Ask
Questions become lookup, trend, review, consult, or research actions.
@health what are rods current concerns?local-first · fuzzy input · alias-only · own-risk
Your private, own-risk health concierge.
Talk to your health data in messy human language. LLM Health turns fuzzy requests like @health whats my mercury?, add today's weight 92.7kg, or plot liver with weight overlay into local review artifacts, timelines, diagnostic gaps, test candidates, consults, and research queues.
health CLI@health/healthfuzzy routerhealth-v2MIT
required before use
Big disclaimer: this is own-risk software, not medical care.
No medical advice: LLM Health does not diagnose, prescribe, order tests, treat disease or create a clinician relationship. It organizes records, calculations, questions and research notes. You are responsible for all decisions, delays, experiments, purchases, interventions and omissions.
Do not use it for emergencies or urgent symptoms. Do not start, stop, combine or change medications, supplements, procedures, devices or preventive protocols solely because an agent suggested or questioned something.
Every private HUB must explicitly accept the agreement once. The CLI stores only the agreement version and timestamp in agreement.json.
get running
Install, read agreement, initialize HUB.
The default path is ~/health. Use any private local or synced directory. Keep raw PDFs, Apple Health exports and clinical dumps out of Git.
brew tap nvk/tap
brew install llm-health
health agreement show
health config hub-path ~/health --init --accept-risk
health doctor
health welcomefuzzy input
Say the thing. The router figures out the command.
Like LLM Wiki, the default interface is not memorizing subcommands. You can be vague, misspell, mix prose with numbers, and let the agent route intent to deterministic local tools. When it writes to your HUB, it should leave reviewable artifacts instead of pretending the chat answer is the record.
Record
Messy context becomes self-reported notes or observations with provenance.
@health add todays weight 92.7kgExplore
Chart requests become health-v2 timelines with profiles, overlays, ranges, exports, and filters.
@health plot liver vs weight 18mo@health whats my mercury?
@health what is rods latest mercury and normal range?
@health why is liver flagged?
@health show liver category for rod, all time
@health overlay weight on ALT and AST for the last 18 months
@health add today's weight of 92.7kg
@health GI is fine, update as self reported
@health record 400mg albendazole single dose two weeks ago
@health close gaps for cara
@health what tests would reduce uncertainty the most?
@health queue deeper research on bilirubin and Gilbert syndrome
@health export the plotted charts as a PDFexamples
More ways to talk to it.
Use exact CLI when you want repeatability. Use fuzzy chat when you want speed. LLM Health should normalize names, aliases, dates, categories, units, and intent before touching your private HUB.
Results & ranges
@health whats my ferritin?@health show Cara's latest abnormal markers@health mercury trend with expected range@health which values are pending and not plotted?
Trends & charts
@health plot all liver markers@health compare ALT, AST, bilirubin and weight@health show 30d, 90d, YTD, 18mo shortcuts@health stack heavy metals and toxins
Wearables & vitals
@health show steps around latest labs@health add BP 118/76 this morning@health import apple health for rod@health summarize sleep and resting HR last month
Self-reported context
@health GI is fine, don't treat it as a concern@health I stopped alcohol for 30 days, mark as context@health record new supplement: magnesium nightly@health note family history of thyroid issues
Reviews & consults
@health review rod after latest labs@health consult internal medicine on my timeline@health toxins consult for mercury and lead@health meds and supplements review for side-effect risk
Diagnostic gaps
@health close gaps for rod@health what should I test next, must-have vs nice-to-have?@health create a test battery for liver and metabolic risk@health what data would change the interpretation?
Research queues
@health deep research ALT elevation vs weight loss@health paper review on mercury blood testing@health find supplement quality options for selenium@health compare mainstream and edge hypotheses, label uncertainty
Family profiles
@health enroll Sol, born 2018@health show only Rod, not Cara@health compare family thyroid markers@health update aliases so no legal names appear
Exports & housekeeping
@health export plotted charts as PDF@health doctor@health sync v2 with the wiki@health package de-identified sample, no dox
interfaces
Use it from a terminal, dashboard, or chat.
Review console
The health CLI is the deterministic core for reads, writes, agreement status, imports, reviews, gaps, batteries, and consults.
Dashboard
The repackaged health-v2 stack provides timelines, chart exports, wearable context, categories, overlays, and local analytics.
