A local multi-agent system running on self-hosted LLMs (Ollama). 3 channels (Telegram/console/MAX), 4 model types (LLM/embed/vision/voice), multimodality (text/PDF/photos/voice), and Planner/Executor/Critic roles with reflection. Tools: email, Yandex Disk, web, OCR, cron scheduler. Semantic mem
16 files for Claude Code, Codex, GitHub Copilot, OpenCode and Gemini CLI: architecture-discipline, async-discipline, automation-discipline, debugging-discipline and 12 more — 2,319 tokens loaded in every session.
.github/copilot-instructions.md A 4 tok AGENTS.md A 2,309 tok CLAUDE.md A 3 tok GEMINI.md A 3 tok .agents/skills/architecture-discipline/SKILL.md A 54 tok .agents/skills/async-discipline/SKILL.md A 38 tok .agents/skills/automation-discipline/SKILL.md A 47 tok .agents/skills/debugging-discipline/SKILL.md A 47 tok .agents/skills/documentation-discipline/SKILL.md A 45 tok .agents/skills/error-handling-discipline/SKILL.md A 47 tok .agents/skills/git-discipline/SKILL.md A 50 tok .agents/skills/pr-review-discipline/SKILL.md A 48 tok .agents/skills/prompt-injection-defense/SKILL.md B 47 tok .agents/skills/pull-request-discipline/SKILL.md A 44 tok .agents/skills/skill-authoring/SKILL.md A 52 tok .agents/skills/testing-discipline/SKILL.md A 37 tok These files are radif-ru/ai-multi-agent-system's own configuration — they tell Claude Code, Codex, GitHub Copilot, OpenCode and Gemini CLI how to work on this repository, so they are not mods to install elsewhere. Copy one as a starting point and replace the parts that are about this project.