算法艺术创作
algorithmic-art
通过代码和可调参数创作生成艺术与交互式视觉效果。
core-runtime【设计与创意】Maintain, debug, and design low-level autogen_core applications: runtimes, routed agents, handlers, topics, subscriptions, tools, model messages, component config, cancellation, intervention, and distributed runtime boundaries.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Use this sub-skill when working on existing AutoGen Python autogen_core applications that need low-level event-driven runtime control. AutoGen is in maintenance mode; for new greenfield agent systems, consider Microsoft Agent Framework, but use this skill for AutoGen maintenance, debugging, migration, and compatibility work.
SingleThreadedAgentRuntime lifecycle, send_message, publish_message, stop, stop_when_idle, and runtime state.RoutedAgent, message_handler, event, rpc, handler type annotations, match predicates, and unhandled messages.DefaultTopicId, TopicId, AgentId, default_subscription, type_subscription, TypeSubscription, and topic-to-agent mapping.FunctionTool, StaticWorkbench, model messages, chat completion contexts, component serialization, cancellation, and intervention handlers.../agentchat-workflows/ for AssistantAgent, teams, group chat, high-level tools-as-agents, streaming chat tasks, termination, and AgentChat state.../extensions-integrations/ for concrete model clients, autogen_ext gRPC runtime packages, Docker/Jupyter executors, MCP, external services, and provider credentials.../tools-studio-bench/ for AutoGen Studio, AG Bench, Magentic-One CLI, and pyautogen package/version boundary work.send_message to an AgentId) or pub/sub (publish_message to a TopicId).stop() for direct sends or stop_when_idle() after publishes.message_handler(strict=False) only when intentionally tolerating type or return mismatches.FunctionTool config as trusted-code operations; validate provider strings, import namespaces, and schemas before loading.references/api-reference.md summarizes key low-level APIs and constructor signatures.references/runtime-patterns.md gives concrete runtime, routing, topics, tools, contexts, cancellation, and intervention patterns.references/distributed-runtime.md explains distributed/gRPC boundaries without depending on extension packages.references/troubleshooting.md maps common runtime failures to diagnosis and fixes.scripts/core_runtime_smoke.py inspects installed signatures and can run a tiny local no-network runtime example.Run the bundled smoke script in an environment with autogen-core installed:
python scripts/core_runtime_smoke.py --inspect
python scripts/core_runtime_smoke.py --run-local-example
The local example uses only SingleThreadedAgentRuntime, a dataclass message, and an in-process RoutedAgent; it performs no network or service calls.