ContextLy.AI: No RAG, No Lag – Gemini x MCP

This content originally appeared on DEV Community and was authored by Snehanshu Raj

ContextLy.AI – A “No RAG, No Lag” Multimodal Intelligence AI

This is a submission for the Google AI Studio Multimodal Challenge

Ever wondered what happens when you remove the complexity of traditional RAG systems while keeping all the power?

Meet ContextLy.AI – a revolutionary context-aware AI assistant that transforms how we interact with our digital content.

What I Built

ContextLy.AI is a “No RAG – No Lag” application that eliminates the traditional barriers between users and their content. Instead of wrestling with complex vector databases, embeddings, and inference models, users can:
Simply upload PDFs, audio files, or URLs and immediately start having intelligent conversations about their content.

This isn’t just another chatbot but it’s a demonstration of:

• Gemini Flash’s revolutionary multimodal capabilities.
• Seamlessly fused with its Live API – Tool calling SDK integration.
• Combined with the Agentic God: Model Context Protocol (MCP).

While Gemini Flash natively processes text, images, audio, and video simultaneously, its Live API calling capabilities enable real-time database queries, API integrations, and dynamic workflows.

The MCP integration amplifies this power exponentially, allowing the AI to orchestrate complex multi-tool operations while maintaining context across all modalities. The result?

An incredibly intuitive experience that feels like magic but harnesses cutting-edge technology where advanced reasoning meets actionable intelligence!

Demo

Privacy-First Design: You need to have a Gemini API Key (which lives only in browser memory), and none of your data will be saved. It is a session based app running and has built-in warning that protects against accidental data loss upon refresh.

Note: ContextLy.AI is currently optimized for desktop use.
If you’re accessing it on mobile, please switch to “Desktop Site” mode in your browser for the best experience.

How to Use ContextLy.AI:

1. Open the App: Visit this URL to launch ContextLy.AI
2. Enter Your Gemini API Key: Input your key to start. Don’t worry, it’s safe: no data is saved. Without the key, none of the features will work.
3. Key Validation: The app instantly validates your API key. If it’s invalid, an error message will appear.
4. Upload Your Contents: Add PDFs, URLs, or audio files. Once uploaded, they appear immediately in your library.
5. Start Chatting: Switch to the chat tab and explore your content instantly. Ask questions, get summaries, or analyze your files without any delay.
6. Session Privacy: When you leave the session or reload the page, all uploaded data is cleared, and a fresh session starts.

Try it out: Live API.

• Watch a quick Video Demo.

Step-by-Step Screenshots

Step 1: Open the app

Step 2: Enter Gemini API Key, it will get validated instantly

Step 3: You can see two tabs, they will be functional only upon Key validation

Step 4: Upload All your Content in Upload tab

Step 5: You can see them in Library View

Step 6: Start Chatting instantly & Experience Multimodal Cross-Content Insights

Step 7: Convenient Chat Options: Save chat or Start Fresh Conversation

Multimodal Features

Seamless Content Understanding:

• The multimodal capabilities transform how users interact with their information.
• You can upload three different types of resources like a research paper PDF, related audio lectures, and supporting web article links – then ask questions that span all three sources.
• Gemini Flash processes each format natively, understanding context and relationships that would be lost in traditional systems.

Dynamic Content Navigation:

• Through MCP and Tool calling SDK integration, the assistant can dynamically navigate through uploaded content during conversations.
• Ask “Is there any points missing in my Resume which was told in the tips?” and watch as the AI uses tool calls to locate, extract, and reference the exact audio segment and compare it with Resume content.

Cross-Modal Analysis:

• The most impressive feature is cross-modal reasoning. Users can ask questions like “How does the conclusion in this PDF relate to what was discussed in the uploaded audio?”
• The AI seamlessly correlates information across different media types, providing insights that would require manual cross-referencing in traditional systems.

Privacy & Cloud Deployment: ContextLy.AI Privacy Design

How privacy is ensured on Cloud Run:

• Stateless Container Design: Each container instance is configured with concurrency = 1, meaning no two sessions ever share the same container instance. This guarantees that session data from one user never interferes with another.
• Ephemeral Memory Storage: On every page load or refresh, all uploaded content and session data are cleared from memory. Nothing is forever there on disk (it remains there only untill session is active), ensuring no persistent storage leaks sensitive information.
• API Key Safety: Your Gemini API key is stored only in memory during the session and is never written to logs, files, or any external system. This ensures secure access while using the app.
• Public Accessibility Without Compromising Security: Since the challenge asked the app to be deployed on Cloud Run this design was adopted. Privacy and session isolation are rigorously maintained through stateless container design and session-level memory isolation.

How I Used Google AI Studio

My development journey with ContextLy.AI truly showcased the full potential of Google AI Studio as a multimodal development environment. I leveraged the platform in multiple ways:

• Rapid Prototyping & Debugging: Used AI Studio extensively for testing prompts, experimenting with different model configurations, and fine-tuning responses directly in the integrated chat interface.
• Image Analysis Superpowers: Uploaded mockups, wireframes, and application frontend hand drawn designs to get intelligent feedback on:
  • UI design decisions
  • Visual hierarchy and layout improvements
• Multimodal Questioning: The app seamlessly worked with text, PDF, and audio, making development and product feel more interactive and holistic.

This powerful workflow enabled me to:

• Build a solid Python backend for document processing, make Gemini calls integrated with a MCP server.
• Create a creative frontend using HTML + CSS + JavaScript, with fully stateless capabilities and session management logic.
• Get step-by-step guidance to deploy on Cloud Run, making the whole project production-ready.

In short, Google AI Studio wasn’t just a tool, but it was my AI-buddy, helping me brainstorm, debug, design, and deploy ContextLy.AI.

Next Steps (Persistent Version):

I will build a stateful, locally runnable version of ContextLy.AI. This version will allow:

• Persistent uploads and libraries stored locally.
• Richer upload capabilities, supporting larger files and multiple formats.
• Advanced chat features, including history, cross-content comparisons, and threaded conversation tracking.

This version will ensure that users can experience the full ContextLy.AI power both instantly on Cloud Run and eventually in a fully persistent local environment (without the need to worry about data loss).

Thanks for reading and trying!

This content originally appeared on DEV Community and was authored by Snehanshu Raj