously. Also added docs for this route as well. Using correct status codes everywhere now
Added two more routes one to view a specific regret by id
where u need to be logged in if u are viewing a private regret and only u can view ur private regret, no need to be logged in if public regret. and one more get request to view all of ur own regrets u posted. Documentation also done for all routes till now. Now i will work on put and delete routes.
Added two new routes and documentation for them
a /regrets post to post new regrets public or private
and a /regrets get to get all public regrets from database
Added swagger api docs for the two routes created
so far. also docs are accesable at /docs route.
Final Devlog before first ship.
I have finished it now. Added release in github for linux and windows.
Took some time to build for release but done with patchelf. Also fixed the error handling. Instead of panicking it now gives nice messages. Errors and logs are now better, colourful with colour crate. Wrote the readme for it with instructions. This is the first time I have writtern a README this big, if it is this big i use ai generally, but this I wrote myself
Added nice interactive ui using dialoguer
A file explorer like ui in terminal to select input image
Also fixed the bg_removal issue it always fixed to 320x320 now it resizes after applying mask
Also fixed the colour to white for bg when converting transparent bg to jpg
Now trying to build for release to see if it works. As there is a model involved, there is a high chance it won't.
Added one line run functionality
Just one line in terminal and vola
Used clap crate to do this
Will now improve the command line interface when not using args
Both options are there
Done with basic setup of api
two routes login and register done
Will keep adding docs simultaneously
fINALLY I DID IT! i tried so many ways to remove backround, i tried using a model with tract crate but that was a terrible idea , complete waste of time. But then i found the right way i used the model with onnxruntime which helped me remove the backround quite easily as well. 6 hours in, I have to do 9 more for terminal-craft. Next i can add image compression features.
Api for users to post their biggest regrets. Can post a public or a private regret. All users can view public requests. Authentication using jwt token. Logged in users can view,edit, post, delete their regrets. Get regrets by id is also availabe where anyone can get public regrets but only authenticated users can view private ones (only their own private regrets). Also there are search routes one to search all public regrets by name and message. And one to search all your regrets by name and message.
Spent a lot of time trying to remove backround when converting to png, what is a png without having a transparent backround. But i failed. Then i though ok lets use opencv, easy-peasy but it violates the rules of terminal-craft which say the app has to be self-sustained not rely on installing some other software, so have to find another way. There has to be another way. the only progrees since last devlog is that now the webp to png conversion. or vice-versa preserves transaparency, and also i am using clap and dailouger so now the cli ui ux is much better.
Simple conversion from png to jpg done. First project with rust. Excited to see how good the language is, there is a lot of hype about it.
A Terminal tool to convert images from one format to another and also remove backround, made in rust. This allows to ways one using the nice looking ineractive ui inside the terminal there is also a file explorer like selection ui for input image selection. And if u want to run in one line command u can use args.Now, there are more features like batch conversion and palette genration. The demo is release which has zip for windows tar.gz for linux just download, extract and run. The backroudnd removal is done using u2net model. U can do these image conversions in ur own pc no need to go to wierd websites just run from the terminal.
Done with the quiz app
Added images for brainrot characters
Find how much ur brain has rotten
Calulate how much brainrot u have consumed and how much knowledge u have on brainrot characters.
Fixed my routing issue
Working fine the demo now
preloading images for faster loading
Also h-screen sets hight to the height of entire screen on mobile
Which did not look good so i used h-[100dvh] which looks much better
it makes the app's height equal to the dynamic height
Added nice writing animation
Used vara.js
Deployed on vercel
Image optimisation, and some deployment issues left rest is done.
Added share feature(working locally but some issues right now deploying)
better prompt than before therfore much better result.
Done with all the work for my site.
Got nice images for the foods.
Good responsiveness.
Some nice transitions.
A sidebar for mobile.
I guess i am done with this
My personal favourite indian junk food site. A nice static webpage made for grub. With nice responsive sections and some sublte transitions. Very simple but looks good to me. Has quite a few sections The navbar , a mobile sidebar, the food cards, the main intro, the testimonials section, the faq section, a contact section, a footer. Made with just HTML, TailwindCSS, a little js.
ok added nice page ui for displaying the letter
nice cursive handwriting like font
very good looking responsive on mobile as well
Lets see what we can do
added firebase ai for ai genrated letter
this is really nice and easy
i was making the almost same backend for all my ai apps and then had to deploy on gcp everytime since i am just making apps for fun right now firebase is a more suited option i thing for these ai apps of mine.
