Skip to content

Releases: Hackathons-ULT/meditrack

Release list

v.1.0.1

v.1.0.1 Latest
Latest

Choose a tag to compare

View all tags
@Jay-Jay-Tee Jay-Jay-Tee released this 08 Mar 18:40
1.0.1
bfad3ef
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

MediTrack v1.0.1

AI-powered medical timeline system — one patient ID, every hospital.

The first stable release of MediTrack. Everything needed to unify a patient's medical history across providers: semantic event storage, voice dictation, AI-generated clinical summaries, document upload, PDF export, and shareable read-only timelines.

What's included

Core system

  • Patient registration & authentication — email/password accounts with bcrypt hashing; every user receives a globally unique MED-XXXXXXXX Patient ID on signup
  • Medical event ingestion — add events by text or voice with event type, doctor name, hospital name, and optional backdated timestamp; stored as 384-dimensional semantic vectors in Qdrant
  • Semantic search — FastEmbed (BAAI/bge-small-en-v1.5) embeddings with cosine similarity; queries find conceptually related records, not just keyword matches

AI analysis

  • Groq Llama 3.3 70B timeline summaries — full clinical narrative covering visit patterns, gaps, and frequency at ~500 tokens/sec
  • Data quality scoring — Rich / Moderate / Sparse rating based on event count, time span, and content depth
  • Timeline insights — activity rate, continuity rating, longest gap, event type breakdown, and unique provider counts

Document handling

  • File upload — drag-and-drop or picker; images and PDFs up to 10MB stored with UUID filenames
  • Document download — retrieve any uploaded file directly from the timeline at any time

Export & sharing

  • Professional PDF export — ReportLab-generated reports with clinical header, patient ID, timestamp, and full event table; print-ready for referrals or insurance
  • Shareable patient links — public read-only URL (/patient/MED-XXXXXXXX) requires no login; safe to share with providers

Interface

  • Voice-to-medical-note — real-time browser dictation via Web Speech API (Chrome/Edge)
  • Dark mode — full light/dark toggle with CSS variables
  • Responsive layout — mobile, tablet, and desktop

Infrastructure

  • GET /health and GET /api/status endpoints for monitoring
  • Graceful error handling with specific messages for Groq rate limits, auth failures, and missing data
  • Production-ready with Gunicorn + nginx; deployed on Render

Stack

Layer Technology
Backend Flask 3.x, Flask-Login, Flask-CORS
Vector DB Qdrant (cloud), 384-dim cosine similarity
Embeddings FastEmbed — CPU-only, no GPU required
AI Groq llama-3.3-70b-versatile
PDF ReportLab 4.x
Auth bcrypt + Flask-Login
Frontend Vanilla JS, Tailwind CSS, Web Speech API

Setup

git clone https://github.com/your-org/meditrack.git
cd meditrack
python -m venv venv && source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env   # fill in Qdrant + Groq credentials
python app.py

Requires: Python 3.8–3.13, a Qdrant instance, and a Groq API key. Both have free tiers.

Known limitations in this release

  • User accounts are stored in memory — a server restart wipes all accounts (Qdrant event data persists)
  • No rate limiting on any endpoint
  • Shareable links cannot be revoked
  • No RBAC — all authenticated users have identical permissions
  • Not HIPAA/GDPR compliant out of the box

These are tracked and will be addressed in upcoming releases. See CONTRIBUTING.md if you want to help.

Links

MediTrack is an information retrieval and summarization tool. It does not diagnose conditions, suggest treatments, or replace professional medical judgment.

# MediTrack v1.0.0

AI-powered medical timeline system — one patient ID, every hospital.

The first stable release of MediTrack. Everything needed to unify a patient's medical history across providers: semantic event storage, voice dictation, AI-generated clinical summaries, document upload, PDF export, and shareable read-only timelines.

What's included

Core system

  • Patient registration & authentication — email/password accounts with bcrypt hashing; every user receives a globally unique MED-XXXXXXXX Patient ID on signup
  • Medical event ingestion — add events by text or voice with event type, doctor name, hospital name, and optional backdated timestamp; stored as 384-dimensional semantic vectors in Qdrant
  • Semantic search — FastEmbed (BAAI/bge-small-en-v1.5) embeddings with cosine similarity; queries find conceptually related records, not just keyword matches

AI analysis

  • Groq Llama 3.3 70B timeline summaries — full clinical narrative covering visit patterns, gaps, and frequency at ~500 tokens/sec
  • Data quality scoring — Rich / Moderate / Sparse rating based on event count, time span, and content depth
  • Timeline insights — activity rate, continuity rating, longest gap, event type breakdown, and unique provider counts

Document handling

  • File upload — drag-and-drop or picker; images and PDFs up to 10MB stored with UUID filenames
  • Document download — retrieve any uploaded file directly from the timeline at any time

Export & sharing

  • Professional PDF export — ReportLab-generated reports with clinical header, patient ID, timestamp, and full event table; print-ready for referrals or insurance
  • Shareable patient links — public read-only URL (/patient/MED-XXXXXXXX) requires no login; safe to share with providers

Interface

  • Voice-to-medical-note — real-time browser dictation via Web Speech API (Chrome/Edge)
  • Dark mode — full light/dark toggle with CSS variables
  • Responsive layout — mobile, tablet, and desktop

Infrastructure

  • GET /health and GET /api/status endpoints for monitoring
  • Graceful error handling with specific messages for Groq rate limits, auth failures, and missing data
  • Production-ready with Gunicorn + nginx; deployed on Render

Stack

Layer Technology
Backend Flask 3.x, Flask-Login, Flask-CORS
Vector DB Qdrant (cloud), 384-dim cosine similarity
Embeddings FastEmbed — CPU-only, no GPU required
AI Groq llama-3.3-70b-versatile
PDF ReportLab 4.x
Auth bcrypt + Flask-Login
Frontend Vanilla JS, Tailwind CSS, Web Speech API

Setup

git clone https://github.com/your-org/meditrack.git
cd meditrack
python -m venv venv && source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env   # fill in [Qdrant](https://cloud.qdrant.io/) + Groq credentials
python app.py

Requires: Python 3.8–3.13, a Qdrant instance, and a [Groq](https://console.groq.com/keys) API key. Both have free tiers.

Known limitations in this release

  • User accounts are stored in memory — a server restart wipes all accounts (Qdrant event data persists)
  • No rate limiting on any endpoint
  • Shareable links cannot be revoked
  • No RBAC — all authenticated users have identical permissions
  • Not HIPAA/GDPR compliant out of the box

These are tracked and will be addressed in upcoming releases. See [CONTRIBUTING.md](CONTRIBUTING.md) if you want to help.

Links

MediTrack is an information retrieval and summarization tool. It does not diagnose conditions, suggest treatments, or replace professional medical judgment.

Assets 2
Loading