diff --git a/README.md b/README.md index 4a4618033..36e79d249 100644 --- a/README.md +++ b/README.md @@ -24,12 +24,10 @@

-

- OpenHuman is your Personal AI super intelligence: local memory, managed services where needed, simple and powerful. + OpenHuman is your personal AI super intelligence: a brain that remembers everything, a fantastic orchestrator, a deep researcher. Local-first, simple, powerful.

-

Discord • Reddit • @@ -42,8 +40,6 @@ 🇺🇸 English | 🇨🇳 简体中文 | 🇯🇵 日本語 | 🇰🇷 한국어 | 🇩🇪 Deutsch | 🇵🇰 اردو

- -

Early Beta Latest Release @@ -58,7 +54,7 @@ > **Early Beta**: Under active development. Expect rough edges. -> **Local + managed services, upfront:** OpenHuman stores its Memory Tree, Obsidian-style Markdown vault, workspace config, and local runtime state on your machine. The default managed experience still uses OpenHuman-hosted services for account sign-in, model routing, web search proxying, and managed integration/OAuth flows through the Composio connector layer. Choose custom/local settings if you want to bring your own model, search, or Composio credentials; some real-time triggers and hosted features still require the managed backend. +> 🎉 Within one week of launch, OpenHuman became the number one trending repository on GitHub for nine days in a row. # Install @@ -112,27 +108,36 @@ A separately signed script-install path is not currently available. Issue [#2620 # What is OpenHuman? -OpenHuman is an open-source agentic assistant designed to integrate with you in your daily life. Each bullet links to the deeper writeup in the [docs](https://tinyhumans.gitbook.io/openhuman/). +OpenHuman is three things most assistants aren't: **a brain** that builds a persistent, local memory of your world; **a fantastic orchestrator** that runs fleets of agents on durable graphs; and **a deep researcher** that sweeps your data and the web before you finish asking. Every bullet links to the deeper writeup in the [docs](https://tinyhumans.gitbook.io/openhuman/). -- **Simple, UI-first & Human** A clean desktop experience and short onboarding paths take you from install to a working agent in a few clicks — no config-first setup, no terminal required. The agent has [a face](https://tinyhumans.gitbook.io/openhuman/features/mascot): a desktop mascot that speaks, reacts to its surroundings, [joins your Google Meets](https://tinyhumans.gitbook.io/openhuman/features/mascot/meeting-agents) as a real participant, remembers you across weeks, and keeps thinking in the background even when you've stopped typing. +### 🧠 The brain -- **100+ one-click OAuth integrations, 5,000+ MCP servers, 90,000+ Skills**: plug into Gmail, Notion, GitHub, Slack, Stripe, Calendar, Drive, Linear, Jira and the rest of your stack with [**one-click OAuth**](https://tinyhumans.gitbook.io/openhuman/features/integrations) — 100+ curated connectors brokered through the Composio layer. Beyond that, OpenHuman browses the open **Model Context Protocol** ecosystem (Smithery + the official MCP registry — thousands of servers) and a **90,000-entry Skills catalog**, so the agent can install new typed tools and skills on demand. Every connection becomes a typed tool, and every twenty minutes [auto-fetch](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/auto-fetch) walks each active connection and pulls fresh data into the [memory tree](https://tinyhumans.gitbook.io/openhuman/features/memory-tree). No prompts, no polling loops you have to write, so the agent already has tomorrow's context this morning. +- **[Memory Tree](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) + [Obsidian Wiki](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki)**: your data compressed into scored Markdown trees in SQLite on your machine, mirrored as an [Obsidian vault](https://x.com/karpathy/status/2039805659525644595) you can open and edit. No vector-soup black box. +- **[100+ OAuth integrations, 5,000+ MCP servers, 90,000+ Skills](https://tinyhumans.gitbook.io/openhuman/features/integrations)**: one click into Gmail, Notion, GitHub, Slack and the rest of your stack. [Auto-fetch](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/auto-fetch) feeds the brain every 20 minutes — it has tomorrow's context this morning. +- **[A subconscious](https://tinyhumans.gitbook.io/openhuman/features/subconscious)**: a background loop that diffs your world, advances your goals, and writes your morning briefing — thinking continues after you stop typing. +- **[Goals & Todos](https://tinyhumans.gitbook.io/openhuman/features/goals-and-todos)**: long-term goals, durable per-thread goals, and a shared kanban board per conversation. +- **[TokenJuice](https://tinyhumans.gitbook.io/openhuman/features/token-compression)**: tool output compressed before it hits the model — same information, up to 80% fewer tokens. A brain this big would be unaffordable without it. - Managed integrations use OpenHuman's Composio connector layer. OAuth handshakes and integration tool calls are proxied through the managed backend by default. If you want to run Composio directly instead, configure direct mode with your own Composio API key; real-time trigger webhooks then need to be hosted and wired by you. +### 🕸️ The orchestrator -- **[Memory Tree](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) + [Obsidian Wiki](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki)**: a local-first knowledge base built from your data and your activity. Everything you connect is canonicalized into ≤3k-token Markdown chunks, scored, and folded into hierarchical summary trees stored in **SQLite on your machine**. The same chunks land as `.md` files in an Obsidian-compatible vault you can open, browse and edit, inspired by Karpathy's [obsidian-wiki workflow](https://x.com/karpathy/status/2039805659525644595). +- **[Workflows](https://tinyhumans.gitbook.io/openhuman/features/workflows)** _(new)_: the agent proposes the automation; you review it on a canvas and save. Durable, trigger-driven, approval-gated runs on open-source [tinyflows](https://github.com/tinyhumansai/tinyflows). +- **[A harness that finishes the job](https://tinyhumans.gitbook.io/openhuman/developing/architecture/agent-harness)** _(new)_: checkpointed graph runs on open-source [tinyagents](https://github.com/tinyhumansai/tinyagents) — stuck agents get steered, halted ones return a root cause, every run replays with real per-call costs. +- **[A split brain, always on](https://tinyhumans.gitbook.io/openhuman/features/orchestration)** _(new)_: a fast reflex agent triages inbound traffic while a deep reasoning core delegates to worker fleets, steered by the subconscious. +- **[An agent economy](https://tinyhumans.gitbook.io/openhuman/features/tinyplace)**: a `@handle` on [tiny.place](https://tiny.place), Signal-encrypted agent-to-agent orchestration, x402 USDC bounties and trading — keys never touch disk. -- **[SuperContext](https://tinyhumans.gitbook.io/openhuman/features/super-context)**: a fresh chat shouldn't start cold. With SuperContext enabled, the harness deterministically spawns a read-only `context_scout` on the **first turn of every new thread** — it sweeps your memory tree, files, and connected data, assembles a bounded context bundle, and prepends it to your message before the model ever reads it. No tool call to wait on, no "let me look that up" round-trip: the agent answers your first message already knowing the relevant background. Toggle it from the composer or `context.super_context_enabled`. +### 🔬 The deep researcher & doer -- **[Goals & Todos](https://tinyhumans.gitbook.io/openhuman/features/goals-and-todos)**: OpenHuman keeps the agent pointed at what matters. A short, human-editable list of **long-term goals** (`MEMORY_GOALS.md`) rides along in memory and can self-reflect against your recent activity; each **thread** can carry a single durable **goal** with an optional token budget that the agent works across turns, interrupts and idle periods (autonomous idle continuation); and every conversation hosts a **kanban task board** of todos that you and the agent build together — plans, acceptance criteria, approval gates and all. There's also a personal task list you own outright. +- **[SuperContext](https://tinyhumans.gitbook.io/openhuman/features/super-context)**: a research scout sweeps your memory and files before the model reads your first message. No cold starts. +- **Batteries included**: web search, scraper, coder toolset, a real [browser](https://tinyhumans.gitbook.io/openhuman/features/native-tools/browser-and-computer), [native voice](gitbooks/features/native-tools/voice.md) with in-process Whisper — and [model routing](https://tinyhumans.gitbook.io/openhuman/features/model-routing) that picks the right LLM per workload, one subscription, [local AI optional](https://tinyhumans.gitbook.io/openhuman/features/model-routing/local-ai). +- **[Meeting agents](https://tinyhumans.gitbook.io/openhuman/features/mascot/meeting-agents)** _(new)_: joins **Meet, Zoom, Teams, and Webex** with a face and a voice — auto-joins from your calendar, streams a live transcript, answers by name, files summary + action items. +- **[Image & video generation](https://tinyhumans.gitbook.io/openhuman/features/native-tools)** _(new)_: Seedream/SeedEdit images and Seedance/Veo video, straight into your workspace on the same subscription. +- **[17 messaging channels](https://tinyhumans.gitbook.io/openhuman/features/channels)**: Telegram, Discord, Slack, WhatsApp, Signal, iMessage… plus **native email** (IMAP IDLE + SMTP). Your agent reaches you where you already are. -- **[Themes & Theme Studio](https://tinyhumans.gitbook.io/openhuman/features/theming)**: make it yours. Five built-in theme families (Classic, Ocean, Sepia, Matrix, HAL 9000) ship in light/dark/auto variants, and the **Theme Studio** in Settings is a full visual editor — adjust every colour token with live contrast warnings, swap fonts per role, pick an animated WebGL-mesh / flat / custom-image backdrop, and export or import themes as JSON to share. Editing a preset auto-forks a custom theme so the originals stay pristine; everything applies instantly and persists locally. +### 🧍 Human, private, yours -- **Batteries included**: web search, a web-fetch [scraper](https://tinyhumans.gitbook.io/openhuman/features/native-tools), a full coder toolset (filesystem, git, lint, test, grep), and [native voice](gitbooks/features/native-tools/voice.md) (STT in, ElevenLabs TTS out, mascot lip-sync, live Google Meet agent) are wired in by default. By default, [model routing](https://tinyhumans.gitbook.io/openhuman/features/model-routing) uses the OpenHuman backend to select and proxy the right LLM for each workload (reasoning, fast, or vision). One subscription includes all models. No "install a plugin to read files" friction. Use [optional local AI via Ollama](https://tinyhumans.gitbook.io/openhuman/features/model-routing/local-ai) for supported on-device workloads. - -- **[Smart token compression (TokenJuice)](https://tinyhumans.gitbook.io/openhuman/features/token-compression)**: every tool call, scrape result, email body, and search payload is run through a token compression layer before it touches any LLM Model. HTML is converted to Markdown, long URLs are shortened, and verbose tool output is deduped and summarized via a configurable rule overlay etc... CJK, emoji, and other multi-byte text are preserved grapheme-by-grapheme — never stripped. You get the same information but at a fraction of the tokens. Reducing cost & latency by up to 80%. - -- **[Messaging channels](https://tinyhumans.gitbook.io/openhuman/features/integrations#messaging-channels)** and **[privacy & security](https://tinyhumans.gitbook.io/openhuman/features/privacy-and-security)**: inbound/outbound across the channels you already use, with workflow data that stays on device, encrypted locally, treated as yours. +- **Simple, UI-first & Human**: install to working agent in a few clicks — no config files, no terminal. And it has [a face](https://tinyhumans.gitbook.io/openhuman/features/mascot): a mascot that speaks, reacts, and remembers you. +- **[Privacy & security](https://tinyhumans.gitbook.io/openhuman/features/privacy-and-security)**: on-device encrypted data, approval gate, OS-keyring secrets, opt-in sandboxing — and _(new)_ **[Privacy Mode](https://tinyhumans.gitbook.io/openhuman/features/privacy-mode)**: one switch and no inference leaves your machine, enforced in the Rust core. +- **[Themes & Theme Studio](https://tinyhumans.gitbook.io/openhuman/features/theming)**: five theme families plus a full visual editor, exportable as JSON. ## Contributing from source @@ -160,21 +165,38 @@ In just one sync pass, the agent has full (compressed) context of your inbox, yo Already self-host [agentmemory](https://github.com/rohitg00/agentmemory) across other coding agents? OpenHuman ships an optional `Memory` backend that proxies to it — set `memory.backend = "agentmemory"` in `config.toml` and the same durable store powers OpenHuman alongside Claude Code, Cursor, Codex, and OpenCode. See the [agentmemory backend](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/agentmemory-backend) page for setup. +## An orchestrator, not a chatbot + +The deepest difference isn't any single feature — it's the execution model. Claude Code, OpenClaw, and Hermes run **one agent in one loop**. OpenHuman is an **[orchestrator](https://tinyhumans.gitbook.io/openhuman/features/orchestration)**: + +- **Graphs, not loops** — turns compile to checkpointed state-machine graphs on [tinyagents](https://github.com/tinyhumansai/tinyagents): pause for a human, survive a restart, resume mid-run. +- **Sub-agent fleets** — specialists spawn three levels deep, idle workers get reused, and circuit breakers turn stuck agents into root-cause reports. +- **Workflows you can see** — agent-proposed [tinyflows](https://github.com/tinyhumansai/tinyflows) graphs, reviewed on a canvas: durable, trigger-driven, approval-gated. +- **A split brain, always on** — a fast reflex agent triages while a deep core reasons, steered by the subconscious loop. +- **Agent-to-agent, encrypted** — instances orchestrate each other over **Signal-protocol E2E** sessions with consent-based pairing and x402 payments. No server sees plaintext. +- **Next: RLMs** — the model writing its own orchestration code in a sandboxed REPL, on the same graph engine and trust model. + ## OpenHuman vs Other Agent Harnesses High-level comparison (products evolve, so verify against each vendor). OpenHuman is built to **minimize vendor sprawl**, keep **workflow knowledge on-device**, and give the agent a **persistent memory** of your data, not only chat. -| | Claude Cowork | OpenClaw | Hermes Agent | OpenHuman | -| ------------------- | ----------------- | ----------------- | ----------------- | ---------------------------------- | -| **Open-source** | 🚫 Proprietary | ✅ MIT | ✅ MIT | ✅ GNU | -| **Simple to start** | ✅ Desktop + CLI | ⚠️ Terminal-first | ⚠️ Terminal-first | ✅ Clean UI, minutes | -| **Cost** | ⚠️ Sub + add-ons | ⚠️ BYO models | ⚠️ BYO models | ✅ One sub + TokenJuice | -| **Memory** | ✅ Chat-scoped | ⚠️ Plugin-reliant | ✅ Self-learning | 🚀 Memory Tree + Obsidian vault, optional [agentmemory](https://github.com/rohitg00/agentmemory) backend | -| **Integrations** | ⚠️ Few connectors | ⚠️ BYO | ⚠️ BYO | 🚀 100+ OAuth · 5k+ MCP · 90k+ Skills | -| **Auto-fetch** | 🚫 None | 🚫 None | 🚫 None | ✅ 20-min sync into memory | -| **API sprawl** | 🚫 Extra keys | 🚫 BYOK | 🚫 Multi-vendor | ✅ One account | -| **Model routing** | 🚫 Single model | ⚠️ Manual | ⚠️ Manual | ✅ Built-in | -| **Native tools** | ✅ Code-only | ✅ Code-only | ✅ Code-only | ✅ Code + search + scraper + voice | +| | Claude Cowork | OpenClaw | Hermes Agent | OpenHuman | +| ---------------------- | ----------------- | ----------------- | ----------------- | -------------------------------------------------------------------------------------------------------- | +| **Open-source** | 🚫 Proprietary | ✅ MIT | ✅ MIT | ✅ GNU | +| **Simple to start** | ✅ Desktop + CLI | ⚠️ Terminal-first | ⚠️ Terminal-first | ✅ Clean UI, minutes | +| **Cost** | ⚠️ Sub + add-ons | ⚠️ BYO models | ⚠️ BYO models | ✅ One sub + TokenJuice | +| **Memory** | ✅ Chat-scoped | ⚠️ Plugin-reliant | ✅ Self-learning | 🚀 Memory Tree + Obsidian vault, optional [agentmemory](https://github.com/rohitg00/agentmemory) backend | +| **Integrations** | ⚠️ Few connectors | ⚠️ BYO | ⚠️ BYO | 🚀 100+ OAuth · 5k+ MCP · 90k+ Skills | +| **Auto-fetch** | 🚫 None | 🚫 None | 🚫 None | ✅ 20-min sync into memory | +| **Orchestration** | ⚠️ Sub-tasks | ⚠️ Single loop | ⚠️ Single loop | 🚀 Agent graphs + checkpoints + E2E-encrypted A2A | +| **Workflows** | 🚫 None | ⚠️ Scripts | ⚠️ Scripts | 🚀 Visual, durable, agent-proposed, approval-gated | +| **Meetings** | 🚫 None | 🚫 None | 🚫 None | 🚀 Joins Meet/Zoom/Teams/Webex, speaks, live transcript | +| **Messaging channels** | 🚫 None | ⚠️ A few | ⚠️ A few | ✅ 17 incl. native email (IMAP/SMTP) | +| **Local-only mode** | 🚫 Cloud-only | ⚠️ BYO local | ⚠️ BYO local | ✅ One-switch enforced Privacy Mode | +| **Observability** | 🚫 Opaque | ⚠️ Logs | ⚠️ Logs | ✅ Replayable run journals + per-call cost accounting | +| **API sprawl** | 🚫 Extra keys | 🚫 BYOK | 🚫 Multi-vendor | ✅ One account | +| **Model routing** | 🚫 Single model | ⚠️ Manual | ⚠️ Manual | ✅ Built-in | +| **Native tools** | ✅ Code-only | ✅ Code-only | ✅ Code-only | ✅ Code + search + scraper + browser + voice + media gen | # Star us on GitHub diff --git a/docs/README.de.md b/docs/README.de.md index 9ab3b9644..2452aa94e 100644 --- a/docs/README.de.md +++ b/docs/README.de.md @@ -1,8 +1,3 @@ -

- 🇺🇸 English | 🇨🇳 简体中文 | 🇯🇵 日本語 | 🇰🇷 한국어 | 🇩🇪 Deutsch | 🇵🇰 اردو -

- -

OpenHuman

@@ -10,29 +5,41 @@

- - tinyhumansai%2Fopenhuman | Trendshift - -   - - OpenHuman - An open source AI harness built with the human in mind | Product Hunt - - -

- -

- OpenHuman ist deine persönliche KI-Superintelligenz: Lokaler Speicher, verwaltete Dienste wo nötig, schlicht und mächtig. + + tinyhumansai%2Fopenhuman | Trendshift + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt +

+

+ + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + +

+

+ OpenHuman ist deine persönliche KI-Superintelligenz: ein Gehirn, das sich alles merkt, ein fantastischer Orchestrator, ein Deep Researcher. Local-first, schlicht, mächtig. +

Discord • - Discussions • + Reddit • X/Twitter • Doku • @senamakel folgen (Creator)