Chat concierge
Claude Code, Codex, OpenCode, Pi, and portable AGENTS files are supported surfaces. The product story is your private health HUB, not the coding agent.
what it does
Review artifacts, not magic answers.
- Ingest alias-only observations, records, labs, wearables and context notes.
- Generate quick-review cards after new results.
- Open diagnostic gaps and profile-aware test candidates.
- Run broad category consults: internal medicine, liver, toxins/exposures, meds/supplements, habits, family history and more.
- Queue deeper research across evidence lenses while preserving disagreement.
- Ship the repackaged
health-v2analytics/dashboard stack.
what it refuses to pretend
Limits stay visible.
- Not a doctor, not a prescription engine, not an emergency service.
- No raw medical files, source paths, legal names, full DOBs or Apple device/source names in publishable artifacts.
- Tests are
TEST_CANDIDATEcards, not orders. - Medication, supplement and protocol work is framed as review material with red flags and uncertainty.
- Research summaries can be wrong or stale; source quality and conflicts must stay visible.
command equivalents
Exact commands are still there when you want them.
health enroll --alias alex --birth-year 1983 --role adult
health data-wishlist
health dr-visit --profile alex --cadence onboarding
health ingest-note --profile alex --marker ALT --value 76 --unit U/L --category liver --flag high
health result --profile alex --marker mercury
health review --profile alex
health close-gaps --profile alex
health test-battery --profile alex --scope expanded --sources
health consult --profile alex --specialist auto
health-v2
health plugin-pathsFAQ
Questions people should ask before trusting any health agent.
What is LLM Health?
A local-first health intelligence package: CLI, fuzzy agent instructions, private HUB, review artifacts, diagnostic gaps, test candidates, consults, and the health-v2 analytics dashboard.
What does “fuzzy input” mean?
You can write rough requests like @health whats rods mercury. The agent maps intent to lookup, ingest, chart, review, consult, gap-closing, or research actions.
Do I need exact command syntax?
No. Exact CLI is available for repeatability, but normal use can be chatty. Dates, aliases, categories and units can be inferred when safe, or the agent should ask before writing.
Is this medical advice?
No. It is experimental software for organizing records and questions. It does not diagnose, prescribe, order tests, treat disease, or replace a clinician.
Where is my data?
In your local HUB. The website has no backend, accounts, analytics or health-data upload. Use health config hub-path to choose the private directory.
Can I use iCloud/Dropbox?
Yes, but use aliases, keep raw dumps out of Git, and be mindful of every app/agent with filesystem access. Sync is convenience, not a privacy guarantee.
What are aliases?
Profiles should use non-legal labels like rod, cara, or alex. Publishable artifacts should not contain legal names, source paths, full DOBs, or raw device/source identifiers.
What is the HUB?
The HUB is the private folder where profiles, observations, context notes, reviews, research queues, exports and agreement state live. It is the memory, not the chat transcript.
How does it handle different frequencies?
Labs may be sparse, weight may be weekly, Apple Health can be daily or minute-level. The dashboard aggregates contextually instead of forcing everything into one cadence.
What is a diagnostic gap?
A gap is a missing data point or uncertainty that materially limits interpretation. Gap cards explain why a test or history question might reduce uncertainty.
Are test candidates orders?
No. They are prioritized suggestions to discuss, research, or decide on independently. They should show priority, friction, rationale and uncertainty.
What are category consults?
Broad specialist-like passes over a topic area — for example liver, toxins, internal medicine, meds/supplements, habits, family history, metabolic health or wearables.
Can it compare people?
Yes, when profiles are intentionally enrolled. Family context can help spot shared patterns, but the system should keep aliases and avoid doxing.
Can it ingest Apple Health?
Yes. Apple Health exports can become summarized wearable context so charts can compare labs against weight, activity, sleep, heart rate and other time-series signals.
What happens to pending results?
Pending values should be visible as pending and not plotted as numeric dots. The chart should distinguish lab/source flags from interpreted concern.
Can it export charts?
Yes. The v2 dashboard supports plotted chart exports such as PDF so reviews can be shared or archived without uploading data to this website.
What if the agent is wrong?
Assume it can be wrong. Prefer reviewable artifacts, source links, visible uncertainty, normal ranges, raw values, dates and profile context. Correct it with a self-reported note when needed.
How is it different from a dashboard?
A dashboard shows data. LLM Health also remembers context, routes fuzzy requests, opens gaps, queues deeper research, and turns new results into review workflows.