跳到主要内容

health-insights

❖ Communityv0.3.2★ 0

Deterministic, local-first analysis of your own Apple Health data - weekly trends, rule-based concern checks, nutrition versus Dietary Reference Intakes, an IOM energy target, lab results and opt-in modules (GLP-1 monitoring, medication adherence; all off by default). Portable Agent Plugins v1 package of ten skills that drive the separately installed health-insights command line tool (installed from the plugin's own pinned copy); the plugin itself contains no executable. Works on a self-hosted HealthRelay or Health Bridge receiver database.

Open in Hermes Desktop
hermes plugins install health-insights

README

From the reviewed commit d6d09c1 ↗; it updates when the author re-pins.

hermes-health-insights

Deterministic, local-first analysis of your own Apple Health data, packaged as a Hermes Agent plugin (skills) plus a small command line tool. The code computes the numbers and findings; your agent explains them.

Works on the SQLite database written by a self-hosted HealthRelay (or Health Bridge) receiver. Pair it with hermes-healthrelay, which gives your agent read-only MCP access to the same data.

Informational only, not medical advice. The rules are simple thresholds and can be wrong.

Setting up the whole chain (app and receiver, hermes-healthrelay, this plugin, hermes-medlog)? Follow the Full setup guide: one ordered walkthrough with a check after each step.

What it does

Command Purpose
health-insights weekly 7-day vs 28-day trends: sleep, resting heart rate, HRV, weight, SpO2, blood pressure
health-insights concerns Rule-based findings (levels 1 to 3): blood pressure, SpO2/ECG, vitals, labs, workouts, data gaps, nutrition gaps
health-insights nutrition Food log vs Dietary Reference Intakes (default table: adult men 31 to 50, see Limits; pass --dri for other groups), plus an energy line against an IOM energy target
health-insights labs ... Import and trend lab results from Apple Health clinical records
health-insights coverage/daily/baseline/anomalies/bp/workouts/ecg Building blocks and detail views
health-insights demo Write a synthetic database so you can try everything with no real data

Install

Two steps: the plugin adds skills to your agent; the analysis tool is installed separately with pipx.

  1. Install the tool and try it on synthetic data (no health data involved):
    pipx install git+https://github.com/mwdearing/hermes-health-insights@<sha>   # 40-character sha of the release you want
    health-insights --version
    health-insights demo --out demo.sqlite
    health-insights weekly --db demo.sqlite
    
  2. Install the plugin (skills for setup, weekly review, concerns, nutrition, energy target, labs and modules):
    hermes plugins install mwdearing/hermes-health-insights --no-enable
    hermes plugins enable health-insights
    
    Then ask your agent to follow the health-insights-setup skill.

Upgrade

hermes plugins update health-insights refuses installs pinned to a commit. Move both pieces to a new commit (use the 40-character sha of the release you want):

hermes plugins install mwdearing/hermes-health-insights --force --ref <sha>
pipx install --force git+https://github.com/mwdearing/hermes-health-insights@<sha>

Run the first command in an interactive terminal: the plugin declares a PyYAML dependency and Hermes asks before installing it. A non-interactive run skips the dependency and refuses to replace an active plugin (--no-deps cannot replace an active plugin either). --force keeps the plugin enabled or disabled as it was; start a new session afterwards. Check the tool with health-insights --version. Hermes installs the plugin into plugins/health-insights under its home (hermes plugins list shows the name).

Point it at your database

Set the receiver database once; after that no command needs --db. The tool uses the first of these that is set:

  1. --db <path> on the command
  2. the HEALTH_INSIGHTS_BRIDGE_DB environment variable
  3. bridge_db: in ~/.config/health-insights/config.yaml
  4. the first line of ~/.config/healthrelay/db-path (the same file the hermes-healthrelay plugin reads, so one setting serves both)

Check it with health-insights concerns. If the database cannot be opened, the tool prints one line saying so and exits with status 2. The rest of config.yaml is optional (timezone, weight unit, data directory, energy profile).

Uninstall

hermes plugins uninstall health-insights      # also: hermes plugins list, hermes plugins show health-insights
pipx uninstall health-insights

Optionally delete what the tool created: ~/.config/health-insights/ (settings and module switches) and ~/.local/share/health-insights/ (imported lab results and other local data). Your receiver database and ~/.config/healthrelay/db-path belong to HealthRelay and are not touched.

Optional modules (all off by default)

Extra checks for specific situations are opt-in modules. See what exists and switch them with:

health-insights modules list
health-insights modules info <id>
health-insights modules enable <id>     # and: disable <id>

Available now: glp1 (monitoring while on a GLP-1 type medicine, details) and medication_adherence (needs the separate medlog CLI, which is not included). Details, the settings file and how to add a module: docs/MODULES.md. Your agent can do this for you with the health-insights-modules skill, and only when you ask.

Privacy and trust

  • Everything runs on your machine. Nothing is uploaded.
  • Personal integrations are off by default and are not part of this plugin: missed-dose counts need a separate medlog CLI, and there are hooks for a push command and a local model endpoint for wording. Nothing here depends on them; with no model configured a fixed template writes the summary.
  • Health data is sensitive. Use a local model, or one you trust with it, and keep raw values out of public channels.
  • Enabling any Hermes plugin grants its skills full trust. The plugin contains only skills; the CLI is installed separately.

Limits

  • The bundled DRI table is for adult men aged 31 to 50, and nutrition output names it. It is only a rough guide for anyone else: supply your own table with --dri for other groups.
  • The energy target uses the standard IOM equation for adults and needs your profile; it is an estimate.
  • This project is not affiliated with Apple.

Development

pip install -e . pytest
pytest

Tests use synthetic data only. Licensed Apache-2.0.

← Back to the catalog · catalog built Oct 3, 2026