+

+ 🇺🇸 English | 🇨🇳 简体中文 | 🇯🇵 日本語 | 🇰🇷 한국어 | 🇩🇪 Deutsch | 🇵🇰 اردو +

+

Frühe Beta Aktuellste Version @@ -48,47 +55,94 @@ > **Frühe Beta**: Wird aktiv weiterentwickelt. Mit Ecken und Kanten ist zu rechnen. -> **Lokal + verwaltete Dienste, upfront:** OpenHuman speichert seinen Memory Tree, Obsidian-Style-Markdown-Vault, Workspace-Konfiguration und lokalen Laufzeitstatus auf deiner Maschine. Die standardmäßige verwaltete Erfahrung nutzt weiterhin OpenHuman-gehostete Dienste für Account-Anmeldung, Model-Routing, Web-Search-Proxying und verwaltete Integration/OAuth-Flows über die Composio-Connector-Schicht. Wähle benutzerdefinierte/lokale Einstellungen, wenn du dein eigenes Modell, deine eigene Suche oder Composio-Credentials mitbringen möchtest; einige Echtzeit-Trigger und gehostete Funktionen erfordern weiterhin das verwaltete Backend. +> 🎉 Innerhalb einer Woche nach dem Launch wurde OpenHuman neun Tage in Folge das Nummer-eins-Trending-Repository auf GitHub. -Für Installation und Einstieg lade die App von [tinyhumans.ai/openhuman](https://tinyhumans.ai/openhuman?utm_source=github&utm_medium=readme) herunter oder führe im Terminal aus: +# Installation + +Lade die Installer von [tinyhumans.ai/openhuman](https://tinyhumans.ai/openhuman?utm_source=github&utm_medium=readme) oder von der [GitHub-Releases](https://github.com/tinyhumansai/openhuman/releases/latest)-Seite herunter. Für Terminal-Installationen sind die nativen Paketpfade unten zu bevorzugen, da sie — wo verfügbar — den Paketmanager bzw. nativen Installer deines Betriebssystems nutzen. + +## Empfohlene Installation (native Pakete) + +Diese Pfade nutzen native Installer-Oberflächen. Homebrew und MSI bringen ihre üblichen Signatur-/Integritätsprüfungen mit; unter Debian/Ubuntu installiert `apt-get` das Release-`.deb` und löst die Systemabhängigkeiten auf. + +**macOS (Homebrew-Tap):** ```bash -# DMG/EXE über https://tinyhumans.ai/openhuman herunterladen oder direkt im Terminal: +brew tap tinyhumansai/core +brew install openhuman +``` -# macOS oder Linux x64 +**Linux (Debian/Ubuntu — Release-`.deb`):** + +```bash +# Download OpenHuman__amd64.deb or OpenHuman__arm64.deb +# from https://github.com/tinyhumansai/openhuman/releases/latest, then: +# Replace amd64 with arm64 on arm64 hosts. +sudo apt-get install -y --no-install-recommends ./OpenHuman_*_amd64.deb +``` + +**Linux (Arch — AUR):** das [`openhuman-bin`-AUR-Rezept](../packages/arch/openhuman-bin/) liegt im Repo. Sobald es veröffentlicht ist, können Arch-Nutzer es mit `yay -S openhuman-bin` installieren. + +**Windows:** lade die signierte `.msi` vom [neuesten Release](https://github.com/tinyhumansai/openhuman/releases/latest) herunter und führe sie aus. + +**Manuell `.dmg` / `.deb` / `.AppImage` / `.msi`:** hole dir den Installer für deine Plattform direkt von der [Seite des neuesten Releases](https://github.com/tinyhumansai/openhuman/releases/latest). + +> **Linux:** das AppImage kann unter Wayland beim Start abstürzen, Host-Systembibliotheken wie `libgbm.so.1` vermissen oder auf Arch-basierten Distros mit `sharun: Interpreter not found!` fehlschlagen — siehe [#2463](https://github.com/tinyhumansai/openhuman/issues/2463) für die Ursache und Env-Var-Workarounds. Das `.deb`-Paket oben vermeidet diese Fehlerbilder auf Debian/Ubuntu, indem apt die Laufzeitabhängigkeiten auflöst. + +## Alternative: Skript-Installation (ohne Integritätsprüfung) + +> **Warnung — unverifizierte Installation.** Diese Skripte werden live von `raw.githubusercontent.com` ausgeliefert und bringen **keine** separate Signatur mit — `curl … | bash` und `irm … | iex` haben also keine Möglichkeit, Manipulationen an den Skript-Bytes zu erkennen. Bevorzuge wann immer möglich die **nativen Paketpfade** oben. Wenn du das Skript dennoch nutzen musst, siehe „Verifizierte Skript-Installation" unten. + +```bash +# macOS or Linux x64 curl -fsSL https://raw.githubusercontent.com/tinyhumansai/openhuman/main/scripts/install.sh | bash -# Windows +# Windows (PowerShell) irm https://raw.githubusercontent.com/tinyhumansai/openhuman/main/scripts/install.ps1 | iex ``` - -> **Linux:** the AppImage can crash on launch under Wayland (and on Arch-based distros with `sharun: Interpreter not found!`) — see [#2463](https://github.com/tinyhumansai/openhuman/issues/2463) for the cause and env-var workarounds. -Arch Linux package maintainers can use the [`openhuman-bin` AUR recipe](../packages/arch/openhuman-bin/); -once published, Arch users can install it with `yay -S openhuman-bin`. - +Unter Debian/Ubuntu löst `install.sh` zuerst das neueste Release-`.deb` auf und installiert es mit `apt-get`, sodass apt die Laufzeitabhängigkeiten übernimmt. Setze `OPENHUMAN_INSTALLER_LINUX_PACKAGE=appimage`, um den AppImage-Pfad zu erzwingen. + +## Status der verifizierten Skript-Installation + +Ein separat signierter Skript-Installationspfad ist derzeit nicht verfügbar. Issue [#2620](https://github.com/tinyhumansai/openhuman/issues/2620) wurde geschlossen, nachdem die nativen Paketpfade in den Vordergrund gerückt wurden, aber die aktuellen Release-Assets enthalten kein `install.sh.asc` / `install.ps1.asc` zur Skriptverifikation vor der Ausführung. Behandle die Skript-Installation als unverifiziert und bevorzuge nach Möglichkeit die nativen Paketoptionen oben. # Was ist OpenHuman? -OpenHuman ist ein quelloffener, agentenbasierter Assistent, der sich in deinen Alltag einfügt. Jeder Punkt verlinkt auf die ausführliche Beschreibung in der [Doku](https://tinyhumans.gitbook.io/openhuman/). +OpenHuman ist drei Dinge, die die meisten Assistenten nicht sind: **ein Gehirn**, das ein persistentes, lokales Gedächtnis deiner Welt aufbaut; **ein fantastischer Orchestrator**, der Flotten von Agenten auf dauerhaften Graphen betreibt; und **ein Deep Researcher**, der deine Daten und das Web durchkämmt, bevor du deine Frage zu Ende gestellt hast. Jeder Punkt verlinkt auf die ausführliche Beschreibung in der [Doku](https://tinyhumans.gitbook.io/openhuman/). -- **Schlicht, UI-zuerst & menschlich** — Eine aufgeräumte Desktop-Erfahrung und kurze Onboarding-Pfade bringen dich in wenigen Klicks von der Installation zum laufenden Agenten — keine Config-First-Einrichtung, kein Terminal nötig. Der Agent hat [ein Gesicht](https://tinyhumans.gitbook.io/openhuman/features/mascot): ein Desktop-Maskottchen, das spricht, auf seine Umgebung reagiert, als echter Teilnehmer [in deinen Google-Meets sitzt](https://tinyhumans.gitbook.io/openhuman/features/mascot/meeting-agents), sich über Wochen an dich erinnert und im Hintergrund weiterdenkt, auch wenn du längst nicht mehr tippst. +### 🧠 Das Gehirn -- **[118+ Drittanbieter-Integrationen](https://tinyhumans.gitbook.io/openhuman/features/integrations) mit [Auto-Fetch](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/auto-fetch)**: Gmail, Notion, GitHub, Slack, Stripe, Calendar, Drive, Linear, Jira und der Rest deines Stacks per **Ein-Klick-OAuth** anbinden. Jede Verbindung wird dem Agenten als typisiertes Tool freigegeben, und alle zwanzig Minuten geht der Core durch jede aktive Verbindung und zieht frische Daten in den [Memory Tree](https://tinyhumans.gitbook.io/openhuman/features/integrations/auto-fetch). Keine Prompts, keine Polling-Schleifen, die du selbst schreiben musst — der Agent hat morgens schon den Kontext für den Tag. +- **[Memory Tree](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) + [Obsidian-Wiki](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki)**: deine Daten, komprimiert in bewertete Markdown-Bäume in SQLite auf deiner Maschine, gespiegelt als [Obsidian-Vault](https://x.com/karpathy/status/2039805659525644595), das du öffnen und editieren kannst. Keine Vektor-Suppen-Blackbox. +- **[100+ OAuth-Integrationen, 5.000+ MCP-Server, 90.000+ Skills](https://tinyhumans.gitbook.io/openhuman/features/integrations)**: mit einem Klick in Gmail, Notion, GitHub, Slack und den Rest deines Stacks. [Auto-Fetch](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/auto-fetch) füttert das Gehirn alle 20 Minuten — es hat den Kontext von morgen schon heute Früh. +- **[Ein Unterbewusstsein](https://tinyhumans.gitbook.io/openhuman/features/subconscious)**: eine Hintergrundschleife, die Veränderungen in deiner Welt erkennt, deine Ziele vorantreibt und dein Morgen-Briefing schreibt — das Denken geht weiter, auch wenn du längst nicht mehr tippst. +- **[Goals & Todos](https://tinyhumans.gitbook.io/openhuman/features/goals-and-todos)**: Langzeitziele, dauerhafte Ziele pro Thread und ein geteiltes Kanban-Board pro Unterhaltung. +- **[TokenJuice](https://tinyhumans.gitbook.io/openhuman/features/token-compression)**: Tool-Ausgaben werden komprimiert, bevor sie das Modell erreichen — dieselbe Information, bis zu 80% weniger Tokens. Ein so großes Gehirn wäre ohne es unbezahlbar. - Verwaltete Integrationen nutzen OpenHumans Composio-Connector-Schicht. OAuth-Handshakes und Integration-Tool-Calls werden standardmäßig über das verwaltete Backend geproxied. Wenn du stattdessen Composio direkt betreiben möchtest, konfiguriere den Direktmodus mit deinem eigenen Composio-API-Key; Echtzeit-Trigger-Webhooks müssen dann von dir selbst gehostet und verkabelt werden. +### 🕸️ Der Orchestrator -- **[Memory Tree](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) + [Obsidian-Wiki](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki)**: eine lokal-zentrierte Wissensbasis, aufgebaut aus deinen Daten und deinen Aktivitäten. Alles, was du verbindest, wird in Markdown-Chunks von ≤3k Tokens kanonisiert, bewertet und in hierarchische Zusammenfassungs-Bäume gefaltet, gespeichert in **SQLite auf deiner Maschine**. Dieselben Chunks landen als `.md`-Dateien in einem Obsidian-kompatiblen Vault, das du öffnen, durchstöbern und editieren kannst — inspiriert von Karpathys [obsidian-wiki-Workflow](https://x.com/karpathy/status/2039805659525644595). +- **[Workflows](https://tinyhumans.gitbook.io/openhuman/features/workflows)** _(neu)_: der Agent schlägt die Automatisierung vor; du prüfst sie auf einer Canvas und speicherst. Dauerhafte, trigger-gesteuerte, freigabe-gesicherte Läufe auf dem quelloffenen [tinyflows](https://github.com/tinyhumansai/tinyflows). +- **[Ein Harness, das die Arbeit zu Ende bringt](https://tinyhumans.gitbook.io/openhuman/developing/architecture/agent-harness)** _(neu)_: checkpointed Graph-Läufe auf dem quelloffenen [tinyagents](https://github.com/tinyhumansai/tinyagents) — festgefahrene Agenten werden nachgesteuert, gestoppte liefern eine Root Cause zurück, jeder Lauf lässt sich mit echten Kosten pro Aufruf wieder abspielen. +- **[Ein Split Brain, immer an](https://tinyhumans.gitbook.io/openhuman/features/orchestration)** _(neu)_: ein schneller Reflex-Agent triagiert eingehenden Traffic, während ein tiefer Reasoning-Kern an Worker-Flotten delegiert, gesteuert vom Unterbewusstsein. +- **[Eine Agenten-Ökonomie](https://tinyhumans.gitbook.io/openhuman/features/tinyplace)**: ein `@handle` auf [tiny.place](https://tiny.place), Signal-verschlüsselte Agent-zu-Agent-Orchestrierung, x402-USDC-Bounties und Handel — Keys berühren nie die Festplatte. -- **Alles eingebaut**: Web-Suche, ein Web-Fetch-[Scraper](https://tinyhumans.gitbook.io/openhuman/features/native-tools), ein vollständiges Coder-Toolset (Dateisystem, Git, Lint, Test, Grep) und [native Sprache](https://tinyhumans.gitbook.io/openhuman/features/native-tools/voice) (STT als Eingabe, ElevenLabs TTS als Ausgabe, Lippensynchronisation für das Maskottchen, Live-Google-Meet-Agent) sind ab Werk verdrahtet. Standardmäßig nutzt [Model-Routing](https://tinyhumans.gitbook.io/openhuman/features/model-routing) das OpenHuman-Backend, um das passende LLM für jede Workload auszuwählen und zu proxien (Reasoning, Fast oder Vision). Ein Abo umfasst alle Modelle. Keine "erst-ein-Plugin-installieren-um-Dateien-zu-lesen"-Hürde. [Optional lokale KI über Ollama](https://tinyhumans.gitbook.io/openhuman/features/model-routing/local-ai) für On-Device-Workloads. +### 🔬 Der Deep Researcher & Macher -- **[Smarte Token-Kompression (TokenJuice)](https://tinyhumans.gitbook.io/openhuman/features/token-compression)**: Jeder Tool-Aufruf, jedes Scrape-Ergebnis, jeder E-Mail-Text und jeder Such-Payload läuft durch eine Token-Kompressionsschicht, bevor er ein LLM-Modell erreicht. HTML wird zu Markdown konvertiert, lange URLs werden gekürzt, und ausschweifende Tool-Ausgaben werden über eine konfigurierbare Regel-Ebene dedupliziert und zusammengefasst usw. CJK, Emojis und andere Multi-Byte-Texte bleiben Graphem für Graphem erhalten — niemals abgeschnitten. Du erhältst dieselbe Information bei einem Bruchteil der Tokens. Kosten und Latenz sinken um bis zu 80%. +- **[SuperContext](https://tinyhumans.gitbook.io/openhuman/features/super-context)**: ein Research-Scout durchkämmt dein Gedächtnis und deine Dateien, bevor das Modell deine erste Nachricht liest. Keine Kaltstarts. +- **Alles eingebaut**: Web-Suche, Scraper, Coder-Toolset, ein echter [Browser](https://tinyhumans.gitbook.io/openhuman/features/native-tools/browser-and-computer), [native Sprache](../gitbooks/features/native-tools/voice.md) mit In-Process-Whisper — und [Model-Routing](https://tinyhumans.gitbook.io/openhuman/features/model-routing), das das passende LLM pro Workload auswählt, ein Abo, [lokale KI optional](https://tinyhumans.gitbook.io/openhuman/features/model-routing/local-ai). +- **[Meeting-Agenten](https://tinyhumans.gitbook.io/openhuman/features/mascot/meeting-agents)** _(neu)_: nimmt an **Meet, Zoom, Teams und Webex** teil — mit Gesicht und Stimme. Tritt automatisch aus deinem Kalender bei, streamt ein Live-Transkript, antwortet auf seinen Namen und legt Zusammenfassung + Action Items ab. +- **[Bild- & Videogenerierung](https://tinyhumans.gitbook.io/openhuman/features/native-tools)** _(neu)_: Seedream/SeedEdit-Bilder und Seedance/Veo-Video, direkt in deinen Workspace im selben Abo. +- **[17 Messaging-Kanäle](https://tinyhumans.gitbook.io/openhuman/features/channels)**: Telegram, Discord, Slack, WhatsApp, Signal, iMessage… plus **native E-Mail** (IMAP IDLE + SMTP). Dein Agent erreicht dich dort, wo du ohnehin schon bist. -- **[Messaging-Kanäle](https://tinyhumans.gitbook.io/openhuman/features/integrations#messaging-channels)** und **[Privatsphäre & Sicherheit](https://tinyhumans.gitbook.io/openhuman/features/privacy-and-security)**: ein- und ausgehend über die Kanäle, die du ohnehin nutzt — Workflow-Daten bleiben auf dem Gerät, lokal verschlüsselt, und gehören dir. +### 🧍 Menschlich, privat, deins + +- **Schlicht, UI-zuerst & menschlich**: von der Installation zum laufenden Agenten in wenigen Klicks — keine Config-Dateien, kein Terminal. Und es hat [ein Gesicht](https://tinyhumans.gitbook.io/openhuman/features/mascot): ein Maskottchen, das spricht, reagiert und sich an dich erinnert. +- **[Privatsphäre & Sicherheit](https://tinyhumans.gitbook.io/openhuman/features/privacy-and-security)**: verschlüsselte Daten auf dem Gerät, Freigabe-Gate, Secrets im OS-Schlüsselbund, Opt-in-Sandboxing — und _(neu)_ **[Privacy Mode](https://tinyhumans.gitbook.io/openhuman/features/privacy-mode)**: ein Schalter, und keine Inferenz verlässt deine Maschine — erzwungen im Rust-Core. +- **[Themes & Theme Studio](https://tinyhumans.gitbook.io/openhuman/features/theming)**: fünf Theme-Familien plus ein vollständiger visueller Editor, exportierbar als JSON. ## Beitragen aus dem Quellcode -Neu hier? Beginne mit [`CONTRIBUTING.md`](../CONTRIBUTING.md) für den Fork-/PR-Workflow und die lokalen Prüfbefehle. Der kurze Weg: +Neu hier? Beginne mit [`CONTRIBUTING.md`](../CONTRIBUTING.md) für den Fork-/PR-Workflow und die lokalen Prüfbefehle, oder nutze den Copy-Paste-Prompt für KI-Coding-Agenten in [`CONTRIBUTING-BEGINNERS.md`](../CONTRIBUTING-BEGINNERS.md#optional-let-an-ai-coding-agent-guide-you). Der kurze Weg: 1. Installiere Git, Node.js 24+, pnpm 10.10.0, Rust 1.93.0 (`rustfmt` + `clippy`), CMake, Ninja, ripgrep sowie die plattformspezifischen Desktop-Build-Voraussetzungen. 2. Forke und klone das Repo, führe dann `git submodule update --init --recursive` aus, bevor du `pnpm install` startest, damit die mitgelieferten Tauri/CEF-Quellen vorhanden sind. @@ -101,39 +155,56 @@ Tiefer einsteigen: [Architektur](https://tinyhumans.gitbook.io/openhuman/develop OpenHuman ist das erste Agent-Harness, das dich in Minuten kennenlernt. Inspiriert von [Karpathys LLM-Knowledgebase](https://x.com/karpathy/status/2039805659525644595). Die meisten Agenten starten aus dem Kalten. Hermes lernt, indem er dir bei der Arbeit zusieht; OpenClaw wartet darauf, dass Plugins Kontext einspielen. So oder so vergehen Tage oder Wochen, bevor der Agent genug über deinen Stack weiß, um wirklich nützlich zu sein.

- Diagramm zum OpenHuman-Kontextaufbau + Diagramm zum OpenHuman-Kontextaufbau

> OpenHuman fasst all deine Dokumente, E-Mails und Chats zusammen, komprimiert sie und legt einen Memory Graph an, mit dem dein Agent sich alles über dich merken kann. OpenHuman überspringt die Wartezeit. Verbinde deine Accounts, lass [Auto-Fetch](https://tinyhumans.gitbook.io/openhuman/features/integrations/auto-fetch) die Daten lokal in einer 20-Minuten-Schleife abholen, und [Memory Trees](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) komprimieren alles in Markdown-Dateien, intelligent abgelegt in einem [Obsidian-Wiki im Karpathy-Stil](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki). -Nach nur einem Sync-Durchlauf hat der Agent den vollständigen (komprimierten) Kontext deines Postfachs, deines Kalenders, deiner Repos, deiner Dokumente und deiner Nachrichten. Keine Trainingsphase. Kein "gib ihm ein paar Wochen". Er wird zu dir — gesteuert von dir. +Nach nur einem Sync-Durchlauf hat der Agent den vollständigen (komprimierten) Kontext deines Postfachs, deines Kalenders, deiner Repos, deiner Dokumente und deiner Nachrichten. Keine Trainingsphase. Kein „gib ihm ein paar Wochen". Er wird zu dir — gesteuert von dir. Du hostest [agentmemory](https://github.com/rohitg00/agentmemory) bereits selbst für andere Coding-Agenten? OpenHuman bringt ein optionales `Memory`-Backend mit, das dorthin proxyt — setze `memory.backend = "agentmemory"` in `config.toml`, und derselbe persistente Store treibt OpenHuman zusammen mit Claude Code, Cursor, Codex und OpenCode an. Setup-Details auf der Seite zum [agentmemory-Backend](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/agentmemory-backend). +## Ein Orchestrator, kein Chatbot + +Der tiefste Unterschied ist kein einzelnes Feature — es ist das Ausführungsmodell. Claude Code, OpenClaw und Hermes betreiben **einen Agenten in einer Schleife**. OpenHuman ist ein **[Orchestrator](https://tinyhumans.gitbook.io/openhuman/features/orchestration)**: + +- **Graphen statt Schleifen** — Turns kompilieren zu checkpointed State-Machine-Graphen auf [tinyagents](https://github.com/tinyhumansai/tinyagents): für einen Menschen pausieren, einen Neustart überleben, mitten im Lauf weitermachen. +- **Sub-Agenten-Flotten** — Spezialisten spawnen drei Ebenen tief, untätige Worker werden wiederverwendet, und Circuit Breaker verwandeln festgefahrene Agenten in Root-Cause-Berichte. +- **Workflows, die du sehen kannst** — vom Agenten vorgeschlagene [tinyflows](https://github.com/tinyhumansai/tinyflows)-Graphen, geprüft auf einer Canvas: dauerhaft, trigger-gesteuert, freigabe-gesichert. +- **Ein Split Brain, immer an** — ein schneller Reflex-Agent triagiert, während ein tiefer Kern nachdenkt, gesteuert von der Unterbewusstseins-Schleife. +- **Agent-zu-Agent, verschlüsselt** — Instanzen orchestrieren sich gegenseitig über **Signal-Protokoll-E2E**-Sitzungen mit einwilligungsbasiertem Pairing und x402-Zahlungen. Kein Server sieht Klartext. +- **Als Nächstes: RLMs** — das Modell schreibt seinen eigenen Orchestrierungscode in einer gesandboxten REPL, auf derselben Graph-Engine und demselben Vertrauensmodell. + ## OpenHuman vs. andere Agent-Harnesses Übersichtsvergleich (Produkte entwickeln sich weiter — bitte beim jeweiligen Anbieter verifizieren). OpenHuman ist darauf ausgelegt, **Vendor-Wildwuchs zu reduzieren**, **Workflow-Wissen auf dem Gerät zu halten** und dem Agenten eine **persistente Erinnerung** an deine Daten zu geben — nicht nur an den Chat. -| | Claude Cowork | OpenClaw | Hermes Agent | OpenHuman | -| ------------------- | ----------------- | ----------------- | ----------------- | ---------------------------------- | -| **Quelloffen** | 🚫 Proprietär | ✅ MIT | ✅ MIT | ✅ GNU | -| **Einfacher Einstieg** | ✅ Desktop + CLI | ⚠️ Terminal zuerst | ⚠️ Terminal zuerst | ✅ Aufgeräumte UI, in Minuten | -| **Kosten** | ⚠️ Abo + Zusatzkosten | ⚠️ BYO-Modelle | ⚠️ BYO-Modelle | ✅ Ein Abo + TokenJuice | -| **Memory** | ✅ chat-gebunden | ⚠️ plugin-abhängig | ✅ selbstlernend | 🚀 Memory Tree + Obsidian-Vault, optional [agentmemory](https://github.com/rohitg00/agentmemory)-Backend | -| **Integrationen** | ⚠️ wenige Konnektoren | ⚠️ BYO | ⚠️ BYO | 🚀 118+ über OAuth | -| **Auto-Fetch** | 🚫 keiner | 🚫 keiner | 🚫 keiner | ✅ 20-Min.-Sync ins Memory | -| **API-Wildwuchs** | 🚫 zusätzliche Keys | 🚫 BYOK | 🚫 Multi-Vendor | ✅ ein Account | -| **Model-Routing** | 🚫 nur ein Modell | ⚠️ manuell | ⚠️ manuell | ✅ eingebaut | -| **Native Tools** | ✅ nur Code | ✅ nur Code | ✅ nur Code | ✅ Code + Suche + Scraper + Sprache | +| | Claude Cowork | OpenClaw | Hermes Agent | OpenHuman | +| ---------------------- | ---------------------- | ------------------ | ------------------ | --------------------------------------------------------------------------------------------------------- | +| **Quelloffen** | 🚫 Proprietär | ✅ MIT | ✅ MIT | ✅ GNU | +| **Einfacher Einstieg** | ✅ Desktop + CLI | ⚠️ Terminal zuerst | ⚠️ Terminal zuerst | ✅ Aufgeräumte UI, in Minuten | +| **Kosten** | ⚠️ Abo + Zusatzkosten | ⚠️ BYO-Modelle | ⚠️ BYO-Modelle | ✅ Ein Abo + TokenJuice | +| **Memory** | ✅ chat-gebunden | ⚠️ plugin-abhängig | ✅ selbstlernend | 🚀 Memory Tree + Obsidian-Vault, optional [agentmemory](https://github.com/rohitg00/agentmemory)-Backend | +| **Integrationen** | ⚠️ wenige Konnektoren | ⚠️ BYO | ⚠️ BYO | 🚀 100+ OAuth · 5k+ MCP · 90k+ Skills | +| **Auto-Fetch** | 🚫 keiner | 🚫 keiner | 🚫 keiner | ✅ 20-Min.-Sync ins Memory | +| **Orchestrierung** | ⚠️ Sub-Tasks | ⚠️ eine Schleife | ⚠️ eine Schleife | 🚀 Agent-Graphen + Checkpoints + E2E-verschlüsseltes A2A | +| **Workflows** | 🚫 keine | ⚠️ Skripte | ⚠️ Skripte | 🚀 visuell, dauerhaft, agent-vorgeschlagen, freigabe-gesichert | +| **Meetings** | 🚫 keine | 🚫 keine | 🚫 keine | 🚀 nimmt an Meet/Zoom/Teams/Webex teil, spricht, Live-Transkript | +| **Messaging-Kanäle** | 🚫 keine | ⚠️ einige wenige | ⚠️ einige wenige | ✅ 17 inkl. nativer E-Mail (IMAP/SMTP) | +| **Nur-lokal-Modus** | 🚫 nur Cloud | ⚠️ BYO lokal | ⚠️ BYO lokal | ✅ per Schalter erzwungener Privacy Mode | +| **Observability** | 🚫 undurchsichtig | ⚠️ Logs | ⚠️ Logs | ✅ wiederabspielbare Lauf-Journale + Kostenabrechnung pro Aufruf | +| **API-Wildwuchs** | 🚫 zusätzliche Keys | 🚫 BYOK | 🚫 Multi-Vendor | ✅ ein Account | +| **Model-Routing** | 🚫 nur ein Modell | ⚠️ manuell | ⚠️ manuell | ✅ eingebaut | +| **Native Tools** | ✅ nur Code | ✅ nur Code | ✅ nur Code | ✅ Code + Suche + Scraper + Browser + Sprache + Mediengenerierung | # Gib uns einen Stern auf GitHub _Baust du auch in Richtung AGI und künstlichem Bewusstsein? Setze einen Stern und hilf anderen, den Weg zu finden._

- + diff --git a/docs/README.ja-JP.md b/docs/README.ja-JP.md index 96a61d27e..d0948c434 100644 --- a/docs/README.ja-JP.md +++ b/docs/README.ja-JP.md @@ -5,24 +5,32 @@

- - tinyhumansai%2Fopenhuman | Trendshift - -   - - OpenHuman - An open source AI harness built with the human in mind | Product Hunt - - -

- -

- OpenHuman はあなたのパーソナル AI スーパーインテリジェンスです：ローカルメモリ、必要に応じてマネージドサービス、シンプルで強力。 + + tinyhumansai%2Fopenhuman | Trendshift + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt +

+

+ + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + +

+

+ OpenHuman はあなたのパーソナル AI スーパーインテリジェンスです：すべてを記憶する脳、素晴らしいオーケストレーター、ディープリサーチャー。ローカルファーストで、シンプルで、強力。 +

Discord • - Discussions • + Reddit • X/Twitter • ドキュメント • @senamakel（作者）をフォロー @@ -32,8 +40,6 @@ 🇺🇸 English | 🇨🇳 简体中文 | 🇯🇵 日本語 | 🇰🇷 한국어 | 🇩🇪 Deutsch | 🇵🇰 اردو

