The problem FridonAI solves

Firstly, analyzing cryptocurrency price charts often requires specialized knowledge and tools. FridonAI simplifies this process with intuitive features like CoinPriceAnalytics and TextToChart, allowing users to analyze and visualize coin performance using natural language commands.
Finding coins based on specific technical indicators or chart patterns has been cumbersome. With FridonSearch's TechnicalCoinSearcher, users can find coins through detailed textual descriptions. Additionally, the CoinChartSimilaritySearch discovers coins with similar historical or current chart patterns using pre-trained time series models, enhancing search precision and enabling more efficient identification of investment opportunities.
Missing trading opportunities due to lack of real-time notifications is another common issue. FridonNotifier enables users to set up custom notifications for specific indicators and scenarios, ensuring they stay informed of important market changes.
Executing blockchain operations often requires technical expertise, posing a barrier to many users. FridonBlockchain simplifies blockchain interactions through natural language commands, enabling users to perform operations like swapping tokens or checking balances effortlessly. Combined with FridonChat, which unifies all functionalities into a single chat interface, users enjoy a seamless and integrated experience.
The platform also addresses the limited community contribution in crypto tools. By open-sourcing the chat pipeline and providing the FridonAI-Core and FridonAI-Community packages, FridonAI encourages community contributions and makes it easier for developers to create and integrate new plugins and functionalities.
Lastly, FridonAI helps users overcome barriers to learning and utilizing DeFi protocols. Through educational and practical assistance in FridonBlockchain and FridonChat, users are guided through DeFi operations like staking and swapping via simple text commands.

Challenges I ran into

Throughout FridonAI’s development, we encountered significant challenges that required creative solutions:

Chat and Blockchain Communication:
One primary challenge was integrating chat and blockchain functionalities. The chat app, built in Python, processes user messages and plugins to generate responses. A NestJS backend manages user authentication, session storage, and message routing. For blockchain interactions, once a crypto request is made, a transaction is sent to the frontend for the user to sign. Afterward, we sign it and send it off, awaiting the result. To keep the chat app responsive, we implemented Redis and its queue system for asynchronous communication.

Dynamic Chatbot with Plugins:
Building a dynamic chatbot where plugins could be added or removed seamlessly required a modular architecture and robust testing to ensure plugins work consistently within the chatbot.

Data Handling and Processing:
Fetching data from providers like CoinPaprika, updating timeframes, and recalculating indicators demanded an efficient data pipeline. We optimized data retrieval and storage to handle large volumes without compromising performance.

Developer Contributions:
Encouraging open-source contributions required modularizing the codebase and setting clear contribution guidelines. This allowed for easy plugin integration without extensive reconfiguration.

Hosting and Optimizing Models:
Hosting the Coin Price Chart Similarity plugin on HuggingFace proved costly, so we switched to Replicate.com for faster performance at a lower cost. Optimizing this setup improved efficiency and scalability.

These challenges strengthened FridonAI and taught us valuable lessons as we continue to develop and innovate.