Skip to content

Releases: thinkwee/HiMe

HiMe v1.0.2

29 Apr 09:29
@thinkwee thinkwee
v1.0.2
51768b2

Choose a tag to compare

View all tags
HiMe v1.0.2 Latest
Latest

Full Changelog: v1.0.1...v1.0.2

Assets 2
Loading

HiMe v1.0.1

26 Apr 12:10
@thinkwee thinkwee
v1.0.1
9c7f437

Choose a tag to compare

View all tags
HiMe v1.0.1

New Features & Fixes

  • Fix the soul.md reload problem
  • Fix a UTC timestamp question
  • update iOS App to v1.0.1: Now you can delete reports on the iOS app; improve the onboarding experience; faster sync for the first time installing the app

What's Changed

  • Bump eslint-plugin-react-hooks from 7.0.1 to 7.1.1 in /frontend by @dependabot[bot] in #21
  • Update anthropic requirement from <1,>=0.40 to >=0.96.0,<1 in /backend by @dependabot[bot] in #14
  • Update croniter requirement from >=1.3 to >=6.2.2 in /backend by @dependabot[bot] in #16
  • Upgrade React to 19.2.5 and lucide-react to 1.9.0 by @thinkwee in #24
  • Bump fastapi from 0.109.0 to 0.136.1 in /backend by @dependabot[bot] in #15
  • Improve backend coverage for utility serialization and inbox queue behavior by @Copilot in #25

New Contributors

  • @dependabot[bot] made their first contribution in #21
  • @thinkwee made their first contribution in #24
  • @Copilot made their first contribution in #25

Full Changelog: v1.0.0...v1.0.1

Contributors

thinkwee and dependabot
Loading

HiMe v1.0.0

22 Apr 17:41
@thinkwee thinkwee
v1.0.0
324ea99

Choose a tag to compare

View all tags
HiMe v1.0.0

HiMe v1.0.0 — Initial Public Release

Say Hi to Healthy Me.

HiMe is a self-hosted, fully local, open-source personal health AI agent platform. It continuously streams HealthKit samples from your Apple Watch and iPhone into your local server, runs an autonomous agent powered by the LLM provider of your choice.
This is the first public release.

✨ What's in the box

Autonomous agent

  • Single-agent, event-driven architecture — one loop drains three priority queues (chat > triggers > cron) sequentially, eliminating LLM lock contention.
  • 15+ LLM providers first-class: Anthropic, OpenAI, Gemini, Amazon Bedrock, ZhipuAI, plus Azure OpenAI, Google Vertex AI, and OpenAI-compatible endpoints (Groq, Mistral, DeepSeek, MiniMax, xAI, Perplexity, OpenRouter, vLLM).
  • Transparent fallback — 3 consecutive capacity errors auto-switch to a backup provider.
  • Layered prompt system — soul / rules / memory layers assembled at runtime; the agent writes its own memory via update_md_tool over time.
  • Fact verification — every outbound message is cross-checked against tool results before sending, to curb LLM hallucination.

Data ingestion

  • Apple Watch + iPhone HealthKit → WebSocket exporter → backend SQLite.
  • Rolling retention window (default 30 days, configurable).
  • Event triggers — analyses can fire on feature anomalies, not just on schedule.

Interactions

  • iOS app (iPhone + watchOS) with live cat mascot, chat UI, and background HealthKit sync. Install from the App Store.
  • iPhone home-screen widgets and Apple Watch widgets.
  • Telegram and Feishu gateways with evidence-backed card replies.
  • React 18 + Vite dashboard for reports, tasks, skills, prompts, and configuration.
  • Agent-generated personalised pages — the agent writes its own HTML dashboards at runtime, served with strict CSP and a sandboxed Python route handler.

Extension points

  • Skills system — user-authored analysis playbooks as Markdown with YAML frontmatter. Five bundled:
    • baseline_change_point, circadian_alignment, respiratory_screen, sedentary_behavior_audit, sleep_regularity_index
  • Seven core agent tools: sql_tool, code_tool (optional Docker sandbox), reply_user_tool, push_report_tool, create_page_tool, read_skill_tool, update_md_tool.
  • New LLM providers and new tools are a file drop — see AGENTS.md.

Deployment

  • Docker Compose (backend + frontend + watch exporter) in one command.
  • Native Python + Node.js mode for development.
  • Bundled setup.sh interactive wizard for first-time configuration (LLM provider, IM gateway, keys).
  • Bilingual UI (English + 简体中文) across backend, frontend, and iOS.

🚀 Quick start

git clone https://github.com/thinkwee/HiMe.git
cd HiMe
./setup.sh           # interactive wizard, ~2–5 min

Install HiMe on the App Store → Settings → enter your server URL → say hi to your bot.

Full install paths: docs/INSTALL.md. Production patterns: docs/DEPLOYMENT.md.

🧭 For contributors

- Architecture + three design principles — AGENTS.md
- Developer guide — docs/DEVELOPMENT.md
- Contribution process — CONTRIBUTING.md

⚠️ Notes

- HiMe is research-grade software for personal use. It is not a medical device and does not provide diagnoses.
- Released under the PolyForm Noncommercial License 1.0.0. Non-commercial use only.
- Privacy posture: fully self-hosted, no telemetry. Health data stays on your hardware; LLM calls go only to the provider you configure. Details: PRIVACY.md.
- Security disclosures: see SECURITY.md — do not file public issues for vulnerabilities.

---
Questions or feedback? Open a GitHub issue or Discussion.
Loading