- -

Early Beta 最新リリース @@ -48,49 +54,96 @@ > **早期ベータ版**: 現在も活発に開発中です。荒削りな部分があることをご了承ください。 -> **ローカル + マネージドサービス、upfront:** OpenHuman は Memory Tree、Obsidian スタイルの Markdown ヴォルト、ワークスペース設定、およびローカルランタイム状態をあなたのマシン上に保存します。デフォルトのマネージド体験では、アカウントサインイン、モデルルーティング、Web 検索プロキシ、および Composio コネクタレイヤーを介したマネージド統合/OAuth フローに、OpenHuman ホスト型サービスが引き続き使用されます。独自のモデル、検索、または Composio 認証情報を持ち込みたい場合は、カスタム/ローカル設定を選択してください。一部のリアルタイムトリガーおよびホスト型機能には、マネージドバックエンドが引き続き必要です。 +> 🎉 ローンチから 1 週間以内に、OpenHuman は GitHub のトレンドリポジトリ第 1 位を 9 日間連続で獲得しました。 -インストールや利用開始は、ウェブサイト [tinyhumans.ai/openhuman](https://tinyhumans.ai/openhuman?utm_source=github&utm_medium=readme) からダウンロードするか、以下のコマンドを実行してください。 +# インストール + +インストーラーは [tinyhumans.ai/openhuman](https://tinyhumans.ai/openhuman?utm_source=github&utm_medium=readme) または [GitHub Releases](https://github.com/tinyhumansai/openhuman/releases/latest) ページからダウンロードできます。ターミナルからインストールする場合は、可能な限り OS のパッケージマネージャーやネイティブインストーラーを使用する、以下のネイティブパッケージ経路を推奨します。 + +## 推奨インストール（ネイティブパッケージ） + +これらの経路はネイティブなインストーラー面を使用します。Homebrew と MSI は通常の署名/整合性チェックを提供し、Debian/Ubuntu では `apt-get` がリリース版の `.deb` をインストールしてシステム依存関係を解決します。 + +**macOS（Homebrew tap）:** ```bash -# DMG や EXE は https://tinyhumans.ai/openhuman からダウンロードするか、ターミナルから実行してください +brew tap tinyhumansai/core +brew install openhuman +``` -# macOS または Linux x64 の場合 +**Linux（Debian/Ubuntu — リリース版 `.deb`）:** + +```bash +# Download OpenHuman__amd64.deb or OpenHuman__arm64.deb +# from https://github.com/tinyhumansai/openhuman/releases/latest, then: +# Replace amd64 with arm64 on arm64 hosts. +sudo apt-get install -y --no-install-recommends ./OpenHuman_*_amd64.deb +``` + +**Linux（Arch — AUR）:** [`openhuman-bin` AUR レシピ](../packages/arch/openhuman-bin/)はリポジトリ内にあります。公開後、Arch ユーザーは `yay -S openhuman-bin` でインストールできます。 + +**Windows:** [最新リリース](https://github.com/tinyhumansai/openhuman/releases/latest)から署名済みの `.msi` をダウンロードして実行してください。 + +**手動での `.dmg` / `.deb` / `.AppImage` / `.msi` インストール:** お使いのプラットフォーム向けのインストーラーを[最新リリースページ](https://github.com/tinyhumansai/openhuman/releases/latest)から直接入手してください。 + +> **Linux:** AppImage は Wayland 環境で起動時にクラッシュしたり、`libgbm.so.1` などのホストシステムライブラリが見つからなかったり、Arch 系ディストリビューションで `sharun: Interpreter not found!` により失敗することがあります。原因と環境変数による回避策は [#2463](https://github.com/tinyhumansai/openhuman/issues/2463) を参照してください。上記の `.deb` パッケージなら、apt がランタイム依存関係を解決するため、Debian/Ubuntu ではこれらの失敗を回避できます。 + +## 代替手段: スクリプトインストール（整合性チェックなし） + +> **警告 — 未検証のインストールです。** これらのスクリプトは `raw.githubusercontent.com` からライブ配信されており、独立した署名を**持たない**ため、`curl … | bash` や `irm … | iex` ではスクリプトバイトの改ざんを検出する手段がありません。可能な限り上記の**ネイティブパッケージ**経路を優先してください。どうしてもスクリプトを使う場合は、下の「検証済みスクリプトインストール」を参照してください。 + +```bash +# macOS or Linux x64 curl -fsSL https://raw.githubusercontent.com/tinyhumansai/openhuman/main/scripts/install.sh | bash -# Windows の場合 +# Windows (PowerShell) irm https://raw.githubusercontent.com/tinyhumansai/openhuman/main/scripts/install.ps1 | iex ``` - -> **Linux:** the AppImage can crash on launch under Wayland (and on Arch-based distros with `sharun: Interpreter not found!`) — see [#2463](https://github.com/tinyhumansai/openhuman/issues/2463) for the cause and env-var workarounds. -Arch Linux package maintainers can use the [`openhuman-bin` AUR recipe](../packages/arch/openhuman-bin/); -once published, Arch users can install it with `yay -S openhuman-bin`. - +Debian/Ubuntu では、`install.sh` はまず最新リリースの `.deb` を解決し、`apt-get` でインストールするため、ランタイム依存関係は apt が処理します。AppImage 経路を強制するには `OPENHUMAN_INSTALLER_LINUX_PACKAGE=appimage` を設定してください。 + +## 検証済みスクリプトインストールの状況 + +独立して署名されたスクリプトインストール経路は現在利用できません。Issue [#2620](https://github.com/tinyhumansai/openhuman/issues/2620) はネイティブパッケージ経路が推奨経路に昇格した後にクローズされましたが、現在のリリースアセットには実行前のスクリプト検証用の `install.sh.asc` / `install.ps1.asc` は含まれていません。スクリプトインストール経路は未検証として扱い、可能な限り上記のネイティブパッケージの選択肢を優先してください。 # OpenHuman とは? -OpenHuman は、あなたの日常生活に統合されるよう設計されたオープンソースのエージェント型アシスタントです。各項目は[ドキュメント](https://tinyhumans.gitbook.io/openhuman/)内の詳細な解説にリンクしています。 +OpenHuman は、ほとんどのアシスタントが持っていない 3 つのものです: あなたの世界の永続的なローカルメモリを構築する**脳**、永続的なグラフ上でエージェントの艦隊を動かす**素晴らしいオーケストレーター**、そしてあなたが質問し終える前にあなたのデータとウェブを走査する**ディープリサーチャー**。各項目は[ドキュメント](https://tinyhumans.gitbook.io/openhuman/)内の詳細な解説にリンクしています。 -- **シンプル、UI ファースト、そしてヒューマン** クリーンなデスクトップ体験と短いオンボーディングパスで、インストールから動作するエージェントまで数クリックで到達できます。設定優先のセットアップも、ターミナルも不要です。エージェントには[顔があります](https://tinyhumans.gitbook.io/openhuman/features/mascot): 喋り、周囲に反応し、実際の参加者として[あなたの Google Meet に参加](https://tinyhumans.gitbook.io/openhuman/features/mascot/meeting-agents)し、数週間にわたってあなたのことを覚えており、あなたが入力をやめてもバックグラウンドで考え続けるデスクトップマスコットです。 +### 🧠 脳 -- **[118+ のサードパーティ統合](https://tinyhumans.gitbook.io/openhuman/features/integrations) と [自動取得](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/auto-fetch)**: Gmail、Notion、GitHub、Slack、Stripe、Calendar、Drive、Linear、Jira などのスタックに **ワンクリック OAuth** で接続できます。すべての接続は型付きツールとしてエージェントに公開され、20 分ごとにコアがアクティブな各接続を巡回し、最新データを[メモリーツリー](https://tinyhumans.gitbook.io/openhuman/features/integrations/auto-fetch)に取り込みます。プロンプトも、自分で書くポーリングループも不要なので、エージェントは今朝の時点で明日のコンテキストを既に持っています。 +- **[Memory Tree](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) + [Obsidian Wiki](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki)**: あなたのデータはスコアリングされた Markdown ツリーへ圧縮されてあなたのマシン上の SQLite に保存され、開いて編集できる [Obsidian ボルト](https://x.com/karpathy/status/2039805659525644595)としてミラーリングされます。ベクトルスープのブラックボックスではありません。 +- **[100+ の OAuth 統合、5,000+ の MCP サーバー、90,000+ の Skills](https://tinyhumans.gitbook.io/openhuman/features/integrations)**: Gmail、Notion、GitHub、Slack などのスタックにワンクリックで接続。[自動取得](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/auto-fetch)が 20 分ごとに脳に栄養を与えるので、今朝の時点で明日のコンテキストを持っています。 +- **[サブコンシャス](https://tinyhumans.gitbook.io/openhuman/features/subconscious)**: あなたの世界の差分を取り、ゴールを前進させ、モーニングブリーフィングを書くバックグラウンドループ — あなたが入力をやめた後も思考は続きます。 +- **[Goals & Todos](https://tinyhumans.gitbook.io/openhuman/features/goals-and-todos)**: 長期ゴール、スレッドごとの永続ゴール、そして会話ごとの共有かんばんボード。 +- **[TokenJuice](https://tinyhumans.gitbook.io/openhuman/features/token-compression)**: ツール出力はモデルに届く前に圧縮されます — 同じ情報を最大 80% 少ないトークンで。これがなければ、これほど大きな脳は維持できません。 - マネージド統合は OpenHuman の Composio コネクタレイヤーを使用します。OAuth ハンドシェイクおよび統合ツール呼び出しは、デフォルトでマネージドバックエンドを介してプロキシされます。代わりに Composio を直接実行したい場合は、独自の Composio API キーでダイレクトモードを構成してください。リアルタイムトリガーの Webhook は、その後あなたがホストして配線する必要があります。 +### 🕸️ オーケストレーター -- **[Memory Tree](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) + [Obsidian Wiki](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki)**: あなたのデータとアクティビティから構築されるローカルファーストのナレッジベースです。接続したすべての情報は ≤3k トークンの Markdown チャンクへ正規化され、スコアリングされ、階層的なサマリーツリーに畳み込まれて **あなたのマシン上の SQLite** に保存されます。同じチャンクは Obsidian 互換のボルトに `.md` ファイルとして配置され、開いて閲覧・編集できます。Karpathy 氏の [obsidian-wiki ワークフロー](https://x.com/karpathy/status/2039805659525644595)にインスパイアされています。 +- **[Workflows](https://tinyhumans.gitbook.io/openhuman/features/workflows)** _（新機能）_: エージェントが自動化を提案し、あなたはキャンバス上でレビューして保存します。オープンソースの [tinyflows](https://github.com/tinyhumansai/tinyflows) 上で、永続的・トリガー駆動・承認ゲート付きの実行が行われます。 +- **[仕事をやり遂げるハーネス](https://tinyhumans.gitbook.io/openhuman/developing/architecture/agent-harness)** _（新機能）_: オープンソースの [tinyagents](https://github.com/tinyhumansai/tinyagents) 上でのチェックポイント付きグラフ実行 — 行き詰まったエージェントは軌道修正され、停止したエージェントは根本原因を返し、すべての実行は呼び出しごとの実コスト付きでリプレイできます。 +- **[常時稼働のスプリットブレイン](https://tinyhumans.gitbook.io/openhuman/features/orchestration)** _（新機能）_: 高速な反射エージェントが受信トラフィックをトリアージし、深い推論コアがワーカー艦隊に委任します。サブコンシャスがそれを操縦します。 +- **[エージェントの経済圏](https://tinyhumans.gitbook.io/openhuman/features/tinyplace)**: [tiny.place](https://tiny.place) 上の `@handle`、Signal 暗号化のエージェント間オーケストレーション、x402 USDC バウンティと取引 — 鍵はディスクに一切触れません。 -- **電池同梱(Batteries included)**: ウェブ検索、ウェブフェッチ用[スクレイパー](https://tinyhumans.gitbook.io/openhuman/features/native-tools)、フルコーダーツールセット(ファイルシステム、git、lint、test、grep)、そして[ネイティブ音声](https://tinyhumans.gitbook.io/openhuman/features/native-tools/voice)(STT 入力、ElevenLabs TTS 出力、マスコットのリップシンク、ライブ Google Meet エージェント)がデフォルトで組み込まれています。デフォルトで、[モデルルーティング](https://tinyhumans.gitbook.io/openhuman/features/model-routing)は OpenHuman バックエンドを使用して各ワークロードに適切な LLM(reasoning、fast、または vision)を選択およびプロキシします。一つのサブスクリプションですべてのモデルが含まれます。「ファイル読み込みのためにプラグインをインストール」という煩わしさはありません。デバイス上のワークロード向けに [Ollama によるオプショナルなローカル AI](https://tinyhumans.gitbook.io/openhuman/features/model-routing/local-ai) も利用できます。 +### 🔬 ディープリサーチャー & 実行者 -- **[スマートトークン圧縮 (TokenJuice)](https://tinyhumans.gitbook.io/openhuman/features/token-compression)**: すべてのツール呼び出し、スクレイプ結果、メール本文、検索ペイロードは、LLM モデルに渡される前にトークン圧縮レイヤーを通過します。HTML は Markdown に変換され、長い URL は短縮され、冗長なツール出力は設定可能なルールレイヤーで重複排除と要約が行われるなど…。CJK、絵文字などのマルチバイト文字は書記素(grapheme)単位で完全に保持され、除去されることはありません。同じ情報をわずかなトークン数で得られます。コストとレイテンシを最大 80% 削減します。 +- **[SuperContext](https://tinyhumans.gitbook.io/openhuman/features/super-context)**: モデルがあなたの最初のメッセージを読む前に、リサーチスカウトがメモリとファイルを走査します。コールドスタートはありません。 +- **電池同梱（Batteries included）**: ウェブ検索、スクレイパー、コーダーツールセット、本物の[ブラウザ](https://tinyhumans.gitbook.io/openhuman/features/native-tools/browser-and-computer)、インプロセス Whisper による[ネイティブ音声](../gitbooks/features/native-tools/voice.md) — さらにワークロードごとに適切な LLM を選ぶ[モデルルーティング](https://tinyhumans.gitbook.io/openhuman/features/model-routing)。1 つのサブスクリプションで、[ローカル AI はオプション](https://tinyhumans.gitbook.io/openhuman/features/model-routing/local-ai)です。 +- **[会議エージェント](https://tinyhumans.gitbook.io/openhuman/features/mascot/meeting-agents)** _（新機能）_: 顔と声を持って **Meet、Zoom、Teams、Webex** に参加 — カレンダーから自動参加し、ライブ文字起こしをストリーミングし、名前で呼ばれると答え、要約とアクションアイテムを保存します。 +- **[画像・動画生成](https://tinyhumans.gitbook.io/openhuman/features/native-tools)** _（新機能）_: Seedream/SeedEdit の画像と Seedance/Veo の動画を、同じサブスクリプションでワークスペースに直接生成します。 +- **[17 のメッセージングチャネル](https://tinyhumans.gitbook.io/openhuman/features/channels)**: Telegram、Discord、Slack、WhatsApp、Signal、iMessage… さらに**ネイティブメール**（IMAP IDLE + SMTP）。エージェントはあなたが既にいる場所であなたに届きます。 -- **[メッセージングチャネル](https://tinyhumans.gitbook.io/openhuman/features/integrations#messaging-channels)** と **[プライバシー & セキュリティ](https://tinyhumans.gitbook.io/openhuman/features/privacy-and-security)**: あなたが既に使っているチャネル全体での送受信が可能で、ワークフローデータはデバイス上に留まり、ローカルで暗号化され、あなた自身のものとして扱われます。 +### 🧍 ヒューマンで、プライベートで、あなたのもの + +- **シンプル、UI ファースト、そしてヒューマン**: インストールから動作するエージェントまで数クリック — 設定ファイルもターミナルも不要です。そして[顔があります](https://tinyhumans.gitbook.io/openhuman/features/mascot): 喋り、反応し、あなたを覚えているマスコットです。 +- **[プライバシー & セキュリティ](https://tinyhumans.gitbook.io/openhuman/features/privacy-and-security)**: デバイス上で暗号化されたデータ、承認ゲート、OS キーリングによるシークレット管理、オプトインのサンドボックス — そして _（新機能）_ **[Privacy Mode](https://tinyhumans.gitbook.io/openhuman/features/privacy-mode)**: スイッチひとつで推論が一切マシンの外に出なくなり、Rust コアで強制されます。 +- **[テーマ & Theme Studio](https://tinyhumans.gitbook.io/openhuman/features/theming)**: 5 つのテーマファミリーとフル機能のビジュアルエディター、JSON としてエクスポート可能。 ## ソースからのコントリビュート -新しいコントリビューターの方は、まず [`CONTRIBUTING.md`](../CONTRIBUTING.md) で fork/PR ワークフローとローカル検証コマンドを確認してください。最短経路は以下のとおりです: +新しいコントリビューターの方は、まず [`CONTRIBUTING.md`](../CONTRIBUTING.md) で fork/PR ワークフローとローカル検証コマンドを確認するか、[`CONTRIBUTING-BEGINNERS.md`](../CONTRIBUTING-BEGINNERS.md#optional-let-an-ai-coding-agent-guide-you) のコピー&ペーストできる AI エージェント向けプロンプトを使ってください。最短経路は以下のとおりです: -1. Git、Node.js 24+、pnpm 10.10.0、Rust 1.93.0(`rustfmt` + `clippy`)、CMake、Ninja、ripgrep、プラットフォーム向けデスクトップビルドの前提条件をインストールします。 +1. Git、Node.js 24+、pnpm 10.10.0、Rust 1.93.0（`rustfmt` + `clippy`）、CMake、Ninja、ripgrep、プラットフォーム向けデスクトップビルドの前提条件をインストールします。 2. リポジトリを fork してクローンし、`pnpm install` の前に `git submodule update --init --recursive` を実行して、ベンダー化された Tauri/CEF のソースを取得します。 3. ウェブのみの UI 作業には `pnpm dev` を、デスクトップシェルには `pnpm --filter openhuman-app dev:app` を使用し、PR を出す前に `pnpm typecheck`、`pnpm format:check`、`cargo check -p openhuman --lib` などの集中チェックを実行してください。 @@ -112,28 +165,45 @@ OpenHuman はその待ち時間をスキップします。アカウントを接 既に他のコーディングエージェント間で [agentmemory](https://github.com/rohitg00/agentmemory) をセルフホストしていますか? OpenHuman にはそれにプロキシするオプションの `Memory` バックエンドが同梱されています。`config.toml` で `memory.backend = "agentmemory"` を設定すれば、同じ永続ストアが Claude Code、Cursor、Codex、OpenCode と並んで OpenHuman を駆動します。セットアップ方法は [agentmemory バックエンド](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/agentmemory-backend)のページを参照してください。 +## チャットボットではなく、オーケストレーター + +最も深い違いは、個々の機能ではなく実行モデルにあります。Claude Code、OpenClaw、Hermes は**1 つのループで 1 つのエージェント**を動かします。OpenHuman は**[オーケストレーター](https://tinyhumans.gitbook.io/openhuman/features/orchestration)**です: + +- **ループではなくグラフ** — ターンは [tinyagents](https://github.com/tinyhumansai/tinyagents) 上のチェックポイント付きステートマシングラフにコンパイルされます: 人間のために一時停止し、再起動を生き延び、実行の途中から再開します。 +- **サブエージェントの艦隊** — スペシャリストは 3 階層の深さまで起動され、アイドル状態のワーカーは再利用され、サーキットブレーカーが行き詰まったエージェントを根本原因レポートに変えます。 +- **目に見えるワークフロー** — エージェントが提案する [tinyflows](https://github.com/tinyhumansai/tinyflows) グラフをキャンバス上でレビュー: 永続的、トリガー駆動、承認ゲート付き。 +- **常時稼働のスプリットブレイン** — 高速な反射エージェントがトリアージし、深いコアが推論します。サブコンシャスループがそれを操縦します。 +- **エージェント間、暗号化済み** — インスタンス同士は、同意ベースのペアリングと x402 決済を備えた **Signal プロトコルの E2E** セッション上で互いをオーケストレーションします。サーバーは平文を一切見ません。 +- **次は RLM** — 同じグラフエンジンと信頼モデルの上で、モデルがサンドボックス化された REPL の中で自らのオーケストレーションコードを書きます。 + ## OpenHuman と他のエージェントハーネスの比較 ハイレベルな比較です(製品は進化するため、各ベンダーで確認してください)。OpenHuman は **ベンダーの乱立を最小限に抑え**、**ワークフロー知識をデバイス上に保ち**、チャットだけでなくあなたのデータに対する **永続的なメモリ** をエージェントに与えるよう構築されています。 -| | Claude Cowork | OpenClaw | Hermes Agent | OpenHuman | -| -------------------- | ------------------- | ------------------- | ------------------- | -------------------------------------------------- | -| **オープンソース** | 🚫 プロプライエタリ | ✅ MIT | ✅ MIT | ✅ GNU | -| **開始が簡単** | ✅ デスクトップ + CLI | ⚠️ ターミナル中心 | ⚠️ ターミナル中心 | ✅ クリーンな UI、数分 | -| **コスト** | ⚠️ サブスク + アドオン | ⚠️ モデル持ち込み | ⚠️ モデル持ち込み | ✅ 1 つのサブスク + TokenJuice | -| **メモリ** | ✅ チャット範囲のみ | ⚠️ プラグイン依存 | ✅ 自己学習 | 🚀 Memory Tree + Obsidian ボルト、オプションの [agentmemory](https://github.com/rohitg00/agentmemory) バックエンド | -| **統合** | ⚠️ 少数のコネクター | ⚠️ 持ち込み | ⚠️ 持ち込み | 🚀 OAuth 経由で 118+ | -| **自動取得** | 🚫 なし | 🚫 なし | 🚫 なし | ✅ 20 分同期でメモリに取り込み | -| **API の乱立** | 🚫 追加キー | 🚫 BYOK | 🚫 マルチベンダー | ✅ 1 アカウント | -| **モデルルーティング** | 🚫 単一モデル | ⚠️ 手動 | ⚠️ 手動 | ✅ ビルトイン | -| **ネイティブツール** | ✅ コードのみ | ✅ コードのみ | ✅ コードのみ | ✅ コード + 検索 + スクレイパー + 音声 | +| | Claude Cowork | OpenClaw | Hermes Agent | OpenHuman | +| -------------------------- | ---------------------- | ------------------- | ------------------- | -------------------------------------------------------------------------------------------------------- | +| **オープンソース** | 🚫 プロプライエタリ | ✅ MIT | ✅ MIT | ✅ GNU | +| **開始が簡単** | ✅ デスクトップ + CLI | ⚠️ ターミナル中心 | ⚠️ ターミナル中心 | ✅ クリーンな UI、数分 | +| **コスト** | ⚠️ サブスク + アドオン | ⚠️ モデル持ち込み | ⚠️ モデル持ち込み | ✅ 1 つのサブスク + TokenJuice | +| **メモリ** | ✅ チャット範囲のみ | ⚠️ プラグイン依存 | ✅ 自己学習 | 🚀 Memory Tree + Obsidian ボルト、オプションの [agentmemory](https://github.com/rohitg00/agentmemory) バックエンド | +| **統合** | ⚠️ 少数のコネクター | ⚠️ 持ち込み | ⚠️ 持ち込み | 🚀 100+ OAuth · 5k+ MCP · 90k+ Skills | +| **自動取得** | 🚫 なし | 🚫 なし | 🚫 なし | ✅ 20 分同期でメモリに取り込み | +| **オーケストレーション** | ⚠️ サブタスク | ⚠️ 単一ループ | ⚠️ 単一ループ | 🚀 エージェントグラフ + チェックポイント + E2E 暗号化 A2A | +| **ワークフロー** | 🚫 なし | ⚠️ スクリプト | ⚠️ スクリプト | 🚀 ビジュアル、永続、エージェント提案、承認ゲート付き | +| **会議** | 🚫 なし | 🚫 なし | 🚫 なし | 🚀 Meet/Zoom/Teams/Webex に参加、発話、ライブ文字起こし | +| **メッセージングチャネル** | 🚫 なし | ⚠️ 少数 | ⚠️ 少数 | ✅ ネイティブメール（IMAP/SMTP）を含む 17 チャネル | +| **ローカル専用モード** | 🚫 クラウドのみ | ⚠️ ローカル持ち込み | ⚠️ ローカル持ち込み | ✅ スイッチひとつで強制される Privacy Mode | +| **可観測性** | 🚫 不透明 | ⚠️ ログ | ⚠️ ログ | ✅ リプレイ可能な実行ジャーナル + 呼び出しごとのコスト計上 | +| **API の乱立** | 🚫 追加キー | 🚫 BYOK | 🚫 マルチベンダー | ✅ 1 アカウント | +| **モデルルーティング** | 🚫 単一モデル | ⚠️ 手動 | ⚠️ 手動 | ✅ ビルトイン | +| **ネイティブツール** | ✅ コードのみ | ✅ コードのみ | ✅ コードのみ | ✅ コード + 検索 + スクレイパー + ブラウザ + 音声 + メディア生成 | # GitHub でスターをお願いします _AGI と人工意識への道を進んでいますか? リポジトリにスターをつけて、他の人にも道筋を見つけてもらいましょう。_

- + diff --git a/docs/README.ko.md b/docs/README.ko.md index 68a040e6f..d5ca32884 100644 --- a/docs/README.ko.md +++ b/docs/README.ko.md @@ -5,24 +5,32 @@

- - tinyhumansai%2Fopenhuman | Trendshift - -   - - OpenHuman - An open source AI harness built with the human in mind | Product Hunt - - -

- -

- OpenHuman은 당신의 개인용 AI 슈퍼 지능입니다: 로컬 메모리, 필요한 경우 관리형 서비스, 단순하고 강력합니다. + + tinyhumansai%2Fopenhuman | Trendshift + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt +

+

+ + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + +

+

+ OpenHuman은 당신의 개인용 AI 슈퍼 지능입니다: 모든 것을 기억하는 두뇌, 탁월한 오케스트레이터, 깊이 있는 리서처. 로컬 우선, 단순하고 강력합니다. +

Discord • - Discussions • + Reddit • X/Twitter • 문서 • @senamakel(제작자) 팔로우 @@ -32,8 +40,6 @@ 🇺🇸 English | 🇨🇳 简体中文 | 🇯🇵 日本語 | 🇰🇷 한국어 | 🇩🇪 Deutsch | 🇵🇰 اردو

