NYRA

Inspiration

The need for AI applications that work offline and protect user privacy while maintaining cloud-level intelligence inspired NYRA. Google's Gemini Nano enables on-device processing, eliminating the privacy concerns and connectivity requirements of cloud-only solutions.

What it does

NYRA provides six AI APIs (Prompt, Summarizer, Translator, Writer, Proofreader, Rewriter) plus SQL optimization, DevOps analysis and multi-agent orchestration. It processes requests on-device using Gemini Nano when possible and enhances with Gemini Pro for complex tasks, achieving 65% on-device operation rate.

How we built it

Built with Flask backend integrating Google Generative AI SDK, Firebase for authentication and storage, and BigQuery for analytics. Frontend uses vanilla JavaScript with Chart.js and D3.js for visualizations. Architecture implements hybrid routing that attempts on-device processing first before cloud escalation.

Challenges we ran into

Implementing seamless fallback between on-device and cloud processing required careful state management. Optimizing the decision logic to route requests appropriately based on complexity and resource availability proved complex. Managing API rate limits and costs across multiple Google services required extensive error handling.

Accomplishments that we're proud of

Achieved privacy score with 65% of operations running locally. Reduced average response time to 0.7 seconds and cut cloud costs by 78% compared to cloud-only approaches. Successfully integrated six different AI APIs into a unified interface with consistent error handling.

What we learned

Hybrid AI architectures require sophisticated routing logic to balance performance and capabilities. On-device models have limitations in context length and reasoning depth that necessitate cloud backup. Proper caching and state management are critical for maintaining performance across the hybrid boundary.

What's next for NYRA

Add voice input/output using Web Speech API, implement Chrome extension for contextual assistance, expand to mobile with Flutter, add collaborative features with real-time sync and develop custom fine-tuned models for specific enterprise use cases.

Built With

• ai
• api
• chart.js
• d3.js
• firebase
• flask
• gemini-nano
• genai
• generative
• google
• google-bigquery
• hybridai
• javascript
• llm
• promptcloud
• proofreader
• python
• rewriter
• sdk
• summarizer
• translator
• writer