{"id":27698124,"url":"https://github.com/rahuletto/agent-orchestrator","last_synced_at":"2025-07-24T07:34:21.956Z","repository":{"id":289528399,"uuid":"971164862","full_name":"Rahuletto/agent-orchestrator","owner":"Rahuletto","description":"A sophisticated recursive architecture that spawns and manages LLM-driven agents via a master-delegator model in a form of layers","archived":false,"fork":false,"pushed_at":"2025-07-09T13:25:54.000Z","size":3464,"stargazers_count":2,"open_issues_count":0,"forks_count":1,"subscribers_count":2,"default_branch":"main","last_synced_at":"2025-07-19T06:52:40.693Z","etag":null,"topics":["agents","ai","artificial-intelligence","delegation","layers","networks"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"gpl-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/Rahuletto.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2025-04-23T05:43:05.000Z","updated_at":"2025-05-16T01:34:49.000Z","dependencies_parsed_at":"2025-04-25T16:38:48.551Z","dependency_job_id":"3bf1758f-8b6b-4421-b709-490e5b37e049","html_url":"https://github.com/Rahuletto/agent-orchestrator","commit_stats":null,"previous_names":["rahuletto/rao"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/Rahuletto/agent-orchestrator","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Rahuletto%2Fagent-orchestrator","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Rahuletto%2Fagent-orchestrator/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Rahuletto%2Fagent-orchestrator/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Rahuletto%2Fagent-orchestrator/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/Rahuletto","download_url":"https://codeload.github.com/Rahuletto/agent-orchestrator/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Rahuletto%2Fagent-orchestrator/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":266808556,"owners_count":23987450,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","status":"online","status_checked_at":"2025-07-24T02:00:09.469Z","response_time":99,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agents","ai","artificial-intelligence","delegation","layers","networks"],"created_at":"2025-04-25T16:38:27.113Z","updated_at":"2025-07-24T07:34:21.938Z","avatar_url":"https://github.com/Rahuletto.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Autonomous Multi-agent Orchestration\n\n\u003e [!IMPORTANT]\n\u003e This branch consist of new proposal of architecture, Instead of parallel agents, I went with Layers of Agents approach. v1 will be available in [old branch](https://github.com/Rahuletto/agent-orchestrator/tree/v1).\n\nA sophisticated recursive architecture that spawns and manages LLM-driven agents via a master-delegator model in a form of layers (just like a neural layers), with task-specific prompt tuning, execution constraints, and multi-agent recursion without human intervention.\n\n\u003e This uses a concept where a primary agent creates specialized sub-agents on demand. Each agent is dynamically generated with task-specific prompting and execution parameters, enabling efficient problem-solving without human intervention while not having a bias and having opinions from various agents instead of deciding on its own. Uses Ensemble technique.\n\n# Architecture\n\n## Here's how it works\n\n![architecture diagram](/assets/architecture_layers.png)\n\nThe system follows a recursive workflow:\n\n- `Master Agent` analyzes the input task and generates a comprehensive task breakdown with dependency structure.\n- For each subtask, the `Master Agent` creates a specialized `Child Agent` with custom fine-tuned prompts with dependencies.\n- ~`Child Agents` work on their assigned tasks and can create their own `Sub-Child Agents` when further specialization is needed~\n- `Child Agents` are rearranged based on the dependency graph and co-ordinates with other agents while sharing context\n- ~Results flow back up the hierarchy for integration and final output~\n- Results flow through layers of agents with context and intreprets a final output.\n\nThe key innovation is that agent creation and prompt engineering happen automatically at runtime in _layers_, with no predefined agent structures or human-designed prompts.\n\n## Caveats\n\nAs you see, this runs on recrusive layers method. building agents upon agents as it needs. This can result in building an agent layer network as shown below\n![agent tree](/assets/layers.png)\n\nThe recursive nature of this system creates potential challenges:\n\n- **Resource Management:** Each additional agent consumes computational resources\n- **Runtime Concerns:** Deep agent layers can significantly impact completion time as it becomes sequential after a time\n\nWhile the system offers powerful flexibility by giving control to the `Master Agent`, careful monitoring is recommended for resource-intensive tasks.\n\n## Demo\n\nhttps://github.com/user-attachments/assets/ec99d37f-c6c9-4fbd-a07e-1997d88aaa99\n\n# Citations and Resources\n\n1. Emergence AI's 2025 Orchestrator\n   Automatically creates agents and assembles multi-agent systems with minimal human intervention... continuously refining tools through recursive self-improvement\n\n2. ReDel's Recursive Systems (2024)\n   Introduces systems where a root agent \"decomposes tasks into subtasks then delegates to sub-agents\" rather than using human-defined agent graphs.\n\n3. Beyond Better's Orchestrator (2025)\n   Implements \"sub-agents created with specialized capabilities for token efficiency and parallel processing\" through dynamic task analysis.\n\n# License\n\nThis project is licensed under the GNU General Public License (GPL).\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Frahuletto%2Fagent-orchestrator","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Frahuletto%2Fagent-orchestrator","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Frahuletto%2Fagent-orchestrator/lists"}