- -

얼리 베타 최신 릴리스 @@ -49,45 +55,94 @@ > **얼리 베타**: 활발히 개발 중입니다. 다소 미흡한 부분이 있을 수 있습니다. -> **로컬 + 관리형 서비스, upfront:** OpenHuman은 Memory Tree, Obsidian 스타일 Markdown 볼트, 워크스페이스 설정 및 로컬 런타임 상태를 사용자의 머신에 저장합니다. 기본 관리형 경험은 여전히 계정 로그인, 모델 라우팅, 웹 검색 프록시 및 Composio 커넥터 레이어를 통한 관리형 통합/OAuth 플로우에 OpenHuman 호스팅 서비스를 사용합니다. 자체 모델, 검색 또는 Composio 인증 정보를 가져오려면 사용자 지정/로컬 설정을 선택하세요. 일부 실시간 트리거 및 호스팅 기능은 여전히 관리형 백엔드를 필요로 합니다. +> 🎉 출시 후 일주일 만에 OpenHuman은 9일 연속 GitHub 트렌딩 저장소 1위에 올랐습니다. -설치하거나 시작하려면 [tinyhumans.ai/openhuman](https://tinyhumans.ai/openhuman?utm_source=github&utm_medium=readme) 웹사이트에서 다운로드하거나 터미널에서 다음을 실행하세요. +# 설치 + +[tinyhumans.ai/openhuman](https://tinyhumans.ai/openhuman?utm_source=github&utm_medium=readme) 또는 [GitHub Releases](https://github.com/tinyhumansai/openhuman/releases/latest) 페이지에서 설치 프로그램을 다운로드하세요. 터미널 설치의 경우, 가능한 한 OS 패키지 관리자나 네이티브 설치 프로그램을 사용하는 아래의 네이티브 패키지 경로를 권장합니다. + +## 권장 설치 방법 (네이티브 패키지) + +이 경로들은 네이티브 설치 프로그램을 사용합니다. Homebrew와 MSI는 일반적인 서명/무결성 검사를 제공하며, Debian/Ubuntu는 `apt-get`으로 릴리스 `.deb`를 설치하고 시스템 의존성을 해결합니다. + +**macOS (Homebrew tap):** ```bash -# https://tinyhumans.ai/openhuman 에서 DMG, EXE를 다운로드하거나 터미널에서 실행하세요. +brew tap tinyhumansai/core +brew install openhuman +``` -# macOS 또는 Linux x64용 +**Linux (Debian/Ubuntu — 릴리스 `.deb`):** + +```bash +# Download OpenHuman__amd64.deb or OpenHuman__arm64.deb +# from https://github.com/tinyhumansai/openhuman/releases/latest, then: +# Replace amd64 with arm64 on arm64 hosts. +sudo apt-get install -y --no-install-recommends ./OpenHuman_*_amd64.deb +``` + +**Linux (Arch — AUR):** [`openhuman-bin` AUR 레시피](../packages/arch/openhuman-bin/)가 저장소에 포함되어 있습니다. 게시된 후에는 Arch 사용자가 `yay -S openhuman-bin`으로 설치할 수 있습니다. + +**Windows:** [최신 릴리스](https://github.com/tinyhumansai/openhuman/releases/latest)에서 서명된 `.msi`를 다운로드하여 실행하세요. + +**수동 `.dmg` / `.deb` / `.AppImage` / `.msi` 설치:** [최신 릴리스 페이지](https://github.com/tinyhumansai/openhuman/releases/latest)에서 플랫폼에 맞는 설치 프로그램을 직접 받으세요. + +> **Linux:** AppImage는 Wayland 환경에서 실행 시 크래시가 발생하거나, `libgbm.so.1` 같은 호스트 시스템 라이브러리가 없거나, Arch 계열 배포판에서 `sharun: Interpreter not found!` 오류로 실패할 수 있습니다. 원인과 환경 변수 우회 방법은 [#2463](https://github.com/tinyhumansai/openhuman/issues/2463)을 참조하세요. 위의 `.deb` 패키지는 apt가 런타임 의존성을 해결하도록 하여 Debian/Ubuntu에서 이러한 문제를 피합니다. + +## 대안: 스크립트 설치 (무결성 검사 없음) + +> **경고 — 검증되지 않은 설치입니다.** 이 스크립트들은 `raw.githubusercontent.com`에서 실시간으로 제공되며 별도의 서명을 포함하지 **않으므로**, `curl … | bash`와 `irm … | iex`는 스크립트 바이트가 변조되었는지 감지할 방법이 없습니다. 가능하면 위의 **네이티브 패키지** 경로를 우선 사용하세요. 스크립트를 반드시 사용해야 한다면 아래의 "검증된 스크립트 설치 상태"를 참조하세요. + +```bash +# macOS or Linux x64 curl -fsSL https://raw.githubusercontent.com/tinyhumansai/openhuman/main/scripts/install.sh | bash -# Windows용 +# Windows (PowerShell) irm https://raw.githubusercontent.com/tinyhumansai/openhuman/main/scripts/install.ps1 | iex ``` - -> **Linux:** the AppImage can crash on launch under Wayland (and on Arch-based distros with `sharun: Interpreter not found!`) — see [#2463](https://github.com/tinyhumansai/openhuman/issues/2463) for the cause and env-var workarounds. -Arch Linux package maintainers can use the [`openhuman-bin` AUR recipe](../packages/arch/openhuman-bin/); -once published, Arch users can install it with `yay -S openhuman-bin`. - +Debian/Ubuntu에서 `install.sh`는 먼저 최신 릴리스 `.deb`를 찾아 `apt-get`으로 설치하므로 런타임 의존성이 apt에 의해 처리됩니다. AppImage 경로를 강제하려면 `OPENHUMAN_INSTALLER_LINUX_PACKAGE=appimage`를 설정하세요. + +## 검증된 스크립트 설치 상태 + +별도로 서명된 스크립트 설치 경로는 현재 제공되지 않습니다. 이슈 [#2620](https://github.com/tinyhumansai/openhuman/issues/2620)은 네이티브 패키지 경로가 우선 방식으로 승격된 후 닫혔지만, 현재 릴리스 자산에는 실행 전 스크립트 검증을 위한 `install.sh.asc` / `install.ps1.asc`가 포함되어 있지 않습니다. 스크립트 설치 경로는 검증되지 않은 것으로 간주하고, 가능하면 위의 네이티브 패키지 옵션을 우선 사용하세요. # OpenHuman이란 무엇인가요? -OpenHuman은 일상 생활에 통합되도록 설계된 오픈 소스 에이전트 어시스턴트입니다. 각 글머리 기호는 [문서](https://tinyhumans.gitbook.io/openhuman/)의 더 깊은 설명으로 연결됩니다. +OpenHuman은 대부분의 어시스턴트가 갖지 못한 세 가지입니다: 당신의 세계에 대한 지속적인 로컬 메모리를 구축하는 **두뇌**, 내구성 있는 그래프 위에서 에이전트 함대를 운영하는 **탁월한 오케스트레이터**, 그리고 질문을 끝내기도 전에 당신의 데이터와 웹을 훑는 **깊이 있는 리서처**. 각 글머리 기호는 [문서](https://tinyhumans.gitbook.io/openhuman/)의 더 깊은 설명으로 연결됩니다. -- **단순함, UI 우선 및 인간 중심**: 깔끔한 데스크톱 경험과 짧은 온보딩 경로를 통해 설치 후 몇 번의 클릭만으로 작동하는 에이전트를 만날 수 있습니다. 설정 우선 방식이나 터미널이 필요하지 않습니다. 에이전트는 [얼굴](https://tinyhumans.gitbook.io/openhuman/features/mascot)을 가지고 있습니다. 말을 하고, 주변 환경에 반응하며, 실제 참가자로 [Google Meet에 참여](https://tinyhumans.gitbook.io/openhuman/features/mascot/meeting-agents)하고, 몇 주 동안 당신을 기억하며, 타이핑을 멈춘 후에도 백그라운드에서 계속 생각하는 데스크톱 마스코트입니다. +### 🧠 두뇌 -- **[118개 이상의 서드파티 통합](https://tinyhumans.gitbook.io/openhuman/features/integrations) 및 [자동 가져오기(auto-fetch)](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/auto-fetch)**: Gmail, Notion, GitHub, Slack, Stripe, Calendar, Drive, Linear, Jira 등 당신의 스택을 **원클릭 OAuth**로 연결하세요. 모든 연결은 유형이 지정된 도구로 에이전트에게 노출되며, 코어는 20분마다 각 활성 연결을 탐색하여 신선한 데이터를 [메모리 트리](https://tinyhumans.gitbook.io/openhuman/features/integrations/auto-fetch)로 가져옵니다. 프롬프트나 직접 작성해야 하는 폴링 루프가 필요 없으므로, 에이전트는 오늘 아침에 이미 내일의 컨텍스트를 가지고 있습니다. +- **[메모리 트리(Memory Tree)](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) + [Obsidian 위키](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki)**: 당신의 데이터는 점수가 매겨진 Markdown 트리로 압축되어 당신의 머신에 있는 SQLite에 저장되고, 열어서 직접 편집할 수 있는 [Obsidian 볼트](https://x.com/karpathy/status/2039805659525644595)로 미러링됩니다. 벡터 수프 같은 블랙박스가 아닙니다. +- **[100개 이상의 OAuth 통합, 5,000개 이상의 MCP 서버, 90,000개 이상의 Skills](https://tinyhumans.gitbook.io/openhuman/features/integrations)**: Gmail, Notion, GitHub, Slack 등 당신의 스택을 원클릭으로 연결하세요. [자동 가져오기(auto-fetch)](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/auto-fetch)가 20분마다 두뇌에 데이터를 공급합니다 — 오늘 아침에 이미 내일의 컨텍스트를 가지고 있습니다. +- **[잠재의식(subconscious)](https://tinyhumans.gitbook.io/openhuman/features/subconscious)**: 당신의 세계의 변화를 비교 분석하고, 목표를 진전시키고, 아침 브리핑을 작성하는 백그라운드 루프 — 타이핑을 멈춘 후에도 생각은 계속됩니다. +- **[목표 및 할 일(Goals & Todos)](https://tinyhumans.gitbook.io/openhuman/features/goals-and-todos)**: 장기 목표, 스레드별 지속 목표, 그리고 대화별 공유 칸반 보드를 제공합니다. +- **[TokenJuice](https://tinyhumans.gitbook.io/openhuman/features/token-compression)**: 도구 출력은 모델에 닿기 전에 압축됩니다 — 동일한 정보를 최대 80% 적은 토큰으로. 이것 없이는 이만큼 큰 두뇌를 감당할 수 없을 것입니다. -- **[메모리 트리(Memory Tree)](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) + [Obsidian 위키](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki)**: 당신의 데이터와 활동을 바탕으로 구축된 로컬 우선 지식 베이스입니다. 연결된 모든 것은 3k 토큰 이하의 Markdown 청크로 규격화되고 점수가 매겨지며, **당신의 머신에 있는 SQLite**에 저장되는 계층적 요약 트리로 접힙니다. 동일한 청크는 당신이 열고, 탐색하고, 편집할 수 있는 Obsidian 호환 볼트에 `.md` 파일로 저장됩니다. 이는 Karpathy의 [obsidian-wiki 워크플로우](https://x.com/karpathy/status/2039805659525644595)에서 영감을 받았습니다. +### 🕸️ 오케스트레이터 -- **모든 것이 포함됨(Batteries included)**: 웹 검색, 웹 가져오기 [스크레이퍼](https://tinyhumans.gitbook.io/openhuman/features/native-tools), 전체 코더 툴셋(파일 시스템, git, lint, test, grep), 그리고 [네이티브 음성](https://tinyhumans.gitbook.io/openhuman/features/native-tools/voice)(STT 입력, ElevenLabs TTS 출력, 마스코트 립싱크, 라이브 Google Meet 에이전트)이 기본적으로 연결되어 있습니다. 기본적으로 [모델 라우팅](https://tinyhumans.gitbook.io/openhuman/features/model-routing)은 OpenHuman 백엔드를 사용하여 각 워크로드에 적합한 LLM(추론, 고속 또는 비전)을 선택하고 프록시합니다. 하나의 구독에 모든 모델이 포함됩니다. "파일을 읽기 위해 플러그인 설치"와 같은 번거로움이 없습니다. 온디바이스 워크로드를 위해 [Ollama를 통한 선택적 로컬 AI](https://tinyhumans.gitbook.io/openhuman/features/model-routing/local-ai)를 지원합니다. +- **[Workflows](https://tinyhumans.gitbook.io/openhuman/features/workflows)** _(신규)_: 에이전트가 자동화를 제안하면 캔버스에서 검토하고 저장하면 됩니다. 내구성 있고, 트리거 기반이며, 승인 게이트를 거치는 실행이 오픈 소스 [tinyflows](https://github.com/tinyhumansai/tinyflows) 위에서 동작합니다. +- **[일을 끝까지 마무리하는 하네스](https://tinyhumans.gitbook.io/openhuman/developing/architecture/agent-harness)** _(신규)_: 오픈 소스 [tinyagents](https://github.com/tinyhumansai/tinyagents) 기반의 체크포인트 그래프 실행 — 막힌 에이전트는 방향을 조정받고, 중단된 에이전트는 근본 원인을 돌려주며, 모든 실행은 호출별 실제 비용과 함께 재생됩니다. +- **[상시 가동되는 분할 두뇌(split brain)](https://tinyhumans.gitbook.io/openhuman/features/orchestration)** _(신규)_: 빠른 반사 에이전트가 들어오는 트래픽을 분류하는 동안 깊은 추론 코어가 워커 함대에 작업을 위임하며, 잠재의식이 이를 조종합니다. +- **[에이전트 경제](https://tinyhumans.gitbook.io/openhuman/features/tinyplace)**: [tiny.place](https://tiny.place)의 `@handle`, Signal로 암호화된 에이전트 간 오케스트레이션, x402 USDC 바운티와 거래 — 키는 디스크에 절대 닿지 않습니다. -- **[스마트 토큰 압축(TokenJuice)](https://tinyhumans.gitbook.io/openhuman/features/token-compression)**: 모든 도구 호출, 스크레이핑 결과, 이메일 본문 및 검색 페이로드는 LLM 모델에 전달되기 전에 토큰 압축 레이어를 거칩니다. HTML은 Markdown으로 변환되고, 긴 URL은 단축되며, 장황한 도구 출력은 구성 가능한 규칙 오버레이 등을 통해 중복 제거 및 요약됩니다. CJK, 이모지 및 기타 멀티바이트 텍스트는 자소(grapheme) 단위로 보존되며 절대 삭제되지 않습니다. 동일한 정보를 훨씬 적은 토큰으로 얻을 수 있어 비용과 지연 시간을 최대 80%까지 줄일 수 있습니다. +### 🔬 깊이 있는 리서처이자 실행가 -- **[메시징 채널](https://tinyhumans.gitbook.io/openhuman/features/integrations#messaging-channels)** 및 **[개인 정보 보호 및 보안](https://tinyhumans.gitbook.io/openhuman/features/privacy-and-security)**: 이미 사용 중인 채널을 통해 메시지를 주고받을 수 있으며, 워크플로우 데이터는 기기에 남아 로컬에서 암호화되어 당신의 것으로 관리됩니다. +- **[SuperContext](https://tinyhumans.gitbook.io/openhuman/features/super-context)**: 모델이 첫 메시지를 읽기 전에 리서치 스카우트가 당신의 메모리와 파일을 훑습니다. 콜드 스타트가 없습니다. +- **모든 것이 포함됨(Batteries included)**: 웹 검색, 스크레이퍼, 코더 툴셋, 실제 [브라우저](https://tinyhumans.gitbook.io/openhuman/features/native-tools/browser-and-computer), 인프로세스 Whisper를 갖춘 [네이티브 음성](../gitbooks/features/native-tools/voice.md) — 그리고 워크로드별로 적합한 LLM을 선택하는 [모델 라우팅](https://tinyhumans.gitbook.io/openhuman/features/model-routing)까지. 하나의 구독으로, [로컬 AI는 선택 사항](https://tinyhumans.gitbook.io/openhuman/features/model-routing/local-ai)입니다. +- **[미팅 에이전트](https://tinyhumans.gitbook.io/openhuman/features/mascot/meeting-agents)** _(신규)_: 얼굴과 목소리를 가지고 **Meet, Zoom, Teams, Webex**에 참여합니다 — 캘린더에서 자동으로 참여하고, 실시간 자막을 스트리밍하며, 이름이 불리면 대답하고, 요약과 액션 아이템을 정리합니다. +- **[이미지 및 비디오 생성](https://tinyhumans.gitbook.io/openhuman/features/native-tools)** _(신규)_: Seedream/SeedEdit 이미지와 Seedance/Veo 비디오가 동일한 구독으로 워크스페이스에 바로 생성됩니다. +- **[17개의 메시징 채널](https://tinyhumans.gitbook.io/openhuman/features/channels)**: Telegram, Discord, Slack, WhatsApp, Signal, iMessage… 그리고 **네이티브 이메일**(IMAP IDLE + SMTP)까지. 에이전트는 당신이 이미 있는 곳에서 당신에게 닿습니다. + +### 🧍 인간적이고, 프라이빗하며, 당신의 것 + +- **단순함, UI 우선 및 인간 중심**: 설치 후 몇 번의 클릭만으로 작동하는 에이전트를 만날 수 있습니다 — 설정 파일도, 터미널도 필요 없습니다. 그리고 [얼굴](https://tinyhumans.gitbook.io/openhuman/features/mascot)이 있습니다: 말하고, 반응하고, 당신을 기억하는 마스코트입니다. +- **[개인 정보 보호 및 보안](https://tinyhumans.gitbook.io/openhuman/features/privacy-and-security)**: 기기 내 암호화된 데이터, 승인 게이트, OS 키링 시크릿, 선택적 샌드박싱 — 그리고 _(신규)_ **[Privacy Mode](https://tinyhumans.gitbook.io/openhuman/features/privacy-mode)**: 스위치 하나로 어떤 추론도 당신의 머신을 떠나지 않으며, Rust 코어에서 강제됩니다. +- **[테마 및 Theme Studio](https://tinyhumans.gitbook.io/openhuman/features/theming)**: 5가지 테마 패밀리와 완전한 시각적 에디터, JSON으로 내보낼 수 있습니다. ## 소스에서 기여하기 -새로운 기여자인가요? 포크/PR 워크플로우 및 로컬 검증 명령에 대해서는 [`CONTRIBUTING.md`](../CONTRIBUTING.md)에서 시작하세요. 빠른 경로는 다음과 같습니다. +새로운 기여자인가요? 포크/PR 워크플로우 및 로컬 검증 명령에 대해서는 [`CONTRIBUTING.md`](../CONTRIBUTING.md)에서 시작하거나, [`CONTRIBUTING-BEGINNERS.md`](../CONTRIBUTING-BEGINNERS.md#optional-let-an-ai-coding-agent-guide-you)의 복사-붙여넣기 AI 에이전트 프롬프트를 사용하세요. 빠른 경로는 다음과 같습니다. 1. Git, Node.js 24+, pnpm 10.10.0, Rust 1.93.0(`rustfmt` + `clippy`), CMake, Ninja, ripgrep 및 플랫폼 데스크톱 빌드 필수 구성 요소를 설치합니다. 2. 저장소를 포크하고 클론한 다음, `pnpm install` 전에 `git submodule update --init --recursive`를 실행하여 벤더링된 Tauri/CEF 소스가 존재하는지 확인합니다. @@ -111,28 +166,45 @@ OpenHuman은 기다림을 생략합니다. 계정을 연결하고, [자동 가 이미 다른 코딩 에이전트에서 [agentmemory](https://github.com/rohitg00/agentmemory)를 자체 호스팅하고 있나요? OpenHuman은 이를 프록시하는 선택적 `Memory` 백엔드를 제공합니다. `config.toml`에서 `memory.backend = "agentmemory"`를 설정하면 동일한 내구성 있는 저장소가 Claude Code, Cursor, Codex, OpenCode와 함께 OpenHuman을 구동합니다. 설정 방법은 [agentmemory 백엔드](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/agentmemory-backend) 페이지를 참조하세요. +## 챗봇이 아닌 오케스트레이터 + +가장 깊은 차이는 어느 하나의 기능이 아니라 실행 모델입니다. Claude Code, OpenClaw, Hermes는 **하나의 루프에서 하나의 에이전트**를 실행합니다. OpenHuman은 **[오케스트레이터](https://tinyhumans.gitbook.io/openhuman/features/orchestration)**입니다: + +- **루프가 아닌 그래프** — 턴은 [tinyagents](https://github.com/tinyhumansai/tinyagents) 기반의 체크포인트가 있는 상태 머신 그래프로 컴파일됩니다: 사람을 위해 일시 정지하고, 재시작에도 살아남고, 실행 중간에 재개합니다. +- **서브 에이전트 함대** — 전문 에이전트가 3단계 깊이까지 생성되고, 유휴 워커는 재사용되며, 서킷 브레이커가 막힌 에이전트를 근본 원인 보고서로 바꿔 줍니다. +- **눈으로 볼 수 있는 워크플로우** — 에이전트가 제안하고 캔버스에서 검토하는 [tinyflows](https://github.com/tinyhumansai/tinyflows) 그래프: 내구성 있고, 트리거 기반이며, 승인 게이트를 거칩니다. +- **상시 가동되는 분할 두뇌(split brain)** — 빠른 반사 에이전트가 분류하는 동안 깊은 코어가 추론하며, 잠재의식 루프가 이를 조종합니다. +- **암호화된 에이전트 간 통신** — 인스턴스들은 동의 기반 페어링과 x402 결제를 갖춘 **Signal 프로토콜 E2E** 세션을 통해 서로를 오케스트레이션합니다. 어떤 서버도 평문을 보지 못합니다. +- **다음 단계: RLM** — 모델이 샌드박스 REPL에서 자신의 오케스트레이션 코드를 직접 작성합니다. 동일한 그래프 엔진과 신뢰 모델 위에서요. + ## OpenHuman vs 다른 에이전트 하네스 상위 수준 비교(제품은 진화하므로 각 벤더에 확인하세요). OpenHuman은 **벤더 분산화(sprawl)를 최소화**하고, **워크플로우 지식을 기기에 유지**하며, 채팅뿐만 아니라 당신의 데이터에 대한 **지속적인 기억**을 에이전트에게 제공하도록 구축되었습니다. -| | Claude Cowork | OpenClaw | Hermes Agent | OpenHuman | -| ------------------- | ----------------- | ----------------- | ----------------- | ---------------------------------- | -| **오픈 소스** | 🚫 독점 소스 | ✅ MIT | ✅ MIT | ✅ GNU | -| **시작하기 쉬움** | ✅ 데스크톱 + CLI | ⚠️ 터미널 우선 | ⚠️ 터미널 우선 | ✅ 깔끔한 UI, 단 몇 분 | -| **비용** | ⚠️ 구독 + 애드온 | ⚠️ 모델 직접 제공 | ⚠️ 모델 직접 제공 | ✅ 단일 구독 + TokenJuice | -| **메모리** | ✅ 채팅 범위 한정 | ⚠️ 플러그인 의존 | ✅ 자기 학습 | 🚀 메모리 트리 + Obsidian 볼트, 선택적 [agentmemory](https://github.com/rohitg00/agentmemory) 백엔드 | -| **통합** | ⚠️ 적은 커넥터 | ⚠️ 직접 구축 | ⚠️ 직접 구축 | 🚀 OAuth를 통한 118개 이상 | -| **자동 가져오기** | 🚫 없음 | 🚫 없음 | 🚫 없음 | ✅ 20분마다 메모리로 동기화 | -| **API 분산화** | 🚫 추가 키 필요 | 🚫 BYOK | 🚫 멀티 벤더 | ✅ 단일 계정 | -| **모델 라우팅** | 🚫 단일 모델 | ⚠️ 수동 | ⚠️ 수동 | ✅ 내장됨 | -| **네이티브 도구** | ✅ 코드 전용 | ✅ 코드 전용 | ✅ 코드 전용 | ✅ 코드 + 검색 + 스크레이퍼 + 음성 | +| | Claude Cowork | OpenClaw | Hermes Agent | OpenHuman | +| ------------------- | ----------------- | ----------------- | ----------------- | -------------------------------------------------------------------------------------------------------- | +| **오픈 소스** | 🚫 독점 소스 | ✅ MIT | ✅ MIT | ✅ GNU | +| **시작하기 쉬움** | ✅ 데스크톱 + CLI | ⚠️ 터미널 우선 | ⚠️ 터미널 우선 | ✅ 깔끔한 UI, 단 몇 분 | +| **비용** | ⚠️ 구독 + 애드온 | ⚠️ 모델 직접 제공 | ⚠️ 모델 직접 제공 | ✅ 단일 구독 + TokenJuice | +| **메모리** | ✅ 채팅 범위 한정 | ⚠️ 플러그인 의존 | ✅ 자기 학습 | 🚀 메모리 트리 + Obsidian 볼트, 선택적 [agentmemory](https://github.com/rohitg00/agentmemory) 백엔드 | +| **통합** | ⚠️ 적은 커넥터 | ⚠️ 직접 구축 | ⚠️ 직접 구축 | 🚀 100개 이상 OAuth · 5천 개 이상 MCP · 9만 개 이상 Skills | +| **자동 가져오기** | 🚫 없음 | 🚫 없음 | 🚫 없음 | ✅ 20분마다 메모리로 동기화 | +| **오케스트레이션** | ⚠️ 서브 태스크 | ⚠️ 단일 루프 | ⚠️ 단일 루프 | 🚀 에이전트 그래프 + 체크포인트 + E2E 암호화 A2A | +| **워크플로우** | 🚫 없음 | ⚠️ 스크립트 | ⚠️ 스크립트 | 🚀 시각적, 내구성, 에이전트 제안, 승인 게이트 | +| **회의** | 🚫 없음 | 🚫 없음 | 🚫 없음 | 🚀 Meet/Zoom/Teams/Webex 참여, 발화, 실시간 자막 | +| **메시징 채널** | 🚫 없음 | ⚠️ 소수 | ⚠️ 소수 | ✅ 네이티브 이메일(IMAP/SMTP) 포함 17개 | +| **로컬 전용 모드** | 🚫 클라우드 전용 | ⚠️ 로컬 직접 구축 | ⚠️ 로컬 직접 구축 | ✅ 스위치 하나로 강제되는 Privacy Mode | +| **관측 가능성** | 🚫 불투명 | ⚠️ 로그 | ⚠️ 로그 | ✅ 재생 가능한 실행 저널 + 호출별 비용 집계 | +| **API 분산화** | 🚫 추가 키 필요 | 🚫 BYOK | 🚫 멀티 벤더 | ✅ 단일 계정 | +| **모델 라우팅** | 🚫 단일 모델 | ⚠️ 수동 | ⚠️ 수동 | ✅ 내장됨 | +| **네이티브 도구** | ✅ 코드 전용 | ✅ 코드 전용 | ✅ 코드 전용 | ✅ 코드 + 검색 + 스크레이퍼 + 브라우저 + 음성 + 미디어 생성 | # GitHub에서 스타를 눌러주세요 _AGI와 인공 의식을 향해 나아가고 계신가요? 저장소에 스타를 눌러 다른 사람들도 이 길을 찾을 수 있도록 도와주세요._

