tinyhumansai/openhuman
Openhuman
OpenHuman is an AI application framework for building private intelligence systems and personal AI layers. It provides a system for deploying private AI assistants that execute technical tasks and manage personal knowledge bases.
The project features a model-agnostic request proxy that routes AI workloads to different large language models based on requirements for reasoning, speed, or vision. It integrates an OAuth-driven data integrator to synchronize personal information from external services into a local knowledge base composed of hierarchical Markdown summaries. The framework also includes a voice interface with synchronized avatars for participation in video conferencing.
The system covers autonomous agent orchestration with sandboxed tool execution for coding, web research, and filesystem manipulation. It implements a headless JSON-RPC server architecture for remote client access and includes a token-reduction pipeline to compress payloads and optimize model context. Security is handled through local data encryption, secure credential storage, and agent execution isolation within containers or OS jails.
The core logic can be deployed as a self-hosted containerized server to maintain data privacy and support local model execution.
Features
- Agentic LLM Frameworks - Provides a framework for deploying private AI assistants that execute technical tasks using a toolset for coding and web research.
- Private AI Infrastructure - Provides a comprehensive framework for hosting and managing personal AI assistants and agentic workflows on private, self-hosted servers.
- Agent Access Controls - Enforces security tiers and command classification to restrict the autonomous actions AI agents can perform.
- Agent Memory Stores - Provides a persistent memory store shared across agents and tools to maintain a consistent user context.
- Agent Task Execution - Executes proactive technical tasks using tools for web searching, coding, filesystem manipulation, and computer control.
- Model Routers - Directs AI workloads to specific large language models based on reasoning, speed, or vision requirements.
- AI Knowledge Management - Builds a private intelligence system that synchronizes external data into a local Markdown-based knowledge base.
- Autonomous Agent Orchestration - Orchestrates autonomous agents to execute technical workflows, web research, and filesystem manipulations.
- Local Model Execution - Executes AI workloads on-device using local model providers to ensure data privacy and offline availability.
- Model Routing - Routes AI workloads to different large language models based on specific requirements for reasoning, speed, or vision.
- Markdown-Based Knowledge Bases - Transforms unstructured third-party data into hierarchical Markdown summaries stored in a local database for agent memory.
- AI Knowledge Bases - Implements a local database that converts third-party data into hierarchical Markdown summaries to serve as memory for AI models.
- Local Knowledge Bases - Implements a system for parsing, indexing, and managing document collections locally via hierarchical Markdown summaries.
- Model Request Proxies - Provides a model-agnostic proxy that intercepts and routes requests to various interchangeable AI model providers.
- RPC Servers - Implements a headless JSON-RPC server to decouple backend AI logic from desktop and mobile client interfaces.
- Isolated Execution Sandboxes - Isolates agent execution within Docker containers or OS-level jails to prevent unauthorized host system access.
- Local Privacy Solutions - Ensures data privacy by storing workflow data and personal information using local encrypted storage.
- OAuth Integration Layers - Integrates third-party services via OAuth to synchronize personal information for use as AI agent context.
- Filesystem Root Restrictions - Restricts agents to specific read/write root directories using a fail-closed policy for filesystem isolation.
- Agentic Assistant Interfaces - Provides a desktop interface featuring a visual mascot that operates in the background and joins video meetings.
- Agent Deployment - Provides systems for provisioning and configuring private AI agent instances on the desktop.
- Agent Toolsets - Provides agents with a toolset for web scraping, filesystem operations, git management, and speech synthesis.
- Voice Agents - Integrates speech-to-text, text-to-speech, and synchronized avatars for participation in live video conferencing calls.
- AI Voice and Video Integration - Integrates speech-to-text and text-to-speech with synchronized avatars for participation in live video conferencing.
- Context Optimization Tools - Provides utilities to compress verbose tool outputs and route tasks to specific models to optimize the AI context window.
- Voice Interfaces - Provides a conversational system integrating speech-to-text, text-to-speech, and lip-synced avatars.
- Token Reduction Pipelines - Converts HTML to Markdown and summarizes verbose payloads to reduce latency and token costs before model transmission.
- Web Research Tools - Includes built-in tools for performing live web searches and scraping content to gather information for AI reasoning.
- Automated Context Synchronization - Automates the recurring fetching of personal information from external services to maintain an up-to-date AI context.
- Manual Memory Management - Allows users to browse, edit, and link processed data using a folder of Markdown files and a wiki-style workflow.
- Containerized Deployments - Packages the JSON-RPC server into a minimal runtime image for consistent deployment across environments.
- Self-Hosted AI Infrastructure - Provides tools for deploying and managing AI services and models on private hardware to maintain data sovereignty.
- Self-Hosted Deployment Platforms - Supports running the headless application backend in a containerized environment on private infrastructure.
- Event Systems - Coordinates low-latency synchronization across system domains using an event-based broadcasting architecture.
- Data Encryption - Secures sensitive workflow data and credentials on-device using authenticated encryption and OS-level credential managers.
- Credential Storage - Integrates with operating system credential managers to store secrets securely instead of using plain text.
- Memory-Based Key Protection - Protects master keys and decrypted buffers using authenticated encryption and memory-zeroing techniques.
- Non-Retentive Processing - Processes message content without retaining data for training or long-term storage to ensure user privacy.
- Remote Access Security - Secures remote access to the headless JSON-RPC server using bearer tokens configured via environment variables.
- Container-Based Sandboxes - Isolates autonomous agent tool execution within Docker containers or OS jails to protect the host system.
- Skill Sandboxing - Running external skills within defined boundaries to isolate behavior and protect the system.
- Event Bus Systems - Implements a typed broadcast system and request-response registry for decoupled communication between internal system domains.
- Headless Server Hosting - Deploys core logic as a headless JSON-RPC server on remote providers to enable multi-device access.
- AI Agents - Privacy-focused local desktop AI agent with memory systems.
- AI Assistant Tools - Agentic assistant with desktop UI and integrations.
- AI Assistants and Tools - Agentic assistant with desktop UI and memory.
