展示HN：Neuron – 认知多智能体推理架构
目前大多数编排框架仍然像脆弱的链条一样运作——在面对矛盾、长期记忆或动态路由时会崩溃。
Neuron 是一种认知多智能体架构，它以电路而非链条的方式进行思考。多个智能体并行协作，实时调整其路径，并在长时间的交互中保持持久的上下文。
关键组件：
- 智能体：输入、推理、响应、记忆
- 电路：动态路由而非线性链条
- 记忆：情节性 + 上下文持久性
- 监控：完整的推理追踪以便观察
为什么这很重要：
- 处理矛盾输入而不崩溃
- 在长时间会话中保持状态
- 为复杂推理任务提供并行协调
- 透明的日志以便调试和建立信任
GitHub 仓库： [https://github.com/ShaliniAnandaPhD/Neuron](https://github.com/ShaliniAnandaPhD/Neuron)
评估笔记本： [https://www.notion.so/shalini-ananda-phd/Neuron-Evaluation-Notebook-1cec18ea2aa18002b7acf9c1791ca8ea](https://www.notion.so/shalini-ananda-phd/Neuron-Evaluation-Notebook-1cec18ea2aa18002b7acf9c1791ca8ea)
教程系列： [https://www.notion.so/shalini-ananda-phd/Neuron-Framework-Tutorial-Series-Cognitive-Architecture-for-Modular-AI-1fec18ea2aa180b5b1dff554f651bb01](https://www.notion.so/shalini-ananda-phd/Neuron-Framework-Tutorial-Series-Cognitive-Architecture-for-Modular-AI-1fec18ea2aa180b5b1dff554f651bb01)
关于我/背景： [https://www.notion.so/shalini-ananda-phd/Shalini-Ananda-PhD-ad2d228146624c9ea6f5db919bfe5433](https://www.notion.so/shalini-ananda-phd/Shalini-Ananda-PhD-ad2d228146624c9ea6f5db919bfe5433)
希望能收到 HN 社区的反馈——特别是如果您在使用传统工具时遇到过相同的崩溃点。
查看原文
Most orchestration frameworks today still behave like fragile chains — they break when faced with contradictions, long-term memory, or dynamic routing.<p>Neuron is a cognitive multi-agent architecture that thinks in circuits instead of chains. Multiple agents collaborate in parallel, adapt their pathways in real time, and keep persistent context across extended interactions.<p>Key components<p>Agents: Intake, Reasoning, Response, Memory<p>Circuits: Dynamic routing instead of linear chaining<p>Memory: Episodic + contextual persistence<p>Monitoring: Full reasoning traces for observability<p>Why it matters<p>Handles contradictory inputs without collapsing<p>Maintains state across extended sessions<p>Parallel coordination for complex reasoning tasks<p>Transparent logs for debugging & trust<p>GitHub repo: <a href="https://github.com/ShaliniAnandaPhD/Neuron" rel="nofollow">https://github.com/ShaliniAnandaPhD/Neuron</a><p>Evaluation Notebook: <a href="https://www.notion.so/shalini-ananda-phd/Neuron-Evaluation-Notebook-1cec18ea2aa18002b7acf9c1791ca8ea" rel="nofollow">https://www.notion.so/shalini-ananda-phd/Neuron-Evaluation-N...</a><p>Tutorial Series: <a href="https://www.notion.so/shalini-ananda-phd/Neuron-Framework-Tutorial-Series-Cognitive-Architecture-for-Modular-AI-1fec18ea2aa180b5b1dff554f651bb01" rel="nofollow">https://www.notion.so/shalini-ananda-phd/Neuron-Framework-Tu...</a><p>About me / context: <a href="https://www.notion.so/shalini-ananda-phd/Shalini-Ananda-PhD-ad2d228146624c9ea6f5db919bfe5433" rel="nofollow">https://www.notion.so/shalini-ananda-phd/Shalini-Ananda-PhD-...</a><p>Would love feedback from the HN community — especially if you’ve run into the same breakdown points with traditional tools.