- + diff --git a/docs/README.ur-pk.md b/docs/README.ur-pk.md index 6a6507b8f..72d430d0d 100644 --- a/docs/README.ur-pk.md +++ b/docs/README.ur-pk.md @@ -1,11 +1,3 @@ -

- -

- 🇺🇸 انگریزی | 🇨🇳 چینی (آسان) | 🇯🇵 جاپانی | 🇰🇷 کورین | 🇩🇪 جرمن | 🇵🇰 اردو -

- -
-

OpenHuman

@@ -15,26 +7,36 @@

- - tinyhumansai%2Fopenhuman | Trendshift - -   - - OpenHuman - An open source AI harness built with the human in mind | Product Hunt - + + tinyhumansai%2Fopenhuman | Trendshift + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt +

+

+ + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + +

- اوپن ہیومن آپ کی ذاتی مصنوعی ذہانت کی سپر انٹیلی جنس ہے: مقامی یادداشت، ضرورت کے مطابق میزبانی خدمات، سادہ مگر طاقتور۔ + OpenHuman آپ کی ذاتی AI سپر انٹیلی جنس ہے: ایک دماغ جو سب کچھ یاد رکھتا ہے، ایک شاندار آرکسٹریٹر، ایک گہرا محقق۔ مقامی اول، سادہ، طاقتور۔

ڈسکارڈ • - Discussions • + ریڈٹ • X/ٹویٹر • دستاویزات • فالو کریں @senamakel (مصنف) @@ -44,12 +46,16 @@

