Arya | Devpost

Inspiration

Arya was inspired by the need to democratize AI-powered file processing through natural language interaction. The project addresses the growing demand for intuitive tools that can handle complex file operations without requiring technical expertise. The inspiration came from recognizing that users often struggle with multiple specialized software tools for different file processing tasks, leading to the vision of a unified platform where users can simply describe what they want to accomplish in plain English or any of 12+ Indian languages.

The multilingual focus was particularly inspired by India's linguistic diversity, aiming to make advanced AI tools accessible to users who prefer communicating in their native languages like Hindi, Gujarati, Tamil, Telugu, Bengali, and others. This aligns with my experience in AI development and building custom web interfaces with AI features using Sarvam AI & Google's generative AI tools.

What it does

Arya is a cutting-edge FastAPI application that combines AI intelligence with comprehensive file processing capabilities. The platform offers several core functionalities:

AI-Powered Intelligence: The system uses Google Gemini 2.0-Flash integration for advanced prompt parsing and function calling, along with Sarvam AI integration for multilingual speech processing. Users can describe tasks in natural language, and the AI automatically determines and executes the appropriate function.

Advanced Speech Capabilities: The platform provides speech-to-text conversion in 12+ Indian languages plus English, text-to-speech generation, automatic language detection, and real-time processing using Web Speech API.

Comprehensive File Processing: Arya handles image processing (compression, format conversion, PDF creation), document conversion (Word to PDF with customizable settings), archive management (extract and analyze ZIP files), and text operations (find and replace across multiple files).

Multilingual Support: The platform supports 100+ languages with complete Google Translate integration, focusing on Indian languages with native script display. This reflects my background in **web development using Node.js and experience with image processing and compression using JavaScript.

How we built it

The project was built using a modern tech stack centered around FastAPI for the backend architecture. The development process involved several key components:

Backend Architecture: The main application entry point is main.py, with a well-organized structure including AI service integrations (gemini_client.py for Google Gemini AI and sarvam_client.py for Sarvam AI services), core functionality modules for various file processing tasks, and comprehensive file management systems.

Frontend Development: The user interface features a modern drag-and-drop interface with visual feedback, dark/light mode switching, responsive design for desktop, tablet, and mobile, and real-time status updates. The frontend assets include style.css, script.js, and custom templates.

AI Integration: The system integrates multiple AI services including Google Gemini 2.0-Flash for intelligent prompt parsing and function calling, and Sarvam AI for multilingual speech processing. This integration allows for natural language processing and smart function detection.

Security Implementation: The platform implements local processing to ensure files aren't stored in the cloud, temporary storage with automatic cleanup, API key security through environment variables, and comprehensive file validation.

Challenges we ran into

Several significant challenges emerged during development:

Multilingual Processing Complexity: Implementing support for 12+ Indian languages with native script display required extensive testing and optimization. Handling different character encodings and ensuring accurate speech recognition across diverse linguistic patterns proved technically demanding.

AI Function Calling Accuracy: Developing a system that could accurately interpret natural language prompts and map them to specific file processing functions required extensive prompt engineering and testing with Google Gemini's API.

File Processing Optimization: Balancing processing speed with quality, especially for image compression and document conversion, required careful optimization. Managing memory usage for large file uploads while maintaining responsive user experience was particularly challenging.

Cross-Platform Compatibility: Ensuring the speech recognition and file processing features worked consistently across different browsers and operating systems required extensive testing and fallback implementations.

Accomplishments that we're proud of

Several key achievements stand out in the Arya project:

Seamless AI Integration: Successfully implementing natural language processing that can accurately interpret user intent and execute appropriate file operations represents a significant technical achievement. The system's ability to understand context and choose the right function automatically is particularly noteworthy.

Comprehensive Language Support: Achieving native support for 12+ Indian languages with proper script rendering and speech processing capabilities makes this platform uniquely accessible to diverse user groups.

Robust Architecture: Building a scalable FastAPI application with proper separation of concerns, comprehensive error handling, and security features shows strong software engineering practices.

What we learned

The development process provided valuable insights across multiple domains:

AI Integration Complexity: Working with multiple AI APIs (Google Gemini and Sarvam AI) taught us about the intricacies of prompt engineering, function calling, and managing API rate limits and responses.

Multilingual Development: Implementing support for diverse languages revealed the complexity of internationalization, character encoding, and cultural considerations in software design.

File Processing Optimization: We learned about balancing processing quality with performance, memory management for large files, and the importance of providing real-time feedback to users.

Security Best Practices: Implementing local file processing, secure API key management, and temporary file cleanup highlighted the importance of privacy-first design in AI applications.

What's next for Arya

The future roadmap for Arya includes several exciting developments:

Enhanced AI Capabilities: Plans include integrating more advanced AI models for better natural language understanding, expanding function calling capabilities to handle more complex workflows, and implementing machine learning for user preference learning.

Extended Language Support: While currently supporting 12+ Indian languages, the goal is to expand to all 22 official Indian languages and improve speech recognition accuracy for regional dialects.

Advanced File Processing: Future versions will include video processing capabilities, advanced OCR for document digitization, collaborative features for team file processing, and integration with cloud storage services.

Enterprise Features: Development plans include API access for enterprise integration, advanced user management and permissions, audit logging for compliance requirements, and custom function development frameworks.

Mobile Application: A dedicated mobile app is planned to complement the web interface, focusing on voice-first interactions and optimized mobile file processing workflows.