WeAware

This is a really cool project and I’m sorry I didn’t get to cover more of its development through devlogs.
There are podcasts being generated, I’ve covered everything in the NoteFromDev.md (https://github.com/Shreeyans2305/WeAwareNews/blob/master/NoteFromDev.md) file in the repository if you have time, please do check it out. I also have a dating app-like swipe left/right for stories you like or don’t like though instead of just providing the headline (unlike other apps) you get in depth information about the news you’re swiping on. I still haven’t been able to figure out the implementation on Sherlock mode, though it is a really cool feature and I hope it is available soon.

I’m really sorry, since completeing this project has had my complete attention I might have lost sight of writing devlogs. But finally WeAware has been completed. It is a completely autonomous news media channel, instagram page, app and website I will post the links in the following devlog do check it out. I am gonna be missing out on 14 hours smhh. The website features a feed with the latest news, its tts, then the podcast lets you listen to the main events that happened yesterday. The goal of this project has been to maximize coverage so all the news that you see has been corroborated through a tonn of sources. New news is gathered everyday now, and clustering happens through a complex Jaccard model and semantic similarity. I lovee what I’ve ended up with.

New podcasts covering the events of the preceeding day are genereated automatically on schedule through Google GCP’s Vertex AI service, the articles too are written using Gemini Models, please sign in to check it out for yourself. The Sherlock mode is experimental but the idea behind it is an agent acting as your private investigator to find out more stuff about any happening you might be interested in. Please do check it out!!

So I finally got the API working. Now Instagram posts can be automated to post the latest news with the most sources which has an image url in one of the sources (Insta @https://www.instagram.com/weawarenews/). This was a tedious process. Only authenticated requests go through and they are authenticated through a header. Since there’s only one team that is going to be using the automation at once, it makes sense and I dont need to generate mutliple API keys. The following are the endpoints:

1. GET: This returns the corroborated story ID, number of sources for that particular story, and the articles.
2. POST: This is used to mark a story as posted so this will help in avoiding duplicate posts!!

Thank you so much for following along!! Hope you have a great day ahead ^^

Ok, so I finally got everything running on Google Cloud Platform. Also trying to run an instagram page (https://www.instagram.com/weawarenews/) where we automate news stories. So the workflow right now is that on my computer news stories are fetched and aggregated from different sources such as bluesky, rss feeds, etc. Once fetched, similar stories are grouped together locally (I want this part to run in the cloud as well). After they are aggregated locally, finally I manually upload them to google cloud. The idea behind this is that an API can be created which will return the latest corroborated news stories with the highest source count. This API we can connect to our website, socials to make an automated media pipeline. I’m sorry I’m not doing the best job explaining it. The contents are a little dense right now, but I hope as I start building up the other parts like the design and frontend, it’ll get easier to understand visually!!
Thank you so much for your time :))

The beginning of WeAware News!!
The goal is to try to reduce bias in the media which has complete control of the sentiment of the populous. Inspired by Ground News. Right now, I’ve implemented a large pipeline for data aggregation. Data is collected from 5 different News APIs, RSS Feeds, Bluesky. If you have any more suggestions for sources please let me know :)) Same stories from different media outlets will be clustered together and stores in a database. The next stage which I plan to implement is the actual multi-agent logic which will check for bias, source credibility, and do its best to provide nothing but the facts. I might have to fine tune an open source model to achieve this but as of now, I just plan to use system prompt to make whichever model I choose sound as neutral as possible. Thank you for reading this :))
This is the first large project I’m working on, any suggestions are welcome!!