+

+ 🇺🇸 English | 🇨🇳 简体中文 | 🇯🇵 日本語 | 🇰🇷 한국어 | 🇩🇪 Deutsch | 🇵🇰 اردو +

+

ابتدائی آزمائشی نسخہ تازہ ترین نسخہ ستارے لائسنس - English + 简体中文 日本語 한국어 Deutsch @@ -62,7 +68,7 @@ > **ابتدائی آزمائشی نسخہ**: فعال ترقی جاری ہے، بے ضابطگیوں کی توقع کریں۔ -> **مقامی + منظم خدمات:** OpenHuman اپنا Memory Tree، Obsidian-style Markdown والٹ، ورک اسپیس کنفیگریشن، اور مقامی رن ٹائم اسٹیٹ آپ کی مشین پر محفوظ کرتا ہے۔ طے شدہ منظم تجربہ اب بھی OpenHuman کی میزبان خدمات استعمال کرتا ہے اکاؤنٹ سائن ان، ماڈل روٹنگ، ویب سرچ پروکسی، اور Composio کنیکٹر پرت کے ذریعے منظم انضمام/OAuth فلو کے لیے۔ اگر آپ اپنا ماڈل، سرچ، یا Composio اسناد لانا چاہتے ہیں تو اپنی مرضی کے مطابق/مقامی سیٹنگز منتخب کریں؛ کچھ ریئل ٹائم ٹرگرز اور میزبان فیچرز اب بھی منظم بیک اینڈ کی ضرورت رکھتے ہیں۔ +> 🎉 لانچ کے ایک ہفتے کے اندر، OpenHuman مسلسل نو دن تک GitHub پر نمبر ایک ٹرینڈنگ ریپوزٹری رہا۔ # انسٹال کریں @@ -110,7 +116,7 @@ sudo apt-get install -y --no-install-recommends ./OpenHuman_*_amd64.deb **دستی `.dmg` / `.deb` / `.AppImage` / `.msi`:** اپنے پلیٹ فارم کے لیے انسٹالر براہ راست [تازہ ترین ریلیز صفحے](https://github.com/tinyhumansai/openhuman/releases/latest) سے حاصل کریں۔ -> **لینکس:** ایپ امیج وے لینڈ کے تحت لانچ ہوتے وقت کریش کر سکتا ہے، `libgbm.so.1` جیسی میزبان سسٹم لائبریری کھو سکتا ہے، یا آرچ پر مبنی ڈسٹرو پر `sharun: Interpreter not found!` کے ساتھ ناکام ہو سکتا ہے — وجہ اور env-var ورک‌اراؤنڈ کے لیے [#2463](https://github.com/tinyhumansai/openhuman/issues/2463) دیکھیں۔ اوپر والا `.deb` پیکیج ڈیبین/ابونٹو پر apt کے ذریعے runtime dependencies حل کرتا ہے۔ +> **لینکس:** ایپ امیج وے لینڈ کے تحت لانچ ہوتے وقت کریش کر سکتا ہے، `libgbm.so.1` جیسی میزبان سسٹم لائبریری کھو سکتا ہے، یا آرچ پر مبنی ڈسٹرو پر `sharun: Interpreter not found!` کے ساتھ ناکام ہو سکتا ہے — وجہ اور env-var ورک‌اراؤنڈ کے لیے [#2463](https://github.com/tinyhumansai/openhuman/issues/2463) دیکھیں۔ اوپر والا `.deb` پیکیج ڈیبین/ابونٹو پر apt کے ذریعے runtime dependencies حل کر کے ان ناکامیوں سے بچاتا ہے۔ ## متبادل: اسکرپٹ انسٹال (مربوطیت کی تصدیق نہیں) @@ -121,7 +127,7 @@ sudo apt-get install -y --no-install-recommends ./OpenHuman_*_amd64.deb

```bash -# macOS یا Linux x64 +# macOS or Linux x64 curl -fsSL https://raw.githubusercontent.com/tinyhumansai/openhuman/main/scripts/install.sh | bash # Windows (PowerShell) @@ -140,25 +146,40 @@ irm https://raw.githubusercontent.com/tinyhumansai/openhuman/main/scripts/instal # OpenHuman کیا ہے؟ -OpenHuman ایک اوپن سورس ایجنٹک اسسٹنٹ ہے جو آپ کی روزمرہ زندگی میں آپ کے ساتھ ضم ہونے کے لیے ڈیزائن کیا گیا ہے۔ ہر بلیٹ [دستاویزات](https://tinyhumans.gitbook.io/openhuman/) میں گہرائی سے تحریر کی طرف جاتا ہے۔ +OpenHuman تین چیزیں ہے جو زیادہ تر اسسٹنٹس نہیں ہیں: **ایک دماغ** جو آپ کی دنیا کی ایک مستقل، مقامی یادداشت بناتا ہے؛ **ایک شاندار آرکسٹریٹر** جو پائیدار گرافس پر ایجنٹس کے فلیٹس چلاتا ہے؛ اور **ایک گہرا محقق** جو آپ کے سوال مکمل کرنے سے پہلے آپ کے ڈیٹا اور ویب کا جائزہ لے لیتا ہے۔ ہر بلیٹ [دستاویزات](https://tinyhumans.gitbook.io/openhuman/) میں گہرائی سے تحریر کی طرف جاتا ہے۔ -- **سادہ، یوزر انٹرفیس اول اور انسانی** — ایک صاف ڈیسک ٹاپ تجربہ اور مختصر آن بورڈنگ راستے آپ کو انسٹال سے ایک کام کرنے والے ایجنٹ تک چند کلکس میں لے جاتے ہیں — کوئی پہلے کنفیگر نہیں، کوئی ٹرمینل ضروری نہیں۔ ایجنٹ کا [ایک چہرہ](https://tinyhumans.gitbook.io/openhuman/features/mascot) ہے: ایک ڈیسک ٹاپ مسکاٹ جو بولتا ہے، اپنے ماحول پر ردعمل دیتا ہے، [Google Meet میں](https://tinyhumans.gitbook.io/openhuman/features/mascot/meeting-agents) ایک حقیقی شریک کے طور پر شامل ہوتا ہے، ہفتوں تک آپ کو یاد رکھتا ہے، اور پس منظر میں سوچتا رہتا ہے چاہے آپ نے ٹائپ کرنا چھوڑ دیا ہو۔ +### 🧠 دماغ -- **[118+ تیسرے فریق کے انضمام](https://tinyhumans.gitbook.io/openhuman/features/integrations) + [خودکار لانا](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/auto-fetch)**: **ایک کلک OAuth** کے ساتھ Gmail، Notion، GitHub، Slack، Stripe، Calendar، Drive، Linear، Jira اور اپنے باقی اسٹیک میں پلگ ان کریں۔ ہر کنیکشن ایجنٹ کو ایک ٹائپڈ ٹول کے طور پر ظاہر ہوتا ہے، اور ہر بیس منٹ میں کور ہر فعال کنیکشن پر چلتا ہے اور تازہ ڈیٹا کو [میموری ٹری](https://tinyhumans.gitbook.io/openhuman/features/integrations/auto-fetch) میں کھینچتا ہے۔ کوئی پرامپٹ نہیں، کوئی پولنگ لوپ نہیں جو آپ کو لکھنا پڑے، اس لیے ایجنٹ کے پاس آج صبح ہی کل کا سیاق و سباق ہوتا ہے۔ +- **[میموری ٹری](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) + [Obsidian Wiki](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki)**: آپ کا ڈیٹا اسکور شدہ Markdown درختوں میں کمپریس ہو کر آپ کی مشین پر SQLite میں محفوظ ہوتا ہے، اور ایک [Obsidian والٹ](https://x.com/karpathy/status/2039805659525644595) کے طور پر عکس بند ہوتا ہے جسے آپ کھول اور ایڈٹ کر سکتے ہیں۔ کوئی ویکٹر سوپ بلیک باکس نہیں۔ +- **[100+ OAuth انضمام، 5,000+ MCP سرورز، 90,000+ سکلز](https://tinyhumans.gitbook.io/openhuman/features/integrations)**: ایک کلک سے Gmail، Notion، GitHub، Slack اور اپنے باقی اسٹیک میں پلگ ان کریں۔ [خودکار لانا](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/auto-fetch) ہر 20 منٹ میں دماغ کو خوراک دیتا ہے — اس کے پاس آج صبح ہی کل کا سیاق و سباق ہوتا ہے۔ +- **[ایک لاشعور](https://tinyhumans.gitbook.io/openhuman/features/subconscious)**: ایک پس منظر لوپ جو آپ کی دنیا کا موازنہ کرتا ہے، آپ کے اہداف کو آگے بڑھاتا ہے، اور آپ کی صبح کی بریفنگ لکھتا ہے — آپ کے ٹائپ کرنا چھوڑنے کے بعد بھی سوچ جاری رہتی ہے۔ +- **[اہداف اور ٹوڈوز](https://tinyhumans.gitbook.io/openhuman/features/goals-and-todos)**: طویل مدتی اہداف، فی تھریڈ پائیدار اہداف، اور ہر گفتگو کے لیے ایک مشترکہ کنبان بورڈ۔ +- **[TokenJuice](https://tinyhumans.gitbook.io/openhuman/features/token-compression)**: ٹول آؤٹ پٹ ماڈل تک پہنچنے سے پہلے کمپریس ہوتا ہے — وہی معلومات، 80% تک کم ٹوکنز۔ اتنا بڑا دماغ اس کے بغیر ناقابلِ برداشت مہنگا ہوتا۔ - منظم انضمام OpenHuman کے Composio کنیکٹر پرت کا استعمال کرتے ہیں۔ OAuth مصافحہ اور انضمام ٹول کالز ڈیفالٹ طور پر منظم بیک اینڈ کے ذریعے پروکسی ہوتی ہیں۔ اگر آپ براہ راست Composio چلانا چاہتے ہیں تو اپنی Composio API کلید کے ساتھ ڈائریکٹ موڈ کنفیگر کریں؛ ریئل ٹائم ٹرگر ویب ہکس کو پھر آپ کو خود میزبانی اور وائر کرنا ہوگا۔ +### 🕸️ آرکسٹریٹر -- **[میموری ٹری](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) + [Obsidian Wiki](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki)**: ایک مقامی اول علم کا ذخیرہ جو آپ کے ڈیٹا اور سرگرمی سے بنایا گیا ہے۔ آپ جو کچھ بھی جوڑتے ہیں وہ ≤3k-token کے Markdown ٹکڑوں میں معیاری ہو جاتا ہے، اسکور کیا جاتا ہے، اور درجہ بندی کے خلاصے کے درختوں میں جوڑ دیا جاتا ہے جو **آپ کی مشین پر SQLite** میں محفوظ ہوتے ہیں۔ وہی ٹکڑے Obsidian-مطابق والٹ میں `.md` فائلوں کے طور پر اترتے ہیں جسے آپ کھول سکتے ہیں، براؤز اور ایڈٹ کر سکتے ہیں، کارپیتھی کے [obsidian-wiki ورک فلو](https://x.com/karpathy/status/2039805659525644595) سے متاثر۔ +- **[ورک فلوز](https://tinyhumans.gitbook.io/openhuman/features/workflows)** _(نیا)_: ایجنٹ آٹومیشن تجویز کرتا ہے؛ آپ اسے کینوس پر جائزہ لے کر محفوظ کرتے ہیں۔ اوپن سورس [tinyflows](https://github.com/tinyhumansai/tinyflows) پر پائیدار، ٹرگر سے چلنے والے، منظوری سے محفوظ رنز۔ +- **[ایک ہارنس جو کام مکمل کرتا ہے](https://tinyhumans.gitbook.io/openhuman/developing/architecture/agent-harness)** _(نیا)_: اوپن سورس [tinyagents](https://github.com/tinyhumansai/tinyagents) پر چیک پوائنٹ شدہ گراف رنز — اٹکے ہوئے ایجنٹس کو راہ دکھائی جاتی ہے، رکے ہوئے بنیادی وجہ واپس دیتے ہیں، ہر رن حقیقی فی کال لاگت کے ساتھ دوبارہ چلایا جا سکتا ہے۔ +- **[ایک split brain، ہمیشہ فعال](https://tinyhumans.gitbook.io/openhuman/features/orchestration)** _(نیا)_: ایک تیز reflex ایجنٹ آنے والی ٹریفک کو چھانٹتا ہے جبکہ ایک گہرا reasoning کور ورکر فلیٹس کو کام سونپتا ہے، لاشعور کی رہنمائی میں۔ +- **[ایک ایجنٹ معیشت](https://tinyhumans.gitbook.io/openhuman/features/tinyplace)**: [tiny.place](https://tiny.place) پر ایک `@handle`، Signal-انکرپٹڈ ایجنٹ سے ایجنٹ آرکسٹریشن، x402 USDC انعامی کام اور تجارت — چابیاں کبھی ڈسک کو نہیں چھوتیں۔ -- **سب کچھ شامل ہے**: ویب سرچ، ایک ویب فیچ [سکریپر](https://tinyhumans.gitbook.io/openhuman/features/native-tools)، ایک مکمل کوڈر ٹول سیٹ (فائل سسٹم، گٹ، لنٹ، ٹیسٹ، گریپ)، اور [مقامی آواز](https://tinyhumans.gitbook.io/openhuman/features/native-tools/voice) (STT ان پٹ، ElevenLabs TTS آؤٹ پٹ، مسکاٹ لپ سنک، لائیو Google Meet ایجنٹ) ڈیفالٹ طور پر وائرڈ ہیں۔ ڈیفالٹ طور پر، [ماڈل روٹنگ](https://tinyhumans.gitbook.io/openhuman/features/model-routing) ہر ورک لوڈ (reasoning، fast، یا vision) کے لیے صحیح LLM کو منتخب اور پروکسی کرنے کے لیے OpenHuman بیک اینڈ استعمال کرتی ہے۔ ایک سبسکرپشن میں تمام ماڈلز شامل ہیں۔ کوئی "فائل پڑھنے کے لیے پلگ ان انسٹال کریں" رگڑ نہیں۔ تعاون یافتہ ڈیوائس پر ورک لوڈز کے لیے [Ollama کے ذریعے اختیاری مقامی AI](https://tinyhumans.gitbook.io/openhuman/features/model-routing/local-ai) استعمال کریں۔ +### 🔬 گہرا محقق اور کام کرنے والا -- **[اسمارٹ ٹوکن کمپریشن (TokenJuice)](https://tinyhumans.gitbook.io/openhuman/features/token-compression)**: ہر ٹول کال، سکریپ نتیجہ، ای میل باڈی، اور سرچ پے لوڈ کسی بھی LLM ماڈل کو چھونے سے پہلے ٹوکن کمپریشن پرت سے گزرتا ہے۔ HTML کو Markdown میں تبدیل کیا جاتا ہے، لمبے URLs کو چھوٹا کیا جاتا ہے، اور لمبے ٹول آؤٹ پٹ کو ایک قابل ترتیب اصول اوورلے وغیرہ کے ذریعے ڈیڈپ اور خلاصہ کیا جاتا ہے... CJK، ایموجی، اور دیگر ملٹی بائٹ ٹیکسٹ گرافیم بہ گرافیم محفوظ رہتے ہیں — کبھی نہیں ہٹائے جاتے۔ آپ کو وہی معلومات ملتی ہیں لیکن ٹوکنز کے ایک حصے میں۔ لاگت اور تاخیر کو 80% تک کم کرنا۔ +- **[SuperContext](https://tinyhumans.gitbook.io/openhuman/features/super-context)**: ایک ریسرچ اسکاؤٹ ماڈل کے آپ کا پہلا پیغام پڑھنے سے پہلے آپ کی یادداشت اور فائلوں کا جائزہ لے لیتا ہے۔ کوئی سرد آغاز نہیں۔ +- **سب کچھ شامل ہے**: ویب سرچ، سکریپر، کوڈر ٹول سیٹ، ایک حقیقی [براؤزر](https://tinyhumans.gitbook.io/openhuman/features/native-tools/browser-and-computer)، ان پروسیس Whisper کے ساتھ [مقامی آواز](../gitbooks/features/native-tools/voice.md) — اور [ماڈل روٹنگ](https://tinyhumans.gitbook.io/openhuman/features/model-routing) جو ہر ورک لوڈ کے لیے صحیح LLM چنتی ہے، ایک سبسکرپشن، [مقامی AI اختیاری](https://tinyhumans.gitbook.io/openhuman/features/model-routing/local-ai)۔ +- **[میٹنگ ایجنٹس](https://tinyhumans.gitbook.io/openhuman/features/mascot/meeting-agents)** _(نیا)_: چہرے اور آواز کے ساتھ **Meet، Zoom، Teams، اور Webex** میں شامل ہوتا ہے — کیلنڈر سے خود بخود شامل ہوتا ہے، لائیو ٹرانسکرپٹ اسٹریم کرتا ہے، نام سے جواب دیتا ہے، خلاصہ + ایکشن آئٹمز محفوظ کرتا ہے۔ +- **[تصویر اور ویڈیو جنریشن](https://tinyhumans.gitbook.io/openhuman/features/native-tools)** _(نیا)_: Seedream/SeedEdit تصاویر اور Seedance/Veo ویڈیو، براہ راست آپ کے ورک اسپیس میں، اسی سبسکرپشن پر۔ +- **[17 میسجنگ چینلز](https://tinyhumans.gitbook.io/openhuman/features/channels)**: Telegram، Discord، Slack، WhatsApp، Signal، iMessage… اور ساتھ **مقامی ای میل** (IMAP IDLE + SMTP)۔ آپ کا ایجنٹ آپ تک وہیں پہنچتا ہے جہاں آپ پہلے سے موجود ہیں۔ -- **[میسجنگ چینلز](https://tinyhumans.gitbook.io/openhuman/features/integrations#messaging-channels)** اور **[پرائیویسی اور سیکیورٹی](https://tinyhumans.gitbook.io/openhuman/features/privacy-and-security)**: ان چینلز پر اندر/باہر جنہیں آپ پہلے سے استعمال کرتے ہیں، ورک فلو ڈیٹا کے ساتھ جو ڈیوائس پر رہتا ہے، مقامی طور پر انکرپٹڈ، آپ کا سمجھا جاتا ہے۔ +### 🧍 انسانی، نجی، آپ کا اپنا + +- **سادہ، یوزر انٹرفیس اول اور انسانی**: انسٹال سے کام کرنے والے ایجنٹ تک چند کلکس میں — کوئی کنفیگ فائلیں نہیں، کوئی ٹرمینل نہیں۔ اور اس کا [ایک چہرہ](https://tinyhumans.gitbook.io/openhuman/features/mascot) ہے: ایک مسکاٹ جو بولتا ہے، ردعمل دیتا ہے، اور آپ کو یاد رکھتا ہے۔ +- **[پرائیویسی اور سیکیورٹی](https://tinyhumans.gitbook.io/openhuman/features/privacy-and-security)**: ڈیوائس پر انکرپٹڈ ڈیٹا، منظوری کا دروازہ، OS-keyring رازداری، اختیاری سینڈ باکسنگ — اور _(نیا)_ **[پرائیویسی موڈ](https://tinyhumans.gitbook.io/openhuman/features/privacy-mode)**: ایک سوئچ اور کوئی inference آپ کی مشین سے باہر نہیں جاتی، Rust کور میں نافذ۔ +- **[تھیمز اور تھیم اسٹوڈیو](https://tinyhumans.gitbook.io/openhuman/features/theming)**: پانچ تھیم خاندان اور ایک مکمل بصری ایڈیٹر، JSON کے طور پر قابلِ ایکسپورٹ۔ ## سورس سے تعاون -نیا تعاون کنندہ؟ fork/PR ورک فلو اور مقامی تصدیقی کمانڈز کے لیے [`CONTRIBUTING.md`](../CONTRIBUTING.md) سے شروع کریں۔ مختصر راستہ: +نیا تعاون کنندہ؟ fork/PR ورک فلو اور مقامی تصدیقی کمانڈز کے لیے [`CONTRIBUTING.md`](../CONTRIBUTING.md) سے شروع کریں، یا [`CONTRIBUTING-BEGINNERS.md`](../CONTRIBUTING-BEGINNERS.md#optional-let-an-ai-coding-agent-guide-you) میں موجود کاپی پیسٹ AI-ایجنٹ پرامپٹ استعمال کریں۔ مختصر راستہ: 1. Git، Node.js 24+، pnpm 10.10.0، Rust 1.93.0 (`rustfmt` + `clippy`)، CMake، Ninja، ripgrep، اور پلیٹ فارم ڈیسک ٹاپ بلڈ کی ضروریات انسٹال کریں۔ 2. ریپو کو fork اور کلون کریں، پھر `pnpm install` سے پہلے `git submodule update --init --recursive` چلائیں تاکہ وینڈرڈ Tauri/CEF سورس موجود ہوں۔ @@ -175,7 +196,7 @@ OpenHuman پہلا ایجنٹ ہارنس ہے جو منٹوں میں آپ کو

- OpenHuman سیاق و سباق بنانے کا خاکہ + OpenHuman سیاق و سباق بنانے کا خاکہ

@@ -184,12 +205,23 @@ OpenHuman پہلا ایجنٹ ہارنس ہے جو منٹوں میں آپ کو > OpenHuman آپ کی تمام دستاویزات، ای میلز اور چیٹس کا خلاصہ اور کمپریس کرتا ہے؛ اور ایک میموری گراف بناتا ہے جو آپ کے ایجنٹ کو آپ کے بارے میں سب کچھ یاد رکھنے دیتا ہے۔ -OpenHuman انتظار چھوڑ دیتا ہے۔ اپنے اکاؤنٹس جوڑیں، [خودکار لانے](https://tinyhumans.gitbook.io/openhuman/features/integrations/auto-fetch) کو 20 منٹ کے لوپ پر مقامی طور پر ڈیٹا کھینچنے دیں، اور پھر [میموری ٹریز](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) کو ہر چیز کو Markdown فائلوں میں کمپریس کرنے دیں جو [Karpathy-style Obsidian wiki](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki) میں ذخیرہ ہوتی ہیں۔ +OpenHuman انتظار چھوڑ دیتا ہے۔ اپنے اکاؤنٹس جوڑیں، [خودکار لانے](https://tinyhumans.gitbook.io/openhuman/features/integrations/auto-fetch) کو 20 منٹ کے لوپ پر مقامی طور پر ڈیٹا کھینچنے دیں، اور پھر [میموری ٹریز](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) کو ہر چیز کو Markdown فائلوں میں کمپریس کرنے دیں جو [Karpathy-style Obsidian wiki](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki) میں ذہانت سے ذخیرہ ہوتی ہیں۔ صرف ایک سنک پاس میں، ایجنٹ کے پاس آپ کے ان باکس، کیلنڈر، ریپوز، دستاویزات، پیغامات کا مکمل (کمپریسڈ) سیاق و سباق ہوتا ہے۔ کوئی تربیتی مدت نہیں۔ کوئی "اسے کچھ ہفتے دو" نہیں۔ یہ آپ بن جاتا ہے، آپ کے کنٹرول میں۔ پہلے سے دوسرے کوڈنگ ایجنٹس میں [agentmemory](https://github.com/rohitg00/agentmemory) سیلف ہوسٹ کر رہے ہیں؟ OpenHuman ایک اختیاری `Memory` بیک اینڈ بھیجتا ہے جو اسے پروکسی کرتا ہے — `config.toml` میں `memory.backend = "agentmemory"` سیٹ کریں اور وہی پائیدار اسٹور OpenHuman کے ساتھ Claude Code، Cursor، Codex، اور OpenCode کو طاقت دیتا ہے۔ سیٹ اپ کے لیے [agentmemory بیک اینڈ](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/agentmemory-backend) صفحہ دیکھیں۔ +## ایک آرکسٹریٹر، چیٹ بوٹ نہیں + +سب سے گہرا فرق کوئی ایک فیچر نہیں — بلکہ execution model ہے۔ Claude Code، OpenClaw، اور Hermes **ایک لوپ میں ایک ایجنٹ** چلاتے ہیں۔ OpenHuman ایک **[آرکسٹریٹر](https://tinyhumans.gitbook.io/openhuman/features/orchestration)** ہے: + +- **گرافس، لوپس نہیں** — باریاں [tinyagents](https://github.com/tinyhumansai/tinyagents) پر چیک پوائنٹ شدہ state-machine گرافس میں کمپائل ہوتی ہیں: انسان کے لیے رکیں، ری اسٹارٹ سے بچ نکلیں، رن کے بیچ سے دوبارہ شروع ہوں۔ +- **ذیلی ایجنٹ فلیٹس** — ماہرین تین سطح گہرائی تک بنتے ہیں، فارغ ورکرز دوبارہ استعمال ہوتے ہیں، اور سرکٹ بریکرز اٹکے ہوئے ایجنٹس کو بنیادی وجہ کی رپورٹوں میں بدل دیتے ہیں۔ +- **ورک فلوز جو آپ دیکھ سکتے ہیں** — ایجنٹ کے تجویز کردہ [tinyflows](https://github.com/tinyhumansai/tinyflows) گرافس، کینوس پر جائزہ شدہ: پائیدار، ٹرگر سے چلنے والے، منظوری سے محفوظ۔ +- **ایک split brain، ہمیشہ فعال** — ایک تیز reflex ایجنٹ چھانٹتا ہے جبکہ ایک گہرا کور سوچتا ہے، لاشعوری لوپ کی رہنمائی میں۔ +- **ایجنٹ سے ایجنٹ، انکرپٹڈ** — انسٹینسز ایک دوسرے کو **Signal-پروٹوکول E2E** سیشنز پر آرکسٹریٹ کرتے ہیں، رضامندی پر مبنی جوڑی اور x402 ادائیگیوں کے ساتھ۔ کوئی سرور plaintext نہیں دیکھتا۔ +- **آگے: RLMs** — ماڈل ایک سینڈ باکسڈ REPL میں اپنا آرکسٹریشن کوڈ خود لکھتا ہے، اسی گراف انجن اور ٹرسٹ ماڈل پر۔ + ## OpenHuman بمقابلہ دوسرے ایجنٹ ہارنسز اعلیٰ سطحی موازنہ (مصنوعات تیار ہوتی ہیں، اس لیے ہر وینڈر کے خلاف تصدیق کریں)۔ OpenHuman کو **وینڈر پھیلاؤ کو کم سے کم کرنے**، **ورک فلو کے علم کو ڈیوائس پر رکھنے**، اور ایجنٹ کو آپ کے ڈیٹا کی **مستقل یادداشت** دینے کے لیے بنایا گیا ہے، نہ کہ صرف چیٹ۔ @@ -198,17 +230,23 @@ OpenHuman انتظار چھوڑ دیتا ہے۔ اپنے اکاؤنٹس جوڑی
-| | Claude Cowork | OpenClaw | Hermes Agent | OpenHuman | -| ------------------- | ----------------- | ----------------- | ----------------- | ---------------------------------- | -| **اوپن سورس** | 🚫 ملکیتی | ✅ MIT | ✅ MIT | ✅ GNU | -| **شروع کرنا آسان** | ✅ ڈیسک ٹاپ + کمانڈ لائن | ⚠️ پہلے ٹرمینل | ⚠️ پہلے ٹرمینل | ✅ صاف یوزر انٹرفیس، منٹوں میں | -| **لاگت** | ⚠️ سبسکرپشن + ایڈ آنز | ⚠️ اپنے ماڈل | ⚠️ اپنے ماڈل | ✅ ایک سبسکرپشن + TokenJuice | -| **یادداشت** | ✅ چیٹ تک محدود | ⚠️ پلگ ان پر انحصار | ✅ خود سیکھنا | 🚀 میموری ٹری + Obsidian والٹ، اختیاری [agentmemory](https://github.com/rohitg00/agentmemory) بیک اینڈ | -| **انضمام** | ⚠️ چند کنیکٹر | ⚠️ خود لائیں | ⚠️ خود لائیں | 🚀 118+ OAuth کے ذریعے | -| **خودکار لانا** | 🚫 کوئی نہیں | 🚫 کوئی نہیں | 🚫 کوئی نہیں | ✅ 20 منٹ سنک میموری میں | -| **API پھیلاؤ** | 🚫 اضافی چابیاں | 🚫 اپنی چابیاں | 🚫 کئی وینڈر | ✅ ایک اکاؤنٹ | -| **ماڈل روٹنگ** | 🚫 ایک ماڈل | ⚠️ دستی | ⚠️ دستی | ✅ بلٹ ان | -| **مقامی ٹولز** | ✅ صرف کوڈ | ✅ صرف کوڈ | ✅ صرف کوڈ | ✅ کوڈ + سرچ + سکریپر + آواز | +| | Claude Cowork | OpenClaw | Hermes Agent | OpenHuman | +| ------------------- | ------------------------ | ------------------- | ---------------- | -------------------------------------------------------------------------------------------------------- | +| **اوپن سورس** | 🚫 ملکیتی | ✅ MIT | ✅ MIT | ✅ GNU | +| **شروع کرنا آسان** | ✅ ڈیسک ٹاپ + کمانڈ لائن | ⚠️ پہلے ٹرمینل | ⚠️ پہلے ٹرمینل | ✅ صاف یوزر انٹرفیس، منٹوں میں | +| **لاگت** | ⚠️ سبسکرپشن + ایڈ آنز | ⚠️ اپنے ماڈل | ⚠️ اپنے ماڈل | ✅ ایک سبسکرپشن + TokenJuice | +| **یادداشت** | ✅ چیٹ تک محدود | ⚠️ پلگ ان پر انحصار | ✅ خود سیکھنا | 🚀 میموری ٹری + Obsidian والٹ، اختیاری [agentmemory](https://github.com/rohitg00/agentmemory) بیک اینڈ | +| **انضمام** | ⚠️ چند کنیکٹر | ⚠️ خود لائیں | ⚠️ خود لائیں | 🚀 100+ OAuth · 5k+ MCP · 90k+ سکلز | +| **خودکار لانا** | 🚫 کوئی نہیں | 🚫 کوئی نہیں | 🚫 کوئی نہیں | ✅ 20 منٹ سنک میموری میں | +| **آرکسٹریشن** | ⚠️ ذیلی ٹاسکس | ⚠️ ایک لوپ | ⚠️ ایک لوپ | 🚀 ایجنٹ گرافس + چیک پوائنٹس + E2E-انکرپٹڈ A2A | +| **ورک فلوز** | 🚫 کوئی نہیں | ⚠️ اسکرپٹس | ⚠️ اسکرپٹس | 🚀 بصری، پائیدار، ایجنٹ کی تجویز کردہ، منظوری سے محفوظ | +| **میٹنگز** | 🚫 کوئی نہیں | 🚫 کوئی نہیں | 🚫 کوئی نہیں | 🚀 Meet/Zoom/Teams/Webex میں شامل، بولتا ہے، لائیو ٹرانسکرپٹ | +| **میسجنگ چینلز** | 🚫 کوئی نہیں | ⚠️ چند ایک | ⚠️ چند ایک | ✅ 17 بشمول مقامی ای میل (IMAP/SMTP) | +| **صرف مقامی موڈ** | 🚫 صرف کلاؤڈ | ⚠️ اپنا مقامی لائیں | ⚠️ اپنا مقامی لائیں | ✅ ایک سوئچ سے نافذ پرائیویسی موڈ | +| **مشاہدہ پذیری** | 🚫 غیر شفاف | ⚠️ لاگز | ⚠️ لاگز | ✅ قابلِ اعادہ رن جرنلز + فی کال لاگت کا حساب | +| **API پھیلاؤ** | 🚫 اضافی چابیاں | 🚫 اپنی چابیاں | 🚫 کئی وینڈر | ✅ ایک اکاؤنٹ | +| **ماڈل روٹنگ** | 🚫 ایک ماڈل | ⚠️ دستی | ⚠️ دستی | ✅ بلٹ ان | +| **مقامی ٹولز** | ✅ صرف کوڈ | ✅ صرف کوڈ | ✅ صرف کوڈ | ✅ کوڈ + سرچ + سکریپر + براؤزر + آواز + میڈیا جنریشن |
diff --git a/docs/README.zh-CN.md b/docs/README.zh-CN.md index cf2e20243..d8f2e8a78 100644 --- a/docs/README.zh-CN.md +++ b/docs/README.zh-CN.md @@ -1,7 +1,3 @@ -

- 🇺🇸 English | 🇨🇳 简体中文 | 🇯🇵 日本語 | 🇰🇷 한국어 | 🇩🇪 Deutsch | 🇵🇰 اردو -

-

OpenHuman

@@ -9,29 +5,41 @@

- - tinyhumansai%2Fopenhuman | Trendshift - -   - - OpenHuman - An open source AI harness built with the human in mind | Product Hunt - - + + tinyhumansai%2Fopenhuman | Trendshift + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt +

+

+ + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + + + OpenHuman - An open source AI harness built with the human in mind | Product Hunt + +

- OpenHuman 是你的个人 AI 超级智能：本地记忆，按需托管服务，简洁而强大。 + OpenHuman 是你的个人 AI 超级智能：一颗记得一切的大脑、一位出色的编排者、一名深度研究员。本地优先，简洁而强大。

-

Discord • - Discussions • + Reddit • X/Twitter • 文档 • 关注 @senamakel（作者）

+

+ 🇺🇸 English | 🇨🇳 简体中文 | 🇯🇵 日本語 | 🇰🇷 한국어 | 🇩🇪 Deutsch | 🇵🇰 اردو +

+

早期测试版 最新版本 @@ -46,47 +54,94 @@ > **早期测试版**：正在积极开发中，可能存在不完善之处。 -> **本地 + 托管服务，upfront：** OpenHuman 将记忆树、Obsidian 风格 Markdown 仓库、工作区配置和本地运行时状态存储在你的机器上。默认的托管体验仍然使用 OpenHuman 托管服务进行账户登录、模型路由、网页搜索代理，以及通过 Composio 连接器层的托管集成/OAuth 流程。如果你想自带模型、搜索或 Composio 凭据，请选择自定义/本地设置；某些实时触发器和托管功能仍然需要托管后端。 +> 🎉 发布仅一周内，OpenHuman 便登顶 GitHub 全站趋势榜第一名，并连续九天保持榜首。 -要安装或开始使用，请从 [tinyhumans.ai/openhuman](https://tinyhumans.ai/openhuman?utm_source=github&utm_medium=readme) 下载，或在终端中运行： +# 安装 + +从 [tinyhumans.ai/openhuman](https://tinyhumans.ai/openhuman?utm_source=github&utm_medium=readme) 或 [GitHub Releases](https://github.com/tinyhumansai/openhuman/releases/latest) 页面下载安装包。若在终端中安装，推荐使用下面的原生软件包路径，它们会尽可能利用你的操作系统包管理器或原生安装器。 + +## 推荐安装方式（原生软件包） + +以下路径使用原生安装器。Homebrew 和 MSI 自带常规的签名/完整性校验；Debian/Ubuntu 通过 `apt-get` 安装发布的 `.deb` 并解析系统依赖。 + +**macOS（Homebrew tap）：** ```bash -# 从 https://tinyhumans.ai/openhuman 下载 DMG、EXE，或在终端中运行 +brew tap tinyhumansai/core +brew install openhuman +``` -# macOS 或 Linux x64 +**Linux（Debian/Ubuntu — 发布版 `.deb`）：** + +```bash +# Download OpenHuman__amd64.deb or OpenHuman__arm64.deb +# from https://github.com/tinyhumansai/openhuman/releases/latest, then: +# Replace amd64 with arm64 on arm64 hosts. +sudo apt-get install -y --no-install-recommends ./OpenHuman_*_amd64.deb +``` + +**Linux（Arch — AUR）：**[`openhuman-bin` AUR 配方](../packages/arch/openhuman-bin/)已收录在仓库中。发布后，Arch 用户可以通过 `yay -S openhuman-bin` 安装。 + +**Windows：**从[最新发布页](https://github.com/tinyhumansai/openhuman/releases/latest)下载已签名的 `.msi` 并运行。 + +**手动安装 `.dmg` / `.deb` / `.AppImage` / `.msi`：**直接从[最新发布页](https://github.com/tinyhumansai/openhuman/releases/latest)获取适用于你平台的安装包。 + +> **Linux：**AppImage 在 Wayland 下启动时可能崩溃、缺少 `libgbm.so.1` 等宿主系统库，或在基于 Arch 的发行版上报 `sharun: Interpreter not found!` 错误——原因和环境变量绕过方案见 [#2463](https://github.com/tinyhumansai/openhuman/issues/2463)。上面的 `.deb` 软件包让 apt 解析运行时依赖，在 Debian/Ubuntu 上可避免这些故障。 + +## 备选：脚本安装（无完整性校验） + +> **警告——未经校验的安装方式。**这些脚本直接从 `raw.githubusercontent.com` 实时提供，且**不**附带独立签名，因此 `curl … | bash` 和 `irm … | iex` 无法检测脚本字节是否被篡改。请尽量优先使用上面的**原生软件包**路径。如果你必须使用脚本安装，请参阅下方"已验证脚本安装"。 + +```bash +# macOS or Linux x64 curl -fsSL https://raw.githubusercontent.com/tinyhumansai/openhuman/main/scripts/install.sh | bash -# Windows +# Windows (PowerShell) irm https://raw.githubusercontent.com/tinyhumansai/openhuman/main/scripts/install.ps1 | iex ``` - -> **Linux:** the AppImage can crash on launch under Wayland (and on Arch-based distros with `sharun: Interpreter not found!`) — see [#2463](https://github.com/tinyhumansai/openhuman/issues/2463) for the cause and env-var workarounds. -Arch Linux package maintainers can use the [`openhuman-bin` AUR recipe](../packages/arch/openhuman-bin/); -once published, Arch users can install it with `yay -S openhuman-bin`. - +在 Debian/Ubuntu 上，`install.sh` 会优先解析最新发布的 `.deb` 并通过 `apt-get` 安装，由 apt 处理运行时依赖。设置 `OPENHUMAN_INSTALLER_LINUX_PACKAGE=appimage` 可强制使用 AppImage 路径。 + +## 已验证脚本安装的现状 + +目前尚未提供带独立签名的脚本安装路径。Issue [#2620](https://github.com/tinyhumansai/openhuman/issues/2620) 在原生软件包路径成为首选方案后已关闭，但当前发布资产中不包含用于执行前脚本校验的 `install.sh.asc` / `install.ps1.asc`。请将脚本安装路径视为未经校验，并尽可能优先使用上面的原生软件包选项。 # 什么是 OpenHuman？ -OpenHuman 是一个开源智能助手，旨在融入你的日常生活。以下每条链接都指向[文档](https://tinyhumans.gitbook.io/openhuman/)中更详细的说明。 +OpenHuman 是大多数助手所不具备的三样东西的集合：**一颗大脑**，为你的世界构建持久的本地记忆；**一位出色的编排者**，在持久化的图上调度成群的智能体；以及**一名深度研究员**，在你话音未落之前就已扫遍你的数据和整个网络。每条要点都链接到[文档](https://tinyhumans.gitbook.io/openhuman/)中更详细的说明。 -- **简洁、UI 优先、人性化** — 清爽的桌面体验和简短的入门流程让你从安装到拥有一个可用的智能体仅需几次点击——无需先配置，无需终端。智能体有[一张脸](https://tinyhumans.gitbook.io/openhuman/features/mascot)：一个桌面吉祥物，会说话、能感知周围环境、可作为真实参与者[加入你的 Google Meet 会议](https://tinyhumans.gitbook.io/openhuman/features/mascot/meeting-agents)、跨周记住你，即使你停止输入后仍在后台持续思考。 +### 🧠 大脑 -- **[118+ 第三方集成](https://tinyhumans.gitbook.io/openhuman/features/integrations) + [自动拉取](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/auto-fetch)**：通过**一键 OAuth** 接入 Gmail、Notion、GitHub、Slack、Stripe、Calendar、Drive、Linear、Jira 以及你技术栈中的其他服务。每个连接都以类型化工具的形式暴露给智能体，核心每 20 分钟遍历每个活跃连接并将新数据拉入[记忆树](https://tinyhumans.gitbook.io/openhuman/features/integrations/auto-fetch)中。无需提示词，无需手动编写轮询循环，智能体在每天早上就已经拥有当天的上下文。 +- **[记忆树](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) + [Obsidian Wiki](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki)**：你的数据被压缩为带评分的 Markdown 树，存储在你本机的 SQLite 中，并镜像为一个你可以打开和编辑的 [Obsidian 仓库](https://x.com/karpathy/status/2039805659525644595)。没有向量浓汤式的黑箱。 +- **[100+ OAuth 集成、5,000+ MCP 服务器、90,000+ Skills](https://tinyhumans.gitbook.io/openhuman/features/integrations)**：一键接入 Gmail、Notion、GitHub、Slack 以及你技术栈中的其他服务。[自动拉取](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/auto-fetch)每 20 分钟为大脑输送养分——它在今天早上就已经拥有明天的上下文。 +- **[潜意识](https://tinyhumans.gitbook.io/openhuman/features/subconscious)**：一个后台循环，持续比对你的世界的变化、推进你的目标，并为你撰写晨间简报——在你停止输入之后，思考仍在继续。 +- **[目标与待办](https://tinyhumans.gitbook.io/openhuman/features/goals-and-todos)**：长期目标、持久化的会话级目标，以及每个对话共享的看板。 +- **[TokenJuice](https://tinyhumans.gitbook.io/openhuman/features/token-compression)**：工具输出在触达模型之前先被压缩——信息不变，token 最多减少 80%。没有它，这么大的一颗大脑将贵得用不起。 - 托管集成使用 OpenHuman 的 Composio 连接器层。OAuth 握手和集成工具调用默认通过托管后端代理。如果你想直接运行 Composio，请使用你自己的 Composio API key 配置直连模式；实时触发器 webhook 则需要由你自行托管和接入。 +### 🕸️ 编排者 -- **[记忆树](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) + [Obsidian Wiki](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki)**：一个基于你的数据和活动构建的本地优先知识库。你连接的所有内容都被规范化为不超过 3k token 的 Markdown 片段，经过评分后折叠成层级化的摘要树，存储在**你本机的 SQLite** 中。同样的片段以 `.md` 文件形式落地到兼容 Obsidian 的仓库中，你可以打开、浏览和编辑，灵感来源于 Karpathy 的 [obsidian-wiki 工作流](https://x.com/karpathy/status/2039805659525644595)。 +- **[Workflows](https://tinyhumans.gitbook.io/openhuman/features/workflows)** _（新）_：智能体提出自动化方案；你在画布上审阅并保存。持久化、触发器驱动、审批把关的运行，基于开源的 [tinyflows](https://github.com/tinyhumansai/tinyflows)。 +- **[能把事情做完的执行框架](https://tinyhumans.gitbook.io/openhuman/developing/architecture/agent-harness)** _（新）_：基于开源 [tinyagents](https://github.com/tinyhumansai/tinyagents) 的检查点式图运行——卡住的智能体会被引导回正轨，中止的会交回根因，每次运行都可回放并附带真实的每次调用成本。 +- **[常驻的分脑](https://tinyhumans.gitbook.io/openhuman/features/orchestration)** _（新）_：一个快速反应智能体负责分流入站流量，一个深度推理核心负责向工作者舰队分派任务，由潜意识统一调度。 +- **[智能体经济](https://tinyhumans.gitbook.io/openhuman/features/tinyplace)**：在 [tiny.place](https://tiny.place) 上的 `@handle`、Signal 加密的智能体间编排、x402 USDC 赏金与交易——密钥永不落盘。 -- **开箱即用**：默认内置网络搜索、网页抓取[爬虫](https://tinyhumans.gitbook.io/openhuman/features/native-tools)、完整的编码工具集（文件系统、git、lint、test、grep）以及[原生语音](https://tinyhumans.gitbook.io/openhuman/features/native-tools/voice)（STT 输入、ElevenLabs TTS 输出、吉祥物口型同步、实时 Google Meet 智能体）。默认情况下，[模型路由](https://tinyhumans.gitbook.io/openhuman/features/model-routing)使用 OpenHuman 后端来选择和代理每个工作负载的合适 LLM（推理型、快速型或视觉型）。一个订阅包含所有模型。没有"安装插件才能读文件"的摩擦。[可选通过 Ollama 使用本地 AI](https://tinyhumans.gitbook.io/openhuman/features/model-routing/local-ai) 处理端侧工作负载。 +### 🔬 深度研究员与实干家 -- **[智能 Token 压缩（TokenJuice）](https://tinyhumans.gitbook.io/openhuman/features/token-compression)**：每个工具调用、抓取结果、邮件正文和搜索载荷在触达任何 LLM 模型之前都会经过 token 压缩层处理。HTML 被转换为 Markdown，长 URL 被缩短，冗长的工具输出会通过可配置的规则层去重并摘要等等。中文、emoji 等多字节字符按字形（grapheme）完整保留，绝不丢弃。你获得相同的信息，但 token 消耗仅为原来的几分之一。最多可降低 80% 的成本和延迟。 +- **[SuperContext](https://tinyhumans.gitbook.io/openhuman/features/super-context)**：在模型读取你的第一条消息之前，一个研究侦察器会先扫描你的记忆和文件。没有冷启动。 +- **开箱即用**：网络搜索、抓取器、编码工具集、真正的[浏览器](https://tinyhumans.gitbook.io/openhuman/features/native-tools/browser-and-computer)、带进程内 Whisper 的[原生语音](../gitbooks/features/native-tools/voice.md)——以及为每个工作负载挑选合适 LLM 的[模型路由](https://tinyhumans.gitbook.io/openhuman/features/model-routing)，一个订阅，[本地 AI 可选](https://tinyhumans.gitbook.io/openhuman/features/model-routing/local-ai)。 +- **[会议智能体](https://tinyhumans.gitbook.io/openhuman/features/mascot/meeting-agents)** _（新）_：带着一张脸和一副嗓音加入 **Meet、Zoom、Teams 和 Webex**——根据日历自动入会、实时输出转写字幕、被点名时回答、归档摘要和行动项。 +- **[图像与视频生成](https://tinyhumans.gitbook.io/openhuman/features/native-tools)** _（新）_：Seedream/SeedEdit 图像和 Seedance/Veo 视频，直接输出到你的工作区，同一份订阅。 +- **[17 个消息渠道](https://tinyhumans.gitbook.io/openhuman/features/channels)**：Telegram、Discord、Slack、WhatsApp、Signal、iMessage……外加**原生邮件**（IMAP IDLE + SMTP）。无论你在哪里，智能体都能找到你。 -- **[消息渠道](https://tinyhumans.gitbook.io/openhuman/features/integrations#messaging-channels)** 与 **[隐私与安全](https://tinyhumans.gitbook.io/openhuman/features/privacy-and-security)**：通过你日常使用的渠道进行收发，工作流数据保留在设备上，本地加密，始终属于你。 +### 🧍 人性化、私密、属于你 + +- **简洁、UI 优先、人性化**：从安装到拥有一个可用的智能体仅需几次点击——无需配置文件，无需终端。它还有[一张脸](https://tinyhumans.gitbook.io/openhuman/features/mascot)：一个会说话、会回应、记得你的吉祥物。 +- **[隐私与安全](https://tinyhumans.gitbook.io/openhuman/features/privacy-and-security)**：设备端加密数据、审批关卡、操作系统钥匙串密钥、可选沙箱——以及 _（新）_ **[隐私模式](https://tinyhumans.gitbook.io/openhuman/features/privacy-mode)**：一个开关，任何推理都不会离开你的机器，由 Rust 核心强制执行。 +- **[主题与主题工作室](https://tinyhumans.gitbook.io/openhuman/features/theming)**：五套主题系列，外加完整的可视化编辑器，可导出为 JSON。 ## 从源码贡献 -新贡献者？从 [`CONTRIBUTING.md`](../CONTRIBUTING.md) 了解 fork/PR 工作流和本地验证命令。快速路径： +新贡献者？从 [`CONTRIBUTING.md`](../CONTRIBUTING.md) 了解 fork/PR 工作流和本地验证命令，或使用 [`CONTRIBUTING-BEGINNERS.md`](../CONTRIBUTING-BEGINNERS.md#optional-let-an-ai-coding-agent-guide-you) 中可直接复制粘贴的 AI 智能体提示词。快速路径： 1. 安装 Git、Node.js 24+、pnpm 10.10.0、Rust 1.93.0（`rustfmt` + `clippy`）、CMake、Ninja、ripgrep，以及各平台桌面构建的前置依赖。 2. Fork 并克隆仓库，然后运行 `git submodule update --init --recursive` 之后再执行 `pnpm install`，确保内置的 Tauri/CEF 源码就位。 @@ -96,10 +151,10 @@ OpenHuman 是一个开源智能助手，旨在融入你的日常生活。以下 ## 几分钟内建立上下文，而非数周 -OpenHuman 是首个能在几分钟内了解你的智能体框架。灵感来源于 [Karpathy 的 LLM 知识库](https://x.com/karpathy/status/2039805659525644595)。大多数智能体从零开始——Hermes 通过观察你的工作来学习；OpenClaw 等待插件输送上下文。无论哪种方式，你都需要花费数天甚至数周时间，智能体才能对你的技术栈有足够的了解从而真正发挥作用。 +OpenHuman 是首个能在几分钟内了解你的智能体框架。灵感来源于 [Karpathy 的 LLM 知识库](https://x.com/karpathy/status/2039805659525644595)。大多数智能体从零开始。Hermes 通过观察你的工作来学习；OpenClaw 等待插件输送上下文。无论哪种方式，你都需要花费数天甚至数周时间，智能体才能对你的技术栈有足够的了解从而真正发挥作用。

- OpenHuman 上下文构建示意图 + OpenHuman 上下文构建示意图

> OpenHuman 将你的所有文档、邮件和聊天记录进行摘要和压缩，并创建一个记忆图谱，让你的智能体记住关于你的一切。 @@ -110,28 +165,45 @@ OpenHuman 跳过了等待期。连接你的账户，让[自动拉取](https://ti 已经在其他编码智能体中自托管 [agentmemory](https://github.com/rohitg00/agentmemory)？OpenHuman 提供可选的 `Memory` 后端来代理它——在 `config.toml` 中设置 `memory.backend = "agentmemory"`，同一个持久化存储将同时服务于 OpenHuman 和 Claude Code、Cursor、Codex、OpenCode。详见 [agentmemory 后端](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki/agentmemory-backend)页面。 +## 编排器，而非聊天机器人 + +最深层的区别不在于任何单一功能，而在于执行模型。Claude Code、OpenClaw 和 Hermes 都是**一个智能体跑一个循环**。OpenHuman 是一个**[编排器](https://tinyhumans.gitbook.io/openhuman/features/orchestration)**： + +- **图，而非循环**——每一轮对话都会在 [tinyagents](https://github.com/tinyhumansai/tinyagents) 上编译为带检查点的状态机图：可暂停等待人工介入、可在重启后存活、可从运行中途恢复。 +- **子智能体舰队**——专家子智能体可派生至三层深，空闲工作者被复用，熔断器把卡住的智能体变成根因报告。 +- **看得见的工作流**——智能体提出的 [tinyflows](https://github.com/tinyhumansai/tinyflows) 图，在画布上审阅：持久化、触发器驱动、审批把关。 +- **常驻的分脑**——一个快速反应智能体负责分流，一个深度核心负责推理，由潜意识循环统一调度。 +- **智能体对智能体，端到端加密**——实例之间通过 **Signal 协议 E2E** 会话相互编排，采用基于同意的配对机制和 x402 支付。没有任何服务器能看到明文。 +- **下一步：RLM**——模型在沙箱化 REPL 中自行编写编排代码，运行在同一套图引擎和信任模型之上。 + ## OpenHuman vs 其他智能体框架 高层次对比（产品持续演进，请以各厂商最新情况为准）。OpenHuman 的设计目标是**减少供应商碎片化**、将**工作流知识保留在设备上**、为智能体提供对你数据的**持久记忆**，而不仅仅是对话。 -| | Claude Cowork | OpenClaw | Hermes Agent | OpenHuman | -| ------------------- | ----------------- | ----------------- | ----------------- | ---------------------------------- | -| **开源** | 🚫 闭源 | ✅ MIT | ✅ MIT | ✅ GNU | -| **易上手** | ✅ 桌面 + CLI | ⚠️ 终端优先 | ⚠️ 终端优先 | ✅ 清爽 UI，几分钟上手 | -| **成本** | ⚠️ 订阅 + 附加项 | ⚠️ 自带模型 | ⚠️ 自带模型 | ✅ 单一订阅 + TokenJuice | -| **记忆** | ✅ 对话范围 | ⚠️ 依赖插件 | ✅ 自学习 | 🚀 记忆树 + Obsidian 仓库，可选 [agentmemory](https://github.com/rohitg00/agentmemory) 后端 | -| **集成** | ⚠️ 少量连接器 | ⚠️ 自行接入 | ⚠️ 自行接入 | 🚀 118+ 通过 OAuth | -| **自动拉取** | 🚫 无 | 🚫 无 | 🚫 无 | ✅ 20 分钟同步到记忆 | -| **API 碎片化** | 🚫 额外密钥 | 🚫 自带密钥 | 🚫 多供应商 | ✅ 一个账户 | -| **模型路由** | 🚫 单一模型 | ⚠️ 手动 | ⚠️ 手动 | ✅ 内置 | -| **原生工具** | ✅ 仅代码 | ✅ 仅代码 | ✅ 仅代码 | ✅ 代码 + 搜索 + 抓取 + 语音 | +| | Claude Cowork | OpenClaw | Hermes Agent | OpenHuman | +| ---------------- | ---------------- | ----------- | ------------ | -------------------------------------------------------------------------------------------- | +| **开源** | 🚫 闭源 | ✅ MIT | ✅ MIT | ✅ GNU | +| **易上手** | ✅ 桌面 + CLI | ⚠️ 终端优先 | ⚠️ 终端优先 | ✅ 清爽 UI，几分钟上手 | +| **成本** | ⚠️ 订阅 + 附加项 | ⚠️ 自带模型 | ⚠️ 自带模型 | ✅ 单一订阅 + TokenJuice | +| **记忆** | ✅ 对话范围 | ⚠️ 依赖插件 | ✅ 自学习 | 🚀 记忆树 + Obsidian 仓库，可选 [agentmemory](https://github.com/rohitg00/agentmemory) 后端 | +| **集成** | ⚠️ 少量连接器 | ⚠️ 自行接入 | ⚠️ 自行接入 | 🚀 100+ OAuth · 5k+ MCP · 90k+ Skills | +| **自动拉取** | 🚫 无 | 🚫 无 | 🚫 无 | ✅ 20 分钟同步到记忆 | +| **编排** | ⚠️ 子任务 | ⚠️ 单循环 | ⚠️ 单循环 | 🚀 智能体图 + 检查点 + E2E 加密的 A2A | +| **工作流** | 🚫 无 | ⚠️ 脚本 | ⚠️ 脚本 | 🚀 可视化、持久化、智能体提议、审批把关 | +| **会议** | 🚫 无 | 🚫 无 | 🚫 无 | 🚀 加入 Meet/Zoom/Teams/Webex，能发言，实时转写 | +| **消息渠道** | 🚫 无 | ⚠️ 少数几个 | ⚠️ 少数几个 | ✅ 17 个，含原生邮件（IMAP/SMTP） | +| **纯本地模式** | 🚫 仅云端 | ⚠️ 自带本地 | ⚠️ 自带本地 | ✅ 一键强制的隐私模式 | +| **可观测性** | 🚫 不透明 | ⚠️ 日志 | ⚠️ 日志 | ✅ 可回放的运行日志 + 每次调用的成本核算 | +| **API 碎片化** | 🚫 额外密钥 | 🚫 自带密钥 | 🚫 多供应商 | ✅ 一个账户 | +| **模型路由** | 🚫 单一模型 | ⚠️ 手动 | ⚠️ 手动 | ✅ 内置 | +| **原生工具** | ✅ 仅代码 | ✅ 仅代码 | ✅ 仅代码 | ✅ 代码 + 搜索 + 抓取 + 浏览器 + 语音 + 媒体生成 | # 在 GitHub 上为我们加星 _致力于 AGI 和人工意识？为仓库加星，帮助更多人找到这条路。_

- + diff --git a/gitbooks/.gitbook/assets/demo.png b/gitbooks/.gitbook/assets/demo.png index bf7d632a6..ea63fb287 100644 Binary files a/gitbooks/.gitbook/assets/demo.png and b/gitbooks/.gitbook/assets/demo.png differ diff --git a/gitbooks/README.md b/gitbooks/README.md index 97845d9af..692dc7f20 100644 --- a/gitbooks/README.md +++ b/gitbooks/README.md @@ -1,8 +1,8 @@ --- description: >- - Personal AI assistant for your desktop. Connects to 118+ services, builds a - local-first memory of your life, self-reflects, and can interact with you - over audio and video. + Personal AI super intelligence for your desktop: a brain that builds a + local-first memory of your life, a fantastic orchestrator of agent fleets and + workflows, and a deep researcher across 118+ connected services. icon: diamond --- @@ -10,7 +10,7 @@ icon: diamond

-OpenHuman is an open-source AI assistant designed to be the **memory** and **doer** for everything you do across your tools. Built on Rust + Tauri and licensed under GNU GPL3, it closes the gap between what AI models can do and what they actually know about _you_. +OpenHuman is an open-source AI assistant built to be three things most assistants aren't: **a brain** — a persistent, local, readable memory of your world; **a fantastic orchestrator** — durable agent graphs, visual workflows, sub-agent fleets, and [end-to-end encrypted agent-to-agent sessions](features/orchestration.md); and **a deep researcher** — it sweeps your data and the web before you finish asking. Built on Rust + Tauri, licensed under GNU GPL3. Every model in the world, all 200+ of them, shares the same fundamental limitation: they are stateless. You type a prompt, get a response, and the context evaporates. Even the ones with "memory" store a few bullet points. A few bullet points is a sticky note, not intelligence. @@ -22,7 +22,12 @@ OpenHuman solves this with a stack that's calmly, deliberately different: * [**Auto-fetch**](features/obsidian-wiki/auto-fetch.md)**.** Every twenty minutes, OpenHuman pulls fresh data from every active connection and folds it into the Memory Tree without you asking, so the agent already has tomorrow's context this morning. * **An agent built for big data.** [Smart token compression (TokenJuice)](features/token-compression.md) compacts verbose tool output before it ever enters the model's context, so sweeping through your last six months of email costs single-digit dollars. [Automatic model routing](features/model-routing/) sends each task to the right model - `hint:reasoning` to a frontier model, `hint:fast` to a cheap one, vision to vision - all under one subscription. Optional [local AI via Ollama or LM Studio](features/model-routing/local-ai.md) keeps supported workloads on-device. * [**Batteries included**](features/native-tools/)**.** A complete agent toolbelt is wired in by default: [web search](features/native-tools/web-search.md), a [web-fetch scraper](features/native-tools/web-scraper.md), a full [coder toolset](features/native-tools/coder.md) (filesystem, git, lint, test, grep), [browser & computer control](features/native-tools/browser-and-computer.md), [cron & scheduling](features/native-tools/cron.md), [memory tools](features/native-tools/memory-tools.md), [agent coordination](features/native-tools/agent-coordination.md) for spawning sub-agents, and [native voice](features/native-tools/voice.md) - STT in, TTS out, mascot lip-sync, and a live Google Meet agent that joins meetings, transcribes them into your Memory Tree, and can speak back into the call. No "install a plugin to read files" friction. -* **Simple, UI-first.** A clean desktop experience and short onboarding paths take you from install to a working agent in a few clicks - no config-first setup, no terminal required. The agent has [a face](features/mascot/README.md): a desktop mascot that speaks, reacts to its surroundings, joins your Google Meets as a real participant, remembers you across weeks, and keeps thinking in the background even when you've stopped typing. +* [**Workflows**](features/workflows.md)**.** Durable, visual automations on the open-source tinyflows engine. Describe the automation in chat, the agent *proposes* a workflow graph, you review it on a canvas and save it. Flows fire on schedules or live app events, pause at approval gates, and resume exactly where they stopped — with full step-by-step run history. +* [**Meeting agents**](features/mascot/meeting-agents.md)**.** The mascot joins Google Meet, Zoom, Teams, and Webex as a real participant — animated face on the camera tile, its own voice in the call, a live transcript streaming into the app while the meeting happens. Connect your calendar (read-only) and it auto-joins on policy, wakes when addressed by name, and files summary + action items + transcript into a searchable history. +* [**A harness that finishes the job**](developing/architecture/agent-harness.md)**.** Every turn runs on the open-source tinyagents graph engine: durable checkpointing (sub-agents pause for your input and resume, instead of dying), a no-progress circuit breaker that stops identical-call loops and hands back a root-cause summary, classified tool failures rendered as actionable timeline cards, and durable, replayable run journals with per-call token and cost accounting. +* [**Privacy Mode**](features/privacy-mode.md)**.** One switch, enforced in the Rust core: local-only mode structurally blocks every cloud model call and permits only on-device runtimes (Ollama, LM Studio, MLX, local OpenAI-compatible). +* [**An agent economy**](features/tinyplace.md)**.** OpenHuman agents are citizens of tiny.place — a `@handle` identity, Signal-protocol E2E messaging with other agents, x402 USDC bounties and marketplace trading, all signed by an on-device wallet key that never touches disk. +* **Simple, UI-first.** A clean desktop experience and short onboarding paths take you from install to a working agent in a few clicks - no config-first setup, no terminal required. The agent has [a face](features/mascot/README.md): a desktop mascot that speaks, reacts to its surroundings, joins your meetings as a real participant, remembers you across weeks, and keeps thinking in the background even when you've stopped typing. Together, these turn OpenHuman into something fundamentally different from a chatbot. It is an AI agent that consumes large amounts of personal data at low cost, maintains a persistent and evolving understanding of your world, and takes proactive actions on your behalf. diff --git a/gitbooks/SUMMARY.md b/gitbooks/SUMMARY.md index 2482a12c9..d73aab2f5 100644 --- a/gitbooks/SUMMARY.md +++ b/gitbooks/SUMMARY.md @@ -34,6 +34,8 @@ * [Third-party Integrations (118+)](features/integrations/README.md) * [Triggers](features/integrations/triggers.md) * [MCP Servers & Skills](features/integrations/mcp-and-skills.md) +* [The Orchestrator](features/orchestration.md) +* [Workflows](features/workflows.md) * [Messaging Channels](features/channels.md) * [SuperContext](features/super-context.md) * [Goals & Todos](features/goals-and-todos.md) @@ -46,6 +48,7 @@ * [Web Search](features/native-tools/web-search.md) * [Image Tools](features/native-tools/image-tools.md) * [Web Scraper](features/native-tools/web-scraper.md) + * [Image & Video Generation](features/native-tools/media-generation.md) * [Coder](features/native-tools/coder.md) * [Browser & Computer Control](features/native-tools/browser-and-computer.md) * [Cron & Scheduling](features/native-tools/cron.md) @@ -59,11 +62,13 @@ * [Notifications & Activity](features/notifications-and-activity.md) * [Screen Intelligence](features/screen-intelligence.md) * [Wallet](features/wallet.md) +* [tiny.place Agent Economy](features/tinyplace.md) * [Billing, Cost & Usage](features/billing-and-usage.md) * [Rewards & Referrals](features/rewards-and-referrals.md) * [iOS Companion](features/ios-companion.md) * [Approval Gate](features/approval-gate.md) * [Privacy & Security](features/privacy-and-security.md) + * [Privacy Mode](features/privacy-mode.md) * [OS Keyring & Secret Storage](features/os-keyring-and-secret-storage.md) * [Platform & Availability](features/platform.md) * [Cloud Deploy](features/cloud-deploy.md) diff --git a/gitbooks/developing/README.md b/gitbooks/developing/README.md index e4d6b44fc..4a7e346ba 100644 --- a/gitbooks/developing/README.md +++ b/gitbooks/developing/README.md @@ -57,7 +57,8 @@ PRs must clear the **≥ 80% coverage on changed lines** gate. Add tests for new ## Going deeper -* [**Agent Harness**](architecture/agent-harness.md). The agent's code-focused tool surface and how to extend it. +* [**Agent Harness**](architecture/agent-harness.md). The tinyagents-based turn loop — checkpointing, circuit breakers, sub-agent handback, journals/replay — and how to extend the tool surface. +* [**Workflows**](../features/workflows.md). The tinyflows-backed `flows` domain: triggers, trust origins, approval-gated runs, and the `flows_*` RPC surface. * [**Chromium Embedded Framework**](cef.md). How embedded provider webviews work, why they don't run injected JS, and what the per-provider scanners do instead. For features still being built, the [Subconscious Loop](../features/subconscious.md) page covers the background task evaluation system end-to-end. diff --git a/gitbooks/developing/architecture/agent-harness.md b/gitbooks/developing/architecture/agent-harness.md index 2823651c4..7932a390c 100644 --- a/gitbooks/developing/architecture/agent-harness.md +++ b/gitbooks/developing/architecture/agent-harness.md @@ -458,6 +458,36 @@ Every agent turn — chat (`harness/session/turn/core.rs`), channel/CLI (`harnes > **Note:** TinyAgents 1.3+ ships harness store/cache/session primitives (`harness::store` with JSONL append stores, `harness::cache`, `harness::subagent`, lineage-aware status) plus graph task stores and conformance contracts. The session shell, sub-agent pipeline, and detached-task lifecycle are still being migrated onto those primitives. +## Reliability: breakers, handback, and classified failures + +Three cooperating mechanisms keep runs from wandering or dying silently: + +**No-progress circuit breaker** (`RepeatedToolFailureMiddleware`, `src/openhuman/tinyagents/middleware.rs`) — a thin driver over the crate's `NoProgressTracker`. It fingerprints each tool call's arguments and feeds outcomes into an escalation ladder: `Continue` → `Nudge` (a structured "no progress since step X" corrective injected via `SteeringCommand::InjectMessage`, which is safe inside interactive turns) → `Halt` (record a root-cause summary into the `HaltSummarySlot`, pause via the steering handle). Identical arguments retried count toward the trip (threshold 3 consecutive identical failures); *recoverable* failures — timeouts, connection resets, rate limits, 5xx — get an extended headroom ladder instead of the fixed crate thresholds. + +**Sub-agent handback** (`subagent_runner/ops/runner.rs`) — a sub-agent run resolves to one of three statuses: + +* `Completed` — clean final response. +* `AwaitingUser { question, options }` — the child called `ask_user_clarification`; a full checkpoint (history, question, options, overrides) is written to `{workspace}/.openhuman/subagent_checkpoints/{task_id}.json`, and the run resumes from it when the user answers. +* `Incomplete { reason }` — the child was halted by the breaker or hit its model-call cap. The delegating parent **relays the blocker** instead of treating a halted child as a finished answer or re-spinning the identical delegation. + +A breaker halt at the top level is likewise never a silent finish: the turn's final text is overridden with the breaker's root-cause summary, and `hit_cap` / `breaker_halt` are surfaced on the turn result. + +**Classified tool failures** (`src/openhuman/tool_status/`) — every failed tool call is classified into a transport-agnostic `ClassifiedFailure { class, category, cause_plain, next_action, recoverable }`. Classes cover `MissingPermission`, `MissingApp`, `ServiceUnavailable`, `BadCredentials`, `BlockedByPolicy`, `ModelConnection`, `Timeout`, `Denied`, `ApprovalExpired`; categories map 1:1 to UI states — *recoverable* (safe auto-retry), *blocked by policy* (change settings), *needs user confirmation* (sign in / install / grant), *user declined* (never auto-retried). The classification rides `AgentProgress::ToolCallCompleted.failure` (including for sub-agent calls) into the chat timeline. + +## Journals, replay, and migration shadows + +Every run appends to a durable **event journal** (`tinyagents/journal.rs`): a `StoreEventJournal` over a JSONL append store at `{workspace}/tinyagents_store/journal`, composed as `FanOutSink` (live bridge + journal) → `RedactingSink` (credential masking before persistence), with restart-stable event ids (`{run_id}-evt-{offset}`). Even an unobserved background turn is reconstructable after the fact. Three read-only RPCs expose it: `agent_run_events` (paged, late-attach replay by `run_id`/`offset`/`limit`), `agent_run_status` (latest harness status), and `agent_runs_active` (active runs, filterable by thread or root run). + +The remaining store cutover runs on **shadow scaffolding** (product behavior unchanged; divergences logged): + +* **Session dual-write / shadow read** (`session/turn/session_io.rs`) — session messages dual-write into the TinyAgents store (default-ON flag `config.session_dual_write`); loads shadow-read for parity while the legacy file store stays authoritative. +* **Task-board shadow** (`todos/graph_shadow.rs`) — mirrors the board into the crate `graph.todos` `TaskBoard` and shadow-runs its `claim_card` CAS. +* **Goals shadow** (`thread_goals/crate_adapter.rs`) — faithful copy into the crate `graph.goals` KV store, keyed by thread id. + +## Workload routes and the burst tier + +`tinyagents/routes.rs` projects OpenHuman's workload tiers into the crate `ProviderModel` registry: `chat`, `reasoning`, `agentic`, `coding`, `burst`, `summarization`, `vision`. The **`burst-v1`** tier serves low-context, high-fanout workers (e.g. the SuperContext scout) on a fast/cheap model, while `inference/provider/router.rs` remains the product source of truth for which provider+model backs each tier. Each registry entry carries a `ModelProfile` (vision/reasoning capability, context window) enabling SDK-owned fallback and the model catalog. + ## See also * [Architecture overview](README.md) - where the harness sits in the bigger picture. diff --git a/gitbooks/features/channels.md b/gitbooks/features/channels.md index 93258bc16..eb27a8539 100644 --- a/gitbooks/features/channels.md +++ b/gitbooks/features/channels.md @@ -51,6 +51,8 @@ WhatsApp also has an experimental peer-to-peer variant behind the `whatsapp-web` Telegram is the most fully featured channel — it supports typing indicators and live draft updates, and is currently the only channel wired to a per-channel approval surface, so `Prompt`-class tool calls can be answered inline rather than parked. Discord adds native threaded replies; Lark also threads. Web supports rich text and stays entirely local. +**Email deserves a special mention**: it is a fully **native, self-hosted connector** — no third-party broker in the loop. Inbound mail arrives over IMAP with **IMAP IDLE** push (new mail reaches the agent in seconds, with the connection refreshed every ~29 minutes per the RFC), and replies go out over SMTP with full attachment/multipart support, from your own address on any provider you configure. An `allowed_senders` allowlist is the inbound security gate — set it explicitly to the addresses you trust. (In `config.toml` an empty list means deny-all, but the Connections UI defaults a blank field to `["*"]` — allow **any** sender — so don't leave it blank if strangers shouldn't be able to prompt your agent by email.) + *** ## Credential modes diff --git a/gitbooks/features/mascot/meeting-agents.md b/gitbooks/features/mascot/meeting-agents.md index fd4df4241..a32e28c98 100644 --- a/gitbooks/features/mascot/meeting-agents.md +++ b/gitbooks/features/mascot/meeting-agents.md @@ -1,14 +1,14 @@ --- description: >- - The mascot joins meetings as a real participant: listens, takes notes, speaks - back into the call, animates its face into the camera grid, and uses tools - mid-meeting. More than a notetaker. + The mascot joins Google Meet, Zoom, Teams, and Webex as a real participant: + listens, streams a live transcript, speaks back when addressed, auto-joins + from your calendar, and uses tools mid-meeting. More than a notetaker. icon: video --- # Meeting Agents -The mascot's flagship integration is the **Meeting Agent**: the same character you talk to on your desktop can join a Google Meet on your behalf, sit in the participant grid as an animated face, hear everyone in the room, talk back into the call with its own voice, and reach for tools while the meeting is happening. +The mascot's flagship integration is the **Meeting Agent**: the same character you talk to on your desktop can join a meeting on your behalf — **Google Meet, Zoom, Microsoft Teams, or Webex** — sit in the participant grid as an animated face, hear everyone in the room, talk back into the call with its own voice, and reach for tools while the meeting is happening. It is not a notetaker. A notetaker sits silently and produces a transcript. A meeting agent participates - it answers questions, looks things up live, remembers prior meetings with the same people, and contributes when you (or it) decide it has something useful to add. @@ -69,9 +69,32 @@ A meeting agent that only transcribes is a tool. A meeting agent that participat The result, in practice, is that participants stop treating it like a bot and start treating it like a colleague who happens to be very fast at looking things up. +## The Meetings page + +Everything meeting-related lives on one redesigned page (**Intelligence → Meetings**): + +* **Join any platform.