{"id":49700040,"url":"https://github.com/w2jmoe/wisepick","last_synced_at":"2026-05-31T05:00:35.973Z","repository":{"id":355813053,"uuid":"1229722312","full_name":"w2jmoe/WisePick","owner":"w2jmoe","description":"Decision Infrastructure for the AI Era. 🧠✨️🤺","archived":false,"fork":false,"pushed_at":"2026-05-30T08:31:26.000Z","size":320,"stargazers_count":14,"open_issues_count":1,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-05-30T10:12:35.751Z","etag":null,"topics":["adapter-pattern","agent-infrastructure","agent-orchestration","agent-runtime","capability-routing","decision-routing","deterministic-orchestration","deterministic-routing","execution-routing","llm-ops","low-latency","mcp","multi-agent","routing-middleware","runtime-intelligence"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/w2jmoe.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,"notice":null,"maintainers":null,"copyright":null,"agents":"AGENTS.md","dco":null,"cla":null}},"created_at":"2026-05-05T10:19:16.000Z","updated_at":"2026-05-30T08:31:30.000Z","dependencies_parsed_at":"2026-05-15T12:01:54.353Z","dependency_job_id":"c06c45a6-7499-4c13-a709-1b53f95716c3","html_url":"https://github.com/w2jmoe/WisePick","commit_stats":null,"previous_names":["w2jmoe/wisepick"],"tags_count":12,"template":false,"template_full_name":null,"purl":"pkg:github/w2jmoe/WisePick","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/w2jmoe%2FWisePick","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/w2jmoe%2FWisePick/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/w2jmoe%2FWisePick/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/w2jmoe%2FWisePick/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/w2jmoe","download_url":"https://codeload.github.com/w2jmoe/WisePick/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/w2jmoe%2FWisePick/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":33719601,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-05-31T02:00:06.040Z","response_time":95,"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":["adapter-pattern","agent-infrastructure","agent-orchestration","agent-runtime","capability-routing","decision-routing","deterministic-orchestration","deterministic-routing","execution-routing","llm-ops","low-latency","mcp","multi-agent","routing-middleware","runtime-intelligence"],"created_at":"2026-05-08T07:05:38.396Z","updated_at":"2026-05-31T05:00:35.910Z","avatar_url":"https://github.com/w2jmoe.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003cdiv align=\"center\"\u003e\n\n\u003cimg src=\"./logo.png\" height=\"120\" alt=\"WisePick logo\" /\u003e\n\n# WisePick | 智选\n\n\u003e **Docs:** [Overview](./README.md) | [Integration \u0026 SDK](./README_API.md) | [Agent Protocol](./AGENTS.md)\n\n**The deterministic layer that helps agent runtimes make better execution decisions.**\n\n**帮助 Agent Runtime 做出更优执行决策的确定性层。**\n\n[![GitHub stars](https://img.shields.io/github/stars/w2jmoe/WisePick?style=flat-square)](https://github.com/w2jmoe/WisePick/stargazers)\n[![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg?style=flat-square)](./LICENSE)\n[![Follow on X](https://img.shields.io/badge/Follow-%40w2jmoe-000000?style=flat-square\\\u0026logo=x)](https://twitter.com/w2jmoe)\n\n\u003cimg width=\"100%\" alt=\"WisePick contrast flowchart\" src=\"https://github.com/user-attachments/assets/32f44428-f6d4-4b3a-aef7-f85c10dc6090\" /\u003e\n\n\n\u003c/div\u003e\n\n---\n\n## 🚀 WisePick Decision API (WPDA)\n\n* Deterministic capability selection for AI agents.\n* 为 AI Agent 提供确定性能力选择。\n\n### What it is | 智选是什么\n\nWisePick is a small decision layer for agent runtimes. It takes an intent, picks the best capability, and improves from execution feedback.\n\n智选是一个面向 Agent Runtime 的轻量决策层：输入意图，选择最合适的能力，并从执行反馈中持续优化。\n\n### Who is it for | 适用于谁\n\n* Agent Operators\n* Agent Runtime Builders\n* Internal Agent Teams\n* Agent SaaS Platforms\n* Agent Hosting Providers\n\nFor teams running real agent workflows.\n\n面向运行真实 Agent Workflow 的团队。\n\n### Why it matters | 为什么重要\n\nWisePick cuts tool searching, retry loops, latency, and cost by reusing execution feedback.\n\n智选通过复用执行反馈，减少工具搜索、重试循环、延迟和成本。\n\n## 🌐 Ecosystem Alignment | 生态兼容\n\n[![YantrikDB](https://img.shields.io/badge/YantrikDB-000000?style=flat-square)](https://github.com/yantrikos/yantrikdb-server)\n[![Langfuse-MCP](https://img.shields.io/badge/Langfuse--MCP-000000?style=flat-square)](https://github.com/avivsinai/langfuse-mcp)\n[![ChainWeaver](https://img.shields.io/badge/ChainWeaver-000000?style=flat-square)](https://github.com/dgenio/ChainWeaver)\n[![Aetheris](https://img.shields.io/badge/Aetheris-000000?style=flat-square)](https://github.com/Colin4k1024/Aetheris)\n[![SafeAgent](https://img.shields.io/badge/SafeAgent-000000?style=flat-square)](https://github.com/azender1/SafeAgent)\n[![open-thymos](https://img.shields.io/badge/open--thymos-000000?style=flat-square)](https://github.com/gryszzz/open-thymos)\n\n---\n\n## 📘 Contents