The problem SupraTweets solves

What Can People Use SupraTweets For?

SupraTweets is an AI-powered Twitter bot currently tailored for the Supra Protocol, designed to revolutionize how protocols, dApps, services, and ecosystems engage with their communities. Leveraging cutting-edge AI technologies, including LLMs (Large Language Models) and RLNs (Reinforcement Learning Networks), SupraTweets automates content creation, user engagement, and analytics, ensuring seamless interaction with protocol users. From summarizing blogs and announcements to conducting user polls and staying updated with trending memes, SupraTweets covers all aspects of user engagement, providing a hassle-free solution for protocols to communicate effectively with their audiences.

1. Automated educational/announcement tweets: SupraTweets periodically posts curated threads summarizing architectural insights, technical updates, and major announcements about the Supra Protocol. By processing official resources such as blogs, whitepapers, and documentation, the bot transforms complex information into concise and engaging Twitter posts.
2. Periodic polls: The bot also conducts polls on a periodic basis, enabling the protocol to make data-driven and user-centric decisions. By understanding which features or services resonate most with the community—such as oracle networks, cross-chain solutions.
3. Keeping up with trendind memes: For the meme fans, the bot also keeps it updated with the trending memes and presents on the user's feeds when the content is relevant!

Architecture

1. Data Gathering Layer: Scrapes and prepares data to be fed to the models.
2. AI Layer: Summarises the tweets, generates memes, creates polls contents.
3. Scheduler: Schedules and manages posts and polls with cron jobs.
4. Feedback Analysis: Analyzes engagement metrics to guide future content.
5. Twitter API Integration: Posts threads, polls, and memes using Tweepy clients.

Challenges we ran into

Challenges Faced and How We Overcame Them

1. Twitter API Limitations : One of the main hurdles we encountered was that the Twitter API did not support many of the functionalities we needed for the bot to work efficiently in the free tier. A lot of the features like twitter replies, mentions could not be completed due to the search limitations of the free tier. Even bigger content tweets were not getting posted so had to workaorund to post the content as threads. Training and fine tuning the models over twitter data and trends too required listening to it periodically which couldn't be done because of limit reaches.
2. Finding the Right LLM for the Hackathon and Fine tuning them: Similarly, a lot of time was spent in looking into a suitable AI model for the features implemented. We had to try different free options available.

Future Scope

1. Search Queries and Reply Functionality: Currently, the developer Twitter API doesn't provide search queries. However, with upgraded plans, we can implement a search query feature to look for mentions of the bot and enable functionality to reply to tweets where the bot is tagged with a question and Supra is being discussed.
2. Models: To keep the bot relevant and context-aware, we plan to integrate fine-tuning, sentiment analysis, and feedback-based learning models.
3. Limitations of Free API: The developer API currently doesn't provide scraping functionality. This limitation makes training the bot on tweets relevant to crypto twitter trends difficult.
4. Adaptability Based on Engagement Metrics: The AI should adapt dynamically to engagement trends. For example: If educational threads or news updates generate higher engagement, the AI will prioritize and focus more on those types of content.