OpenHuman Model Card
这是可索引项目报告证据页：它保留 OpenHuman Model Card 的源材料入口、机制线索和限制提醒；正文仍需 reader/editor 与 academic public-copy review 后才能当作最终结论引用。
OpenHuman Model Card
One Sentence
OpenHuman is a UI-first personal-agent product with local memory, integrations, typed tools, and token compression.
Three Sentences
It overlaps directly with the user’s product-usability concern: the agent must be useful in daily life, not only impressive in a benchmark. The raw capture describes an Obsidian-compatible wiki, local-first memory, 118+ integrations, automatic fetch, typed tools, web search/fetch, voice, messaging, and token compression. In the graph it belongs beside OpenClaw and Hermes as a personal agent runtime.
Model Card
| Field | Value |
|---|---|
| Repository | tinyhumansai/openhuman |
| Source | raw-github/tinyhumansai_openhuman.md |
| Category | Personal agent product and memory system |
| Pattern | local memory/wiki -> integrations -> typed tools/token compression |
| Evidence | AnySearch GitHub extract, 2026-05-22 |
Teaching Use
Use OpenHuman to teach the product side of self-evolution: memory ownership, UI, integrations, recurring fetch, and token compression. It is a useful counterweight to paper-code systems because it asks whether the agent can survive real user workflows.
Evidence And Limits
The raw capture reports 25.1k stars, 2.3k forks, Rust/TypeScript implementation, GPL-3.0 license, and release v0.54.0. The next analysis should verify which parts are self-improving versus conventional personal-assistant features.