{"repo":"munch2u-a11y/Helix-AGI","free":true,"listed":false,"github":"https://github.com/munch2u-a11y/Helix-AGI","clone":"git clone https://github.com/munch2u-a11y/Helix-AGI.git","description":"Helix-AGI is an Agentic AI System that uses a spatial-graph memory retrieval system. Agents receive a rolling adaptive system prompt of memories and skills that afford the agent a highly dynamic center of conceptual gravity. By avoiding long identity files and subagent-heavy workflows, Agents can develop continuously without spiraling costs.","language":"Python","stars":59,"topics":["agi","ai","agentic-ai-architecture","ai-agent","ai-agentic","ai-agents-automation","ai-agents-framework","ai-memory","ai-memory-system","autonomous-agents"],"license":"Apache-2.0","category":"ai-agents","readme_excerpt":"Helix AGI A continuous agent architecture with separated semantic recall, associative memory, and learned task cognition --- What is Helix AGI? Helix AGI is a multi-model agentic system that implements continuous autonomous operation, structured memory consolidation, and adaptive tool learning. Unlike agents that wait for a prompt, execute one chain, and terminate, Helix runs a continuous event-driven pulse with ACTIVE , REGULAR , RESTING , and scheduled DORMANT states. Its preconscious retrieval has deliberately separate jobs: multi-head mRAG supplies high-recall semantic advice from native 1024D embeddings, a read-only Context Office makes specialist desks bid for one shared injection budget over canonical Helix memory, and raw 8D spatial attention plus learned directed cluster transitions add a small non-semantic complement. Affect, stability, provenance, and Lagrangian metadata remain attached through injection. Helix can also turn naturally voiced intentions into durable tasks. In active task-cognition mode the main consciousness sees broad ability beliefs but no tool schemas; focus threads share Helix's memory and identity state, receive only the small authorized capability subset selected for the situation, and return accepted outcomes to the same event stream. Identity text is included only when the task actually depends on selfhood, values, history, relationships, preferences, or characteristic behavior. --- Architecture Overview Experimental Office-First Speaking Pa","default_branch":null,"files":null,"tree":[],"storefront":"/r/munch2u-a11y","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/munch2u-a11y/Helix-AGI/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}