I Wanted My Meeting Agent to Remember with Hindsight

When a meeting ends, the conversation is usually easy to forget.

The difficult part is remembering what was promised, what remained unresolved, what a client cared about, and what should matter when the same project comes up again. RecallMeet is designed around this problem. It combines meeting context with persistent memory using Hindsight so that preparation for a future meeting can use what happened in previous meetings — and can become more useful when the user provides feedback.

1. The Problem

Important information is often spread across multiple meetings.

A typical project may have:

Several meetings discussing the same project.
Commitments made during those meetings.
Unresolved questions or concerns.
Follow-ups that need to happen later.
Important context that becomes difficult to reconstruct weeks later.

A meeting transcript can tell us what was said, but that alone does not necessarily provide useful preparation for the next meeting. The challenge is turning previous conversations into useful long-term context.

2. The Solution

RecallMeet is a personal AI meeting-memory and preparation agent.

Instead of treating every meeting as an isolated conversation, RecallMeet connects meetings through their projects and remembers useful context.

The core loop is:
Remember
↓
Prepare
↓
Get Feedback
↓
Learn
↓
Prepare Better

3. How RecallMeet Works
The system is organized around a few key concepts:

• 1. Meetings — individual meeting records and transcripts.
• 2. Projects — connect related meetings together.
• 3. Commitments — track promises, follow-ups, and due dates.
• 4. Hindsight Memory — preserves useful contextual information for future interactions.
• 5. Prepare Me — retrieves relevant history and creates a personalized preparation.
• 6. Prep Feedback — allows the user to tell the system what the preparation missed or should emphasize next time.
• 7. This makes the system more than a transcript storage application.

4. Where Hindsight Fits

RecallMeet uses its application database for structured information such as meetings, projects, commitments, and due dates.

Hindsight is used for meaningful contextual memory that can help future preparation.

This distinction allows RecallMeet to combine structured project data with persistent agent memory.

5. Prepare Me

The Prepare Me feature brings together relevant project history before an upcoming meeting.

Instead of simply summarizing the latest meeting, it uses previous context and learned feedback to help answer:

What do I need to know before this meeting?

6. Technical Architecture

At a high level, RecallMeet brings together:
┌──────────────────┐
│ RecallMeet UI │
└────────┬─────────┘
│
▼
┌──────────────────┐
│ Backend │
└───────┬──────────┘
│
┌──────────┴──────────┐
▼ ▼
┌──────────────┐ ┌──────────────┐
│ PostgreSQL │ │ Hindsight │
│ Structured │ │ Long-term │
│ application │ │ memory │
│ data │ └──────┬───────┘
└──────────────┘ │
▼
┌──────────────┐
│ LLM │
│ Preparation │
└──────────────┘

7. Conclusion

RecallMeet is built around a simple idea: a useful meeting assistant should remember more than the transcript.

It should remember what happened, connect that information across a project, learn from what the user says was missing, and use that knowledge when preparing for the next interaction.

That is where Hindsight becomes an important part of the architecture — not simply as storage, but as persistent context that can influence future preparation.

Project Deliverables & Links
GitHub Repository: https://github.com/Ho436-art/RecallMeet
Demo Video: https://youtu.be/xuQOZrfyWJU