What I Built

Rememberly is an AI ecosystem that acts as a compassionate companion for individuals experiencing memory loss. It transforms a lifetime of scattered digital memories into interactive, narrative experiences.

Rememberly has two parts that work together.

For Family & Caregivers: A simple interface lets the support network upload and organize photos, videos, voice notes, and written stories into one shared library.

For the Patient: A gentle interface becomes their personal memory companion. They can ask questions and browse memories. The AI proactively surfaces relevant moments from their life.

Both parts are powered by Mistral and Vectara.

Testing Instructions

Option 1: Onboard as a new patient, upload pictures, videos or text to build your library. With that you can test the complete workflow

Option 2: Use pre-configured test accounts to experience the patient/caregiver workflow.

Test Accounts

Email: james5@gmail.com

Password: qahket-pumde6-Rosdub

Code

There's a helpful sequence diagram in the README that illustrates how everything works.

CijeTheCreator / rememberly

Rememberly is an AI ecosystem that acts as a compassionate companion for individuals experiencing memory loss.

Rememberly

Overview

Rememberly is an AI-powered memory companion for families caring for loved ones with dementia and memory loss. It transforms scattered photos, videos, and stories into an interactive memory library with two interfaces: one for caregivers to upload and organize content, and one for patients to explore memories through natural conversation and AI-generated "Cinematic Shows."

The system uses Mistral AI for intelligent reasoning and content generation, Vectara for secure RAG-based memory retrieval (with isolated knowledge boxes per patient), Deepgram for speech synthesis, and Convex for serverless functions and database management. Each patient's memories are kept completely private and isolated.

Sequence Diagram

sequenceDiagram
    participant Caregiver
    participant Patient
    participant Convex
    participant Mistral
    participant Vectara
    participant Deepgram
    Note over Caregiver,Vectara: Media Upload Flow
    Caregiver->>Convex: Upload media (images/videos/text)
    Convex->>Mistral: Extract tags & generate description
    Mistral-->>Convex: Return tags & description
    Convex->>Vectara: Ingest document to patient corpus
    Convex->>Convex: Store media with tags
    Note over Patient,Deepgram:

…

View on GitHub

How I Built It

I built Rememberly with a security-first approach. Despite the short length and hacky nature of the hackathon, the sensitive nature of memory and health data demanded this from the start.

Vectara powers the RAG pipeline. For maximum security, each patient gets their own isolated knowledge box.

Mistral AI powers the agent at the core. It reasons, orchestrates, and decides how to present memories. Sometimes it responds with simple text. Other times it creates immersive Cinematic Shows. When images are uploaded, an AI agent tags them for future use. These tags enable the Cinematic Shows and help answer patient questions.

Deepgram handles speech-to-text conversion.

Convex serves as the database and runs all serverless functions. Everything works together to keep data private while delivering a seamless experience.

DEV Community

Rememberly: Helping dementia patients reconnect with their cherished moments.

The Community