Meetings contain important information such as decisions, commitments, project discussions, and unresolved issues. As the number of meetings increases, remembering all this information becomes difficult.
RecallMeet is a personal AI meeting-memory and preparation agent that remembers a user's previous meetings, connects them through projects, tracks commitments, and prepares the user for upcoming meetings.
The core idea is:
Remember → Prepare → Feedback → Learn → Prepare better.
Each user's meetings and memory remain private and isolated from other users.
Important information is often scattered across multiple meetings of the same project. Users may have to search through old transcripts to remember decisions, commitments, and unresolved issues before a new meeting.
A traditional workflow is:
Meeting → Transcript → Summary
RecallMeet extends this into:
Meeting → Project Context → Long-Term Memory → Future Preparation
Instead of only answering “What happened in this meeting?”, RecallMeet focuses on:
“What do I need to know before my next meeting?”
Users upload meeting transcripts containing the project, date, participants, and conversation.
The system extracts:
Meetings from the same project are connected within the user's private history.
Apollo Meeting 1
↓
Requirements + Security
↓
Apollo Meeting 2
↓
API + Commitment
↓
Apollo Meeting 3
↓
Deployment + Unresolved Issue
↓
Apollo Meeting 4
↓
Client Concerns
↓
Project Memory
↓
"Prepare Me"
↓
Personalized Preparation
The “Prepare Me” feature uses this project history to generate previous discussions, pending commitments, unresolved issues, important context, and recommended discussion points.
RecallMeet separates structured data, AI reasoning, and long-term memory.
The Next.js frontend manages the user interface, while the FastAPI backend handles application logic.
The application database stores users, meetings, projects, commitments, due dates, and statuses. The LLM handles summarization, information extraction, reasoning, and preparation generation.
Hindsight provides long-term contextual memory, storing meaningful experiences, project context, and preparation feedback.
┌──────────────────────┐
│ USER │
│ Login / Dashboard │
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ NEXT.JS UI │
│ │
│ • Meeting Upload │
│ • Projects │
│ • Commitments │
│ • Prepare Me │
│ • Prep Feedback │
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ FASTAPI BACKEND │
│ │
│ • Authentication │
│ • Meeting Management │
│ • Project Management │
│ • Preparation Engine │
└──────────┬───────────┘
│
┌─────────────────────┼─────────────────────┐
│ │ │
▼ ▼ ▼
┌───────────────┐ ┌───────────────┐ ┌────────────────┐
│ PostgreSQL │ │ LLM ENGINE │ │ HINDSIGHT │
│ │ │ │ │ MEMORY │
│ Users │ │ Summarization │ │ Experiences │
│ Meetings │ │ Extraction │ │ Project Context│
│ Projects │ │ Reasoning │ │ Feedback │
│ Commitments │ │ Preparation │ │ Learning │
└───────────────┘ └───────┬───────┘ └───────┬────────┘
│ │
└──────────┬──────────┘
▼
┌──────────────────────┐
│ PREPARATION ENGINE │
│ │
│ Project History │
│ + Commitments │
│ + Issues │
│ + Feedback │
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ PERSONALIZED MEETING │
│ BRIEFING │
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ PREP FEEDBACK │
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ HINDSIGHT LONG-TERM │
│ LEARNING │
└──────────────────────┘
The workflow begins when a user uploads a meeting transcript. RecallMeet analyzes it and extracts useful information such as decisions, commitments, and unresolved issues.
This information is connected to the relevant project. Hindsight maintains meaningful long-term context, while structured information such as commitments and statuses is maintained by the application database.
When the user selects “Prepare Me,” relevant project history is retrieved and used to generate a personalized briefing.
Example:
Previous Discussions: API integration, security, deployment
Pending Commitments: Send API documentation
Unresolved Issues: Client budget concern
Recommended Focus: Follow up on commitments and unresolved concerns.
The most important feature of RecallMeet is its ability to use Prep Feedback to improve future preparation.
For example:
User Feedback:
“You missed the client's budget concern. Focus more on unresolved client concerns next time.”
This feedback becomes part of the user's long-term memory and can influence the next preparation.
┌─────────────────────┐
│ MEETING UPLOAD │
└──────────┬──────────┘
↓
┌─────────────────────┐
│ AI UNDERSTANDS │
│ MEETING │
└──────────┬──────────┘
↓
┌─────────────────────────────┐
│ PROJECT + DECISIONS │
│ COMMITMENTS + ISSUES │
└──────────────┬──────────────┘
↓
┌──────────────────┐
│ HINDSIGHT MEMORY │
└────────┬─────────┘
↓
┌──────────────────┐
│ "PREPARE ME" │
└────────┬─────────┘
↓
┌────────────────────────┐
│ PERSONALIZED PREP │
│ Context + Commitments │
│ Issues + Focus │
└───────────┬────────────┘
↓
┌──────────────────┐
│ PREP FEEDBACK │
└────────┬─────────┘
↓
┌──────────────────┐
│ HINDSIGHT LEARNS │
└────────┬─────────┘
↓
┌────────────────────────┐
│ NEXT PREPARATION │
│ Uses Previous Feedback │
└───────────┬────────────┘
↓
┌──────────────────┐
│ BETTER FOCUS │
└──────────────────┘
The key learning behavior is:
First Preparation → Feedback → Hindsight Memory → Second Preparation → Improved Focus
The user logs into RecallMeet, uploads meeting transcripts, and organizes them by project. The system builds project memory as more meetings are added.
Before an upcoming meeting, the user selects “Prepare Me” to receive a personalized briefing.
LOGIN
↓
UPLOAD MEETING
↓
ANALYZE TRANSCRIPT
↓
EXTRACT INFORMATION
↓
PROJECT MEMORY
↓
"PREPARE ME"
↓
PERSONALIZED PREPARATION
↓
PREP FEEDBACK
↓
HINDSIGHT LEARNS
↓
BETTER FUTURE PREPARATION
The MVP focuses on authentication, meeting uploads, project organization, information extraction, commitment tracking, Hindsight memory, personalized preparation, and Prep Feedback. Features such as real-time transcription, meeting-platform integrations, mobile apps, and cross-user memory are outside the initial MVP.
RecallMeet transforms meeting information into a continuously useful personal memory system.
Instead of treating meetings as isolated conversations, it connects them through project context, remembers important information, tracks commitments, and prepares users for future discussions.
Its central learning loop is:
RecallMeet is therefore more than a meeting summarizer—it is a private, project-aware AI memory that turns past meetings and preparation feedback into progressively better preparation for the user's next meeting.
Project Deliverables & Links
GitHub Repository: https://github.com/Ho436-art/RecallMeet
Demo Video: https://youtu.be/xuQOZrfyWJU