UnderPressure

Lightweight blood pressure tracking web app for Raspberry Pi (or any host). Log morning/evening readings, view trends, earn badges, optional AI insights, and export PDF reports.

Features

• Log: Morning and evening SBP/DBP (40–250 mmHg) with optional notes
• History: View, edit, and delete past entries
• Dashboard: Weekly, monthly, quarterly, yearly trend charts with high-BP zones (SBP ≥ 135, DBP ≥ 85, configurable in Admin)
• Statistics: Aggregated stats, morning vs evening, trends, device breakdown, measurement habits
• AI insights: Optional LLM-generated summary of your BP data (Statistics page). Choose target language (English/Korean), refresh on demand, and use a copyable prompt for external LLM services; raw measurement data is included in the prompt for week/month/quarter ranges when length allows
• Badges: Streaks, first-of-month, and entry milestones (10, 50, 100, 250)
• PDF report: Date-range report with summary table and trend chart (doctor-friendly)
• Backup: Daily email backup of data.json via SMTP (configurable)
• Import: One-time Excel import; rows containing "Week average" are skipped
• Export/restore: Full backup JSON includes records, meta, and settings (including LLM configuration)

Tech Stack

• Backend: Python, FastAPI, JSON file (data.json), optional LLM (Google Gemini via google-genai)
• Frontend: React, Vite, Tailwind CSS, Recharts
• PDF: ReportLab

Quick Start

Virtual environment (.venv)

Use a virtual environment at the project root named .venv for all Python work (backend, migration CLI, tests). Do not rely on the system Python for project dependencies.

# From project root — create once
python3 -m venv .venv

# Install dependencies (paths below avoid needing `activate` first)
.venv/bin/pip install -r backend/requirements.txt
# Windows: .venv\Scripts\pip install -r backend\requirements.txt

Optional: source .venv/bin/activate (Windows: .venv\Scripts\activate) and then use pip / python as usual.

run.py re-executes with .venv’s interpreter when .venv exists, so python3 run.py still picks up the venv for the server. For any other Python command (e.g. migration), call .venv/bin/python explicitly or activate the venv first.

Development

1. Backend (from project root):

   .venv/bin/python run.py

   (Or: activate .venv, then python run.py.) API: http://127.0.0.1:8000

2. Frontend (separate terminal):

   cd frontend && npm install && npm run dev

   App: http://localhost:5173 (proxies /api to backend)

Production (e.g. Raspberry Pi)

1. Copy backend/.env.example to backend/.env and set SMTP vars for backup (optional). For AI insights, set GEMINI_API_KEY (and optionally GEMINI_MODEL, LLM_PROVIDER=gemini); enable and choose provider/model in Admin → System settings → AI / LLM settings.
2. Create .venv, install Python deps, build frontend, then run:

   python3 -m venv .venv
   .venv/bin/pip install -r backend/requirements.txt
   cd frontend && npm run build && cd ..
   .venv/bin/python run.py

   (On Windows, use .venv\Scripts\pip and .venv\Scripts\python.) Open http://<host>:8000 — backend serves the built frontend.

Docker

# Build (includes frontend build in multi-stage Dockerfile)
docker compose up --build

Data is stored in ./data.json; mount it as in docker-compose.yml to persist.

Excel Import

Columns: date, morning_sbp, morning_dbp, evening_sbp, evening_dbp, note. Rows whose first column contains "Week average" are skipped.

• CLI: .venv/bin/python -m backend.migration path/to/file.xlsx [import_wins|keep_existing] (or activate .venv first, then python -m backend.migration …)
• API: POST /api/import with multipart file and query strategy=import_wins|keep_existing

Backup

• Daily at 02:00 (APScheduler), or trigger via POST /api/backup/run
• Status: GET /api/backup/status
• Configure in backend/.env: SMTP_SERVER, SMTP_PORT, SENDER_EMAIL, RECEIVER_EMAIL, PASSWORD

AI Insights

• Statistics page: An "AI insights" block appears between the time-range selector and the Overview. Click Refresh AI insights to generate a summary for the current range; the last result is shown until you refresh again.
• Target language: In the "Prompt for external AI service" area, choose English or Korean; the prompt and (when using the in-app LLM) the response use this language.
• Copyable prompt: The prompt includes aggregated stats and, for week/month/quarter ranges, raw daily records so you can paste it into an external LLM (e.g. ChatGPT, Claude) if the built-in provider is unavailable or rate-limited.
• Admin: Enable AI insights and set Provider (e.g. Gemini, Dummy) and Model (e.g. gemini-2.5-flash) under System settings → AI / LLM settings. API keys are not stored in the app; set GEMINI_API_KEY (and optionally GEMINI_MODEL) in backend/.env.

API Summary

Method	Path	Description
GET	/api/records	List records (optional ?from=&to=)
GET	/api/records/{date}	Get one record
POST	/api/records	Create (body + optional ?date=)
PUT	/api/records/{date}	Update
DELETE	/api/records/{date}	Delete
GET	/api/badges	Badges (streaks, milestones)
GET	/api/aggregated	Trend data (`?period=week
POST	/api/reports/pdf	PDF report (body: {"from":"YYYY-MM-DD","to":"YYYY-MM-DD"})
POST	/api/insights	AI insights (body: from, to, focus, locale; uses settings llm_enabled, llm_provider, llm_model)
GET/PUT	/api/settings	Get/update settings (receiver_email, devices, sbp_high, dbp_high, llm_enabled, llm_provider, llm_model)
POST	/api/import	Excel upload
GET/POST	/api/backup/status, /api/backup/run	Backup status and manual run

Data

• data.json (project root): Records and meta. Structure:

{
  "records": [
    {
      "date": "YYYY-MM-DD",
      "morning_sbp": 120,
      "morning_dbp": 80,
      "evening_sbp": 118,
      "evening_dbp": 78,
      "note": "",
      "device": null
    }
  ],
  "meta": { "last_updated": "ISO8601", "version": 1 }
}

Writes are atomic (temp file + rename) to avoid corruption.

• settings.json (project root): App settings—receiver email, auto backup, devices, high-zone thresholds (sbp_high, dbp_high), and LLM options (llm_enabled, llm_provider, llm_model). Exported and restored with Export/restore in Admin so backups include LLM configuration (API keys stay in .env only).