Contents
Foundations
Foundations
- Why outheisOn sovereignty, cognition, and why the extractive model fails.
- Design PrinciplesOn building tools that support life instead of consuming it.
- Information and SemanticsWhy information is not data — and what follows from that for a personal AI system.
- Attention as ArchitectureLearned attention mechanisms can replace complex sequential processing
- Annotation as Ground TruthOn the structural parallel between supervised learning and human correction — and what it...
- Tags as Scaffolding: A Deprecation PathFull free text analysis is not yet feasible due to general limitations of current models, but...
Design
Design
- DesignHow outheis is designed — the principles, structure, and reasoning behind the architecture.
- Operating System EngineeringMulti-agent AI systems face the same challenges that operating systems solved decades ago....
- Systems SurveyOperating systems and architectural patterns relevant to multi-agent AI system design.
- ArchitectureSystem structure, agent roles, ownership model, message protocol, dispatcher design, vault...
- Data Formats and ConventionsStructure and format of all data in outheis — enabling agents to make reliable assumptions about...
- Related WorkSurvey of existing multi-agent frameworks and how outheis differs.
- Agent Prompts and Communication RulesHow agents communicate — with users, with each other, and with themselves. All prompts are...
- The Hybrid Memory Stack: Where Code Ends and LLM BeginsThe sharp boundary between deterministic structure and learned meaning — why neither pure code...
- The Quality Threshold: Why Capability Doesn't Scale LinearlyOn emergent abilities, the limits of parallelism, and the minimum viable model.
Implementation
Implementation
- ImplementationHow outheis works in practice — agents, data, configuration, and interfaces.
- ArchitectureHow the pieces fit together.
- MemoryWhat outheis remembers about you — and how.
- AgendaTime management through three simple files.
- SkillsWhat agents know about how to act.
- AlanCode intelligence for development environments.
- HiroExternal action agent — not yet implemented.
- SignalSignal Messenger transport via signal-cli JSON-RPC.
- Annotation Feedback LoopHow `>` annotations route into the memory stack — the current gap and the planned implementation.
- VaultHow to structure your knowledge so outheis can work with it.
- ConfigurationComplete reference for `config.json`.
- GuideGetting started with outheis.
- MigrationHow to bring your existing knowledge into outheis.
- Web UILocal administration interface for outheis.
- Data SafetyHow outheis protects against prompt injection and untrusted content.
- Model EvaluationCapability benchmarks for local and cloud models used in outheis agents.
Installation
Installation
- InstallationGetting outheis running — release notes, setup, and configuration.
- Release NotesPublic beta — April 2026.
- Installation
- First StepsAfter `outheis init` and `outheis start`, run through these steps to verify everything works.
- Model SelectionWhich LLM to use with outheis — and for which agent.
- Communicating with outheisoutheis can be reached through three channels. All channels connect to the same dispatcher and...
- Remote Access via WireGuard (iPhone → Mac Mini)This guide documents how to set up secure remote access to the outheis WebUI from an iPhone...