YES Ai: Smart Assistant with Real-Time Tools & Deep Research

In a world full of fragmented AI tools, YES Ai unifies everything — deep research, real-time data, and multilingual conversations — into one secure, intelligent assistant powered by Google Gemini.

Comment

💡 Inspiration: A Unified AI Assistant

Our goal was to move beyond single-purpose apps and create a unified, secure, multi-functional AI agent that truly acts as a personal assistant for the modern web. YES Ai was inspired by the need to consolidate:

  • Real-time information access (Search & Weather)
  • Complex problem-solving (Math & Research)
  • Secure user management (OTP Login)

All seamlessly orchestrated by Google gemini-2.0-flash to demonstrate its full capabilities for the Cloud Run Hackathon in a lightweight, containerized architecture ready for serverless deployment.

✨ Core Features: What YES Ai Delivers

Feature Technical Insight
🔍 Deep Research Mode Fetches and synthesizes multi-source information using external search APIs.
🌐 Multilingual Power Supports fluid conversation in English, Bengali, and Hindi, showcasing the model's language resilience.
☁️ Real-Time Tooling Simultaneously calls external APIs for Weather updates and internal tools for Math from a single user query.
🔒 Secure Authentication Implements a custom, production-ready login system with Email OTP verification and secure bcrypt password hashing.

🚧 Challenges, Learnings, & Future Vision

☁️ Deployment & Performance (Cloud Run Alignment)

The application's architecture is built on a modular, container-ready Python/Streamlit base. This ensures that the entire agent can be easily containerized and deployed as a Serverless Container on Google Cloud Run, guaranteeing auto-scaling on demand, cost efficiency, and high availability—all critical benefits for a production-grade Cloud Run service.

🧠 Key Learnings & Technical Execution

  1. Tool Orchestration Mastery: Learned to structure system prompts to ensure Gemini 2.0 Flash reliably and accurately selects multiple tools (Math, Weather, Search) based on a single, complex user input.
  2. Security Implementation: Hands-on experience implementing robust authentication using bcrypt and OTP—a critical learning for deploying scalable web applications.
  3. Deployment Resilience: Learned secure deployment best practices on Streamlit Cloud, utilizing secrets.toml to hide API keys from the public repository.

🚀 Future Scope: The Hybrid V2 Vision

Our biggest challenge was building the current architecture modularly enough for a future upgrade. This vision showcases our understanding of sustainable, production-grade AI:

Area Planned Upgrade Technical Impact
Model Architecture Hybrid Model Integration (Gemini + Llama 3) Optimized cost-effectiveness and specialized custom response tones.
Accessibility 🔊 Voice Commands & TTS Implementing Speech-to-Text for input and Text-to-Speech for output.
Generative Tools 🖼️ Photo Generation Integration Adding a feature to generate photorealistic images directly within the chat.

This commitment to a Hybrid V2 vision demonstrates an understanding of production-grade, sustainable AI development. Specifically, future integration of Elasticsearch is planned to enhance the Deep Research Mode with highly scalable, contextual, and internal knowledge base search capabilities, aligning with the Elastic Partner Challenge objectives.

🔗 Built With

  • AI Core: Google gemini-2.0-flash
  • Frontend/Backend: Python / Streamlit
  • Authentication: Custom Login System (bcrypt, Email OTP)
  • **Container Ready Base: Python / Streamlit (Designed for Google Cloud Run)

Built With

Share this project:

Updates