Inbox Angel

• 

Inspiration

Like most of us, we’ve all felt lost in our inbox at the absolute worst moments:

After a hospital visit, I’d get a flurry of follow-ups, appointment reminders, and hard-to-understand after-care instructions. When something really mattered like finding feedback or medical reports from my doctor I’d end up scrolling endlessly, only to feel even more stressed and anxious.

There were times a deadline crept up like a bill due today, a school form my younger sibling needed signed, or a meeting suddenly appearing in a wall of unread notifications. The pain point was always the same: with so many emails, the important stuff slipped through the cracks.

That’s where our journey started. Inbox Angel is for anyone who knows that gut-churning feeling of missing something critical in their life's digital flood. We wanted an assistant that would:

Instantly answer questions like “Do I have anything due today?” or “What did the doctor just say?”

Bring calm, clarity, and maybe even a little comfort to your inbox, whether you’re a school kid, college student, busy parent or grandparent.

What it does

Inbox Angel is an email platform with an integrated AI assistant that can answer any question you have about your emails instantly, conversationally, and with direct citations.

Our custom email database mimics a real-world inbox, populated with threaded conversations, messages from schools, family, doctors, banks, and more.

Instead of users having to manually scroll, open, and interpret each email, our built-in assistant uses state-of-the-art natural language understanding to:

Parse and understand all emails and conversation threads.

Let users ask anything—like “What’s due today?”, “Did my doctor reply?”, or “Are there any important deadlines coming up?”

Provide precise, on-demand answers with references to the exact emails, saving users from digital overload and missed information.

Example: A student receives dozens of emails across classes about assignments, results, and reminders. They can simply ask, “What homeworks and assignments are due today?” Inbox Angel’s assistant surfaces the deadlines and even points back to the original mails all in seconds.

Whether you’re a student, adult juggling bills, or a senior navigating medical care, Inbox Angel is your always-on, always-calm, question-answering guide to your email life.

How we built it

We architected Inbox Angel with a focus on simplicity for users and maximum flexibility for innovation. Here’s how all the pieces came together:

Frontend: Our frontend is built in TypeScript using a modern React framework, providing a snappy, intuitive mail client interface designed for all ages and tech backgrounds.

Email Database: To rapidly iterate and showcase real-world utility, we designed and populated a custom database with dummy email threads — covering everything from doctor follow-ups to school assignments, bills, work updates, and personal notices. This allowed us to model real-life pain points across age groups and scenarios.

Inference Engine—Groq: We integrated the Groq platform as the core language model engine, harnessing its industry-leading inferencing speed and reliability. Groq enables Inbox Angel to understand user queries in natural language, no keywords, no rigid forms and process information across every email with near-instant response time.

AI Memory Management—Letta: Leveraging Letta, we implemented high-fidelity memory for contextual awareness. This lets Inbox Angel maintain “conversational memory,” recall related mail, and manage threads without ever getting lost or repetitive. Letta’s API allows for both advanced retrieval and summarization, creating life-like, friendly assistant interactions.

Fish Audio Text to Speech : Used for converting text to speech .

Conversational Assistant: Blending Groq’s language skills with Letta’s memory, we built a system where users can interact with their inbox as if chatting with a helpful friend-ask any question, from “what’s due today?” to “what did my landlord say last?” and get an instant, referenced answer.

Challenges We Faced

Building Inbox Angel showed us that making something “look easy” is, itself, quite a challenge! Some of the hurdles we faced:

Sponsor Integration Complexity: Integrating both Groq (for inferencing) and Letta (for memory/context) required significant adaptation. Each had its own API quirks, response formats, and recommended usage patterns. Ensuring they worked seamlessly, passing tokens, tracking history, and splitting responsibilities between “language” and “memory”, was anything but plug-and-play.

End-to-End Flow Debugging: Simulating a real inbox with threaded dummy emails introduced numerous edge cases: date handling, subject similarity, multi-person threads, and linking user queries to the right email objects. Ensuring accurate, cited responses stretched our backend logic and frontend UI.

Natural Conversation, Not Just Q&A: We wanted Inbox Angel to go beyond brittle keyword searching, to handle context, follow-ups, and “talk like a friend.” This pushed us to iterate deeply on prompts, evaluate conversational “memory,” and refine assistant tone.

Voice (Future Scope): We plan to add speech-to-text (for asking Inbox Angel) and text-to-speech (for having it read emails aloud). This opens up accessibility for visually impaired and less tech-savvy users, but will require thoughtful error handling (accents, ambient noise, summarization fidelity) and even deeper sponsor integration.

Performance Tuning: Handling queries over a large, multi-threaded mail DB with near-instant answers meant real attention to caching, model selection, and UI loading states, especially with the Groq API, where speed is both the promise and the challenge!

Every integration, technical hurdle, and UI/UX decision made Inbox Angel more robust, friendly, and real-world ready, while reminding us how many invisible details go into making tech genuinely helpful.

Accomplishments that we're proud of

End-to-end working demo: Shipped a fully functional prototype that can answer complex user queries about email threads, with direct citations for every answer.

Groq + Letta integration: Successfully blended Groq inferencing and Letta retrieval APIs for advanced conversational assistant functionality, something not documented or plug-and-play!

Universal user experience: Designed an intuitive mail client interface approachable by users of any age or tech background. Accessibility and simplicity were prioritized at every step.

Real-time Q&A: Achieved instant query answering across thousands of mails validating our system's efficiency and scalability.

Hackathon pace: Built and refined the entire project within a highly time-constrained weekend, overcoming multiple technical pivots and unexpected blockers.

What we learned

Conversational context is hard: Handling nuanced follow-ups, ambiguous questions, and “real human” conversation isn’t solved by keyword search. True AI assistance requires context, memory, and tone.

APIs are not one-size-fits-all: Integrating advanced APIs (Groq, Letta) takes deep experimentation, especially when managing state, request volume, and blending outputs for conversational tasks.

Accessibility matters: Features like voice support, readable UI, and natural interaction dramatically widen who can benefit from digital tools. We realized that making tech “inclusive” delivers impact beyond what we anticipated.

Performance tuning pays off: Fast response times are essential for user satisfaction. Caching strategies and batch retrievals can bring response time .

Teamwork and rapid iteration: Hackathons demand clear communication, quick pivots, and shared commitment. Every feature and improvement was a group effort, many key ideas emerged under pressure!

What's next for Inbox Angel

Future Scope: We plan next to add voice-based interactions, speech-to-text for queries and text-to-speech for responses. This will make Inbox Angel more natural and accessible, especially for older adults, visually impaired users, and anyone who just wants a hands-free, real conversation with their email.

Our sponsor integrations weren’t just bolt-ons—they were essential ingredients, making Inbox Angel’s assistant experience feel not just technically impressive, but truly helpful, fast, and ready for real life..

Built With

• fish
• groq
• letta
• typescript