RecallMeet

An AI agent that remembers past interactions and prepares you for what comes next.

Comment

Inspiration

I was inspired by a common problem in meetings: important details, decisions, deadlines, and client requirements can easily be forgotten or lost in long conversations. I wanted to build an AI agent that could remember past interactions and help people prepare for what comes next.

What it does

RecallMeet is an AI memory agent that remembers important information from previous interactions and helps users prepare for what comes next.

It follows a simple workflow:

Retain → Recall → Prepare

  • Retain: Stores important details such as decisions, deadlines, budgets, concerns, and commitments.
  • Recall: Retrieves relevant information from previous conversations when needed.
  • Prepare: Uses the recalled information to prepare the user for their next meeting.

For example, if a client originally had a 5-month deadline and later changed it to 6 months because of security requirements, RecallMeet can remember why the deadline changed and bring that context into the next meeting.

How we built it

We built RecallMeet using:

  • Frontend: HTML, CSS, and JavaScript
  • Backend: Python and FastAPI
  • AI Memory: Hindsight
  • API: Hindsight memory API

The application stores important interaction details in Hindsight and retrieves relevant memories based on the user's request.

The core architecture is:

$$ \text{Interaction} \rightarrow \text{Retain} \rightarrow \text{Recall} \rightarrow \text{Prepare} $$

We also tested the system with multiple clients to make sure information from one client was not mixed with another.

Challenges we ran into

One of our biggest challenges was getting the AI to retrieve the right memories at the right time.

We also faced challenges with:

  • Connecting the frontend and backend
  • Integrating the Hindsight API
  • Debugging FastAPI endpoints
  • Handling memory retrieval
  • Keeping different client information isolated
  • Testing different real-world meeting scenarios

We solved these problems through repeated testing, debugging, and improving our memory workflow.

Accomplishments that we're proud of

We are proud that we turned an idea into a working AI memory agent instead of building only a concept or mockup.

Our main accomplishments include:

  • Built a working Retain → Recall → Prepare workflow
  • Integrated long-term AI memory using Hindsight
  • Created a functional frontend and FastAPI backend
  • Successfully recalled important client information
  • Demonstrated client-specific memory isolation
  • Created a real-world meeting preparation use case
  • Documented and demonstrated the project

What we learned

Through RecallMeet, we learned how AI memory can be used to solve practical problems beyond simple question answering.

We learned:

  • How to work with long-term AI memory
  • How to build APIs using FastAPI
  • How to connect a frontend with a backend
  • How to integrate external APIs
  • How to debug and test an AI application
  • How important context and memory are for AI agents
  • How to design an AI system around a real-world problem

Most importantly, we learned that building an AI agent is not only about generating responses — it is also about remembering the right context and using it at the right time.

What's next for RecallMeet

Our next goal is to turn RecallMeet from a working prototype into a more complete meeting assistant.

Future improvements include:

  • Automatic meeting transcription
  • Calendar integration
  • Email and meeting-history integration
  • User authentication and multiple user accounts
  • Smarter meeting preparation
  • Action-item and follow-up tracking
  • A more advanced AI agent that can proactively identify important changes

Our long-term vision is to make RecallMeet an AI assistant that doesn't just remember what happened, but understands why it happened and what should happen next.

Built With

Share this project:

Updates

Submission history