I started making this then I reyi cannot hard code the letter messages of the user may vary drastically also how will I give them option to share I am not storing result i need to use some backend. Firebase maybe
A love letter generator site cause why not. Have not seen anything like this here yet, will be quiet unique i guess. Just enter ur name, for whom u are righting, what messages u want to have in the letter. And done nice love letter genrated in site which will be writtern stroke by stroke in a nice animation on the site itself in an old paper that looks pretty nice(see for urself).Ther is option to share the letter to whover u want to send that was generated by u and they can see it in site onlyl. Storing using firebase cloud firestore. And ai letter genration using firebase ai.
Added more solid error handling
Also fixed issue of single server duplication
And sometimes setup failing
Now uploading to mega for large videos
Was having trouble with compatibility with latest python
Had to swich to python 3.10
then it worked fine
Added two options to upload on cloud
gofile.io and file.io
One primary and one for backup
Now works for bigger videos as well
Added multiple server support
Sending messages now as embed on advice of friend
so someone doesn not annoyingly ping everyone everytime
embed buttons to choose server
Uploaded project to pypi
Now hopefully it gets certified
Last deglog before shipping
My first personal project that i will ship here(i have shipped some ysws)
Really took a lot of time but i am finally done
Deployed backend on gcp
frontend on vercel checked if it is working, It is
Hoping for the best with this one
Deployed frontend on vercel
Ready to deploy backend
Added delete modal to replace the alert
Almost done, will ship soon
Realistic loading
Added actual content and working links
Made it responsive
And most importantly added audio
background audio with switch to og audio feature
do try it
Also buildings have audio as well on clicking and moving
Made the box that comes after clickign building
actually like coc a bit as much as i could
And added coc actual font
added demo link now uploaded on vercel
the data inside buildings is placeholder for now which ai gave me
will add real data about me later
Added the coc layout
added main buildings
added villagers intro with animation
Also clicking buildings opens different secitons
Can drag buildings anywhere
Hovering buildings shows name and what they contain
A unique, interactive portfolio inspired by Clash of Clans (CoC), built with React and Vite. Move buildings, click to explore portfolio sections, and enjoy a playful UI reminiscent of the classic mobile game.
Added api to get suggested restaurants and hotels
nice cards to display in frontend
also fixed some frontend issues with location/destination confusion
used some random images for restaurant and hotel images in cards
Almost done now will ship soon i guess
Added Weathe Api
Fetches weather for the next days when you plan a holiday trip
For the next n days where n is no. of days of your trip
also new weather data is loaded when you view your saved trip from dashboard
Uploaded large video to cloud host
temporarily and generated a link to give in discord
switched to modern slash command
added starup guide for servers with help command
almost done
Basic bot setup complete
Works fine
Just run a command with url
And downloads the video and sends it in discord
Works for small file but not for large files
Need to work on that
TubeFetch is a Discord bot that allows users to download YouTube videos directly from their Discord servers. With simple commands, users can fetch and download videos without leaving Discord, making content sharing seamless and efficient.
My first discord bot
This one was fun to make
Using discord.py library
You can dm anything to the bot and it sends the message anonymously in the confessions group of your server
it also allows anonymous replies to those messags
easy setup and help commands
so anyone can use it
Whispr is a Discord bot that lets users send anonymous confessions via DM, which are posted in a public server channel. ✨ Features: DM-based confessions, auto 👍👎 reactions, mod queue (optional), anonymous replies, and spam filtering. No names. Just truth.
All features now working
Finished all routes
Added save and retrieve trip logic
Added toasts from react-toastify for better ui
Provided better navigation
Added error handling to give good looking messages even when there is error
Installed bs4 and rich
Used dev tools to get html and css info
Fetched trending GitHub developers and their data
Followed all guidelines
Python script to crawl and get trending developers for each language.
Added user authentication and database
Use ouath and jwt for authentication
Used mongodb for database
Added a dashboard to view saved trips
and therefore added feature to save trips, view them and delete them, store them in db
Frontend Setup done finally
Can finally call my app a full stack app now
Fully Responsive Simple and Working
Added actual api calls to my backend using axios
Proper error handling and formatting of the response
Experimented with different llms to check which one fits best
Setup My backend api routes, added hugging face models using hugging face - hub. Designed a result to format to keep the result of the ai consistent. Designed a prompt so that ai sends result in the specified format. Used hugging face hub inference client. Tried using text-generation, but had to shift to conversational due to almost no model supporting text-genration. Basic backend setup complete.
Holiday Planner is a full-stack AI-powered travel planning application that helps users effortlessly plan holidays, generate personalized itineraries, and discover destination suggestions tailored to their preferences and budget. The platform integrates real-time weather forecasts, accommodation and restaurant recommendations, and provides a seamless dashboard for managing all your trips. Built with FastAPI (Python) for the backend and React + Vite for the frontend, it offers a modern, responsive, and user-friendly experience.
This was widely regarded as a great move by everyone.