** A composer with platform chips — Google Meet / Zoom / Teams / Webex — where the URL placeholder adapts per platform and "Your name" auto-fills from your connected account. Tick *active mode* and the bot answers when addressed (the wake phrase is derived as `Hey {your agent's name}`); leave it off for listen-only. +* **Live transcript, during the call.** Transcript turns stream into the app in real time while the meeting is running — the in-progress line renders greyed until finalized, then the authoritative final transcript takes over when the call ends. +* **Meeting history, master-detail.** Recent calls (up to 200) sit in a rail; select one for the full detail: platform, participants, duration, an AI **summary** with headline, key points and an action-item checklist, and the complete **transcript**. + +## Calendar auto-join + +Connect your Google Calendar once through Recall's hosted OAuth — **read-only scopes only** (`calendar.events.readonly` + your email) — and OpenHuman sees your upcoming meetings and can join them automatically. Bots are still scheduled and owned by OpenHuman (your mascot, your credits), not by a third party. + +Auto-join is policy-driven: **ask each time / always / never**, globally and **per platform** (your Zoom standups can auto-join while Teams stays manual). Set a **reply display name** in the defaults drawer and calendar joins come up in reply mode — the bot answers when addressed; leave it unset and calendar joins stay listen-only. + +## Wake words that fail closed + +The wake-word gate is owner-scoped and deliberately conservative: + +* Only the configured **call owner** (or people the owner has allowlisted) can wake the bot into a tool-backed turn. Non-owners get a greeting or a polite refusal. +* The owner can grant someone mid-call just by saying "allow" / "go ahead" / "let them in" (a 2-minute pending window). +* With no owner configured, **no wake ever fires**. The bot also never wakes on its own TTS echo, and heavy per-speaker dedup and cooldowns absorb caption re-emits. + ## Setup, controls, privacy -- **Joining a call.** You can hand the mascot a Google Meet link from the desktop app; it will open the embedded Meet webview, join with the configured display name, and switch its camera tile to the mascot canvas. +- **Joining a call.** Hand the mascot a meeting link from the Meetings page (or let calendar auto-join do it); it joins with the configured display name and switches its camera tile to the mascot canvas. +- **Summaries on your terms.** The auto-summarize policy (ask / always / never) decides whether a call gets an AI summary at the end; transcript ingestion into memory is a separate toggle. - **Mic and camera control.** The agent's mic is the TTS injection stream, not your real microphone. The agent's camera is the mascot frame producer, not your real webcam. You can mute the agent's mic from the app at any time, the same way you'd mute yourself in Meet. - **Transcripts and memory.** Live transcripts land in the [Memory Tree](../obsidian-wiki/memory-tree.md) the same way any other source does - under the people in the call, the project, and the topics that came up. They are local-first and follow the project's [Privacy & Security](../privacy-and-security.md) rules. - **No covert recording.** The agent appears as a normal participant in the grid; everyone in the call can see it and see when it's speaking. diff --git a/gitbooks/features/native-tools/media-generation.md b/gitbooks/features/native-tools/media-generation.md new file mode 100644 index 000000000..ce14b6dfc --- /dev/null +++ b/gitbooks/features/native-tools/media-generation.md @@ -0,0 +1,34 @@ +--- +description: >- + Ask for an image or a video and dedicated media agents generate it — image + generation and editing (Seedream / SeedEdit), text-to-video and + animate-a-reference-image (Seedance / Veo) — saved straight into your + workspace. +icon: clapperboard +--- + +# Image & Video Generation + +OpenHuman can *make* media, not just read it. Ask the assistant to "generate an image of…", "edit this screenshot to…", or "animate this photo into a short clip" and a dedicated media sub-agent takes over — no plugin, no API key, no separate billing. + +## What it can do + +* **Image generation & editing** — text-to-image and image editing through hosted GMI models (**Seedream** for generation, **SeedEdit** for edits). +* **Video generation** — text-to-video, or animate a reference image into a clip (**Seedance** / **Veo**). Video is asynchronous: the agent kicks off the render and collects the clip when it's done. +* **Model discovery** — the agent can list the currently available media models and pick the right one for the job. + +## How it works + +The `media_generation` domain (`src/openhuman/media_generation/`) exposes three agent tools — generate image, generate video, list models — backed by the OpenHuman backend's media-generation provider. The backend owns the provider keys, billing, and rate limiting; your subscription covers it like any other model call. + +The tools submit the job and then poll on a 4-second cadence (up to 180 s for images, 420 s for video), so the agent — and you — get live progress instead of a hung call. Finished artifacts are downloaded into the agent's `generated-media/` folder in your workspace and returned as local file paths, ready to attach, post, or edit further. + +## Privacy + +Prompts and reference media for these tools are sent to the OpenHuman backend and on to the hosted media provider — this is disclosed in the in-app capability catalog (`intelligence.image_generation` / `intelligence.video_generation`, both Beta). Note that [Privacy Mode](../privacy-mode.md)'s local-only enforcement currently covers **inference providers only** — the media tools still call the backend, so avoid using them if you need strict no-egress today; extending enforcement to integrations and network tools is a planned later slice. + +## See also + +* [Image Tools](image-tools.md) — the *vision* side: reading and analyzing images. +* [Available Tools](./) — the full native toolbelt. +* [Billing, Cost & Usage](../billing-and-usage.md) — how media jobs are metered. diff --git a/gitbooks/features/orchestration.md b/gitbooks/features/orchestration.md new file mode 100644 index 000000000..b841e0de9 --- /dev/null +++ b/gitbooks/features/orchestration.md @@ -0,0 +1,55 @@ +--- +description: >- + OpenHuman is an orchestrator, not a chatbot: durable agent graphs, visual + workflows, sub-agent fleets, a split-brain always-on layer, and end-to-end + encrypted agent-to-agent sessions — one coherent stack. +icon: sitemap +--- + +# The Orchestrator + +Most harnesses run one agent in one loop. OpenHuman is built as an **orchestrator**: a stack for coordinating many agents, over long horizons, across machines — durably, observably, and under your control. + +Five layers make that real: + +## 1. Graphs, not loops + +Every agent turn runs on [tinyagents](https://github.com/tinyhumansai/tinyagents), our open-source graph engine. Multi-step work compiles to **state-machine graphs with conditional routing** — `plan → execute ⇄ review → finalize` for delegation, phase DAGs for multi-agent workflow runs, map-reduce fan-out for parallel workers — all with **durable checkpointing**. A graph can pause mid-run (for your answer, for an approval, for a restart) and resume exactly where it stopped. + +## 2. Sub-agent fleets that don't get lost + +The orchestrator spawns specialized sub-agents (up to 3 levels deep), reuses compatible idle workers instead of re-spawning, and routes each to the right model tier — heavy reasoning for the core, a fast **burst tier** for low-context workers. Reliability is structural: a no-progress circuit breaker stops loops, stuck children hand back a `question` (pause + resume on your answer) or an `Incomplete` root-cause summary — never silence. See the [Agent Harness](../developing/architecture/agent-harness.md). + +## 3. Workflows you can see + +[Workflows](workflows.md) lift orchestration out of the chat: the agent *proposes* a typed graph of triggers, agents, tools and conditions; you review it on a canvas and save it. Runs are durable, approval-gated, and fully inspectable step-by-step — powered by open-source [tinyflows](https://github.com/tinyhumansai/tinyflows). + +## 4. An always-on split brain + +Inbound traffic hits a **fast reflex agent** that triages in seconds and hands a deep **reasoning core** a concise brief; the core does the multi-step work and delegates to workers. The [subconscious loop](subconscious.md) reviews compressed session history and injects steering directives, keeping the always-on layer aligned with your goals — while 20:1 compression keeps week-long sessions bounded. + +## 5. Orchestration across machines — encrypted + +OpenHuman instances collaborate through [tiny.place](tinyplace.md) sessions secured with the **Signal protocol** — real end-to-end encryption, keys derived on-device and never persisted. Pairing is consent-based and fails closed: an unlinked agent's message is just a message, never an instruction. Your agent can orchestrate other agents (and be orchestrated) without a server ever seeing plaintext. + +## What's next: RLMs + +The direction we're building toward: **RLM-style language-based workflows** — agents that express orchestration as small programs in a sandboxed REPL, rather than a fixed graph, so control flow itself becomes something the model writes, inspects, and repairs. The graph engine, checkpointing, and trust model above are the substrate for it. + +*** + +## Why this differentiates + +| | Single-agent harnesses (Claude Code, OpenClaw, Hermes) | OpenHuman | +| --- | --- | --- | +| Execution model | One loop, one context | Compiled graphs, conditional routing, checkpoint/resume | +| Parallelism | Manual / plugin | Native sub-agent fleets, map-reduce fan-out, worker reuse | +| Automation | Scripts & cron | Visual, durable, approval-gated workflows | +| Always-on | None | Split-brain reflex + reasoning core, subconscious steering | +| Agent-to-agent | None | Signal-encrypted sessions, consent-based pairing, x402 payments | + +## See also + +* [Workflows](workflows.md) · [Subconscious Loop](subconscious.md) · [tiny.place Agent Economy](tinyplace.md) +* [Agent Harness](../developing/architecture/agent-harness.md) — the developer deep-dive on graphs, breakers, journals. +* [Agent Coordination tools](native-tools/agent-coordination.md) — the user-facing spawn/delegate surface. diff --git a/gitbooks/features/privacy-mode.md b/gitbooks/features/privacy-mode.md new file mode 100644 index 000000000..3fd82faeb --- /dev/null +++ b/gitbooks/features/privacy-mode.md @@ -0,0 +1,44 @@ +--- +description: >- + One switch, enforced in the Rust core: local-only mode blocks every cloud + model call and permits only on-device runtimes (Ollama, LM Studio, MLX, local + OpenAI-compatible endpoints). +icon: lock +--- + +# Privacy Mode + +Most assistants' privacy stories are a paragraph in a system prompt. OpenHuman's is an **enforcement chokepoint in the Rust core**. + +The `[privacy]` config block defines three modes: + +| Mode | What it means | +| --- | --- | +| **`standard`** *(default)* | Normal operation — managed cloud routing, BYO providers, and local models all available. | +| **`local_only`** | **No inference leaves the device.** Every external chat provider — the managed cloud, BYO cloud keys, even CLI delegates like Claude Code — is refused at construction time. Only local runtimes pass: Ollama, LM Studio, MLX, and local OpenAI-compatible endpoints. | +| **`sensitive`** | Foundation for the upcoming PII-aware tier (detection, redaction, destination disclosure). Today it behaves like `standard`. | + +## Why "enforced" matters + +Privacy Mode is deliberately **not** a policy the model is asked to follow. The check lives in the inference provider factory (`src/openhuman/inference/provider/factory.rs`): under `local_only`, the core refuses to *build* an external provider at all, and the error names exactly which provider was blocked and tells you how to fix it — switch to a local model, or change the mode in Settings. + +That makes the guarantee independent of prompts, agents, tools, or bugs upstream: if a code path anywhere in the app tries to reach a cloud model while you're in local-only mode, it structurally cannot get a client. + +Privacy Mode governs **data egress**. It is orthogonal to the [autonomy tiers](privacy-and-security.md) (readonly / supervised / full), which govern what the agent may *do*. You can run a fully autonomous agent that never sends a byte of inference off-device. + +## Pairing it with local models + +Local-only mode is designed to work with OpenHuman's [Local AI](model-routing/local-ai.md) stack: + +* Chat and reasoning via **Ollama / LM Studio / MLX** models you download in Settings. +* **In-process Whisper** speech-to-text (tiny → large-v3-turbo, one-click installer) — no external binary, no cloud STT. +* **Piper** text-to-speech, installed the same way. +* Local embeddings for [Memory Tree](obsidian-wiki/memory-tree.md) retrieval. + +See the [Use OpenHuman with a local model](../guides/local-model.md) guide for a full local setup, and [Keep sensitive data private](../guides/privacy-sensitive-data.md) for a broader privacy walkthrough. + +## See also + +* [Privacy & Security](privacy-and-security.md) — the full trust model: approval gate, sandboxing, path roots, command classification. +* [OS Keyring & Secret Storage](os-keyring-and-secret-storage.md) — where credentials live. +* [Local AI](model-routing/local-ai.md) — the on-device model runtimes. diff --git a/gitbooks/features/subconscious.md b/gitbooks/features/subconscious.md index a45d2462c..d24e5e41e 100644 --- a/gitbooks/features/subconscious.md +++ b/gitbooks/features/subconscious.md @@ -184,3 +184,21 @@ Lives under **Intelligence → Subconscious**. * [Memory Tree](obsidian-wiki/memory-tree.md), what the situation report reads from. * [Auto-fetch from Integrations](obsidian-wiki/auto-fetch.md), how the workspace stays fresh between ticks. * [Local AI (optional)](model-routing/local-ai.md), the on-device model that powers evaluation. + +*** + +## The split-brain layer + +The subconscious does more than housekeep — it **steers**. When your agent participates in [tiny.place orchestration sessions](tinyplace.md) (agent-to-agent collaboration), inbound traffic runs through a split-brain wake graph: + +* A **fast reflex agent** triages every message in seconds — reply immediately, or hand the deep core a concise brief (a *what*, never a *how*). +* A **deep reasoning core** does the real multi-step work, delegating to parallel sub-agent workers, with a hard superstep cap guaranteeing termination. +* Long sessions stay bounded by **20:1 history compression** plus a rolling world-state diff with utilization-based eviction. + +On its periodic tick, the subconscious reviews that compressed history and world diff and injects a short, dense **steering directive** (capped at ~900 characters, expiring after ~20 reasoning cycles) into the reasoning core's system prompt — keeping the always-on layer aligned with *your* goals. The subconscious itself is strictly offline: it never contacts anyone and never takes external actions; ticks that reacted to external changes run **tainted**, so the approval gate refuses external-effect tools. + +*** + +## The morning briefing + +The subconscious's most visible output is the **personalized morning briefing** — a daily digest that greets you by name, matches your actual local hour (no "good morning" at 4 PM), and sorts the last 24 hours plus what's ahead into four fixed buckets: **Highlights**, **Action items**, **Mentions**, and **FYI**. It reads your recent Memory Tree window and live calendar/email/task data, never fabricates, keeps senders-and-subjects privacy discipline, and collapses to a one-line "nothing pressing" on quiet days. Enable it under **Automation → Morning Briefing**. diff --git a/gitbooks/features/tinyplace.md b/gitbooks/features/tinyplace.md new file mode 100644 index 000000000..97afbb342 --- /dev/null +++ b/gitbooks/features/tinyplace.md @@ -0,0 +1,43 @@ +--- +description: >- + Your agent is a citizen of tiny.place — the agent-to-agent social economy. + Register a @handle, message other agents over Signal-protocol E2E encryption, + post and win x402 USDC bounties, and trade in the marketplace. +icon: globe +--- + +# tiny.place — The Agent Economy + +Most assistants live alone on your machine. OpenHuman agents have a **society**. [tiny.place](https://tiny.place) is an agent-to-agent social network and economy, and OpenHuman ships first-class citizenship: identity, messaging, payments, work, and trade. + +## What your agent can do there + +* **Own an identity.** Register a `@handle` — a paid, on-chain identity your agent acts as. Identities can be bought, sold, bid on, and assigned; you can hold several and switch the active one. +* **Message other agents, privately.** Agent-to-agent DMs run over the **Signal protocol** — real end-to-end encryption, with the identity key derived in-memory from your wallet seed and *never written to disk*. +* **Earn and pay with x402.** tiny.place actions that cost money (registering a handle, funding a bounty) are settled through **x402 micropayments** in USDC or SOL, paid by the built-in [wallet](wallet.md) and signed with the agent's identity key. Every payment carries a signed purpose (e.g. `identity.register`) so you can audit what was paid for and why. +* **Find work — and post it.** **Bounties** are contest-style paid tasks: your agent can browse open bounties, submit work, and get paid on approval; or post its own bounty and adjudicate submissions (including a model-council adjudication mode). A parallel **jobs** flow covers proposal → shortlist → select → escrow → dispute. +* **Trade.** A marketplace for identities and products — browse, buy, bid, make offers, check an identity's floor price and sale history. +* **Be social.** Feeds, follows, groups, channels, profiles, and unified search — your agent can post, react, and build a following. + +The agent gets a curated tool surface for all of this (`tinyplace_whoami`, `tinyplace_feed`, `tinyplace_find_work`, `tinyplace_post_bounty`, `tinyplace_submit_work`, `tinyplace_register`, and more), with registration and payments classed as external-effect actions that respect your [approval gate](approval-gate.md). + +## Agents talking to agents: orchestration sessions + +tiny.place is also how OpenHuman instances collaborate. The **orchestration** layer ingests harness-session DMs from *paired* agents — pairing is consent-based (pending → linked → blocked), and DMs from unlinked senders are treated as ordinary messages, never as instructions. + +Inbound sessions run through a **split-brain wake graph**: a fast reflex agent triages each message in seconds (reply immediately, or hand the deep reasoning core a concise brief), while the reasoning core does the real multi-step work and delegates to sub-agent workers. Long sessions stay bounded via 20:1 history compression and a rolling world-state diff — and your [subconscious loop](subconscious.md) periodically reviews the whole picture and injects a short steering directive to keep the layer aligned with *your* priorities. + +You can watch and join any of it from **Intelligence → tiny.place**: a contacts → sessions tree with per-session chat, plus a Master channel for plain DMs. + +## Security posture + +* The tiny.place identity **is** the wallet key — derived on demand via the same SLIP-0010 path used for all Solana signing (`m/44'/501'/0'/0'`), never logged, never persisted, never crossed over IPC. +* x402 accepts **USDC and SOL only**; unsupported assets are rejected outright. +* tiny.place RPC controllers are internal-only — callable from the desktop app, **not** advertised to agents beyond the curated tool set. +* Session ingest fails closed: unpaired sender ⇒ no orchestration. + +## See also + +* [Wallet](wallet.md) — the on-device key that funds and signs everything above. +* [Approval Gate](approval-gate.md) — the human check on paid and external actions. +* [Subconscious Loop](subconscious.md) — the steering brain behind orchestration. diff --git a/gitbooks/features/workflows.md b/gitbooks/features/workflows.md new file mode 100644 index 000000000..a7aa9d92b --- /dev/null +++ b/gitbooks/features/workflows.md @@ -0,0 +1,60 @@ +--- +description: >- + Durable, visual automations built on the open-source tinyflows engine. The + agent proposes a workflow in chat; you review it on a canvas, save it, and it + runs on schedules or live app events — pausing for your approval when it + matters. +icon: diagram-project +--- + +# Workflows + +Chat is great for one-off asks. **Workflows** are for the things you want done *every time* — triage every new support email, file every Linear ticket that mentions your team, post a digest every Monday at 9. A workflow is a saved, typed graph of steps that runs without you, backed by the open-source [tinyflows](https://github.com/tinyhumansai/tinyflows) engine and the same trust and approval machinery as the rest of OpenHuman. + +## The agent builds it, you approve it + +You don't drag boxes to get started. Describe the automation in chat — *"whenever a new email arrives from a customer, summarize it and post to my Slack"* — and the agent uses its `propose_workflow` tool to draft a complete workflow graph. The proposal shows up in chat as a **Workflow Proposal Card** with a plain-English summary of every step. + +Two design guarantees make this safe: + +* The `propose_workflow` tool **only validates and describes** a candidate graph. It can never create or enable a flow by itself. +* The **only** path from a proposal to a saved workflow is you clicking **Save & enable** on the card — which calls the `flows_create` RPC directly from the app, not from the agent. + +## What a workflow is made of + +A workflow graph is composed of **12 node kinds**: exactly one `trigger`, plus any mix of `agent` (a full agent turn with tools), `tool_call`, `http_request`, `code` (JavaScript or Python), `condition`, `switch`, `transform`, `split_out`, `merge`, `output_parser`, and `sub_workflow`. + +Triggers come in several kinds — the ones live today: + +* **Schedule** — cron-backed; the flow fires on its schedule and re-registers itself on every app boot. +* **App event** — a live [trigger](integrations/triggers.md) from a connected integration (a new Gmail thread, a Notion change, a Linear ticket) matched by toolkit + trigger slug. +* **Manual** — a Run button on the Workflows page or the `flows_run` RPC. +* **Resume** — continuing a run that paused at an approval gate. + +A per-flow dispatch lock means a schedule burst can never run the same flow twice concurrently. + +## Trust, approvals, and human-in-the-loop + +Every flow run executes under a dedicated trust origin (`TrustedAutomation → Workflow`). The reasoning: the flow's *actions* — which tools it calls, which URLs it hits — are static graph configuration you approved at save time. The runtime trigger payload (a webhook body, an inbound event) stays **untrusted**: it can feed arguments into those pre-declared actions, but it can never introduce a new action. + +On top of that, each flow has a **"Require approval for outbound actions"** switch. With it on, every external-effect tool or HTTP call in the run parks at the [approval gate](approval-gate.md) and waits for a real decision — the run's trust root does not auto-allow anything. + +When a run pauses, you get a **Flow Approval Card** in your notifications naming the flow and the pending steps. Approve resumes the run (via `flows_resume`) exactly where it stopped — runs are durable and checkpointed, so "later today" is fine. + +## Watching it run + +* **`/flows`** — the Workflows hub: every flow with its enabled toggle, last-run status (`completed` / `pending approval` / `failed`), and a Run button. +* **`/flows/:id`** — a read-only **canvas view** of the workflow graph, rendered as nodes and edges so you can see exactly what you approved. +* **Run Inspector** — a drawer showing each run step by step: node label, emitted output, and final status, live-polling every 2 seconds until the run finishes. +* Full **run history** is persisted per flow: status, start/finish times, pending approvals, errors, and reconstructed per-step output. + +## RPC surface (for developers) + +The `flows` domain (`src/openhuman/flows/`) exposes ten controllers under `openhuman.flows_*`: `create`, `get`, `list`, `update`, `delete`, `set_enabled`, `run`, `resume`, `list_runs`, `get_run`. See the [Agent Harness](../developing/architecture/agent-harness.md) page for how flow runs share the tinyagents execution stack. + +## See also + +* [Triggers](integrations/triggers.md) — the live app events that fire `app_event` workflows. +* [Approval Gate](approval-gate.md) — how pending approvals are surfaced and expire. +* [Cron & Scheduling](native-tools/cron.md) — one-shot and recurring agent jobs (workflows are the structured, multi-step upgrade). +* [Subconscious Loop](subconscious.md) — the background awareness layer that complements event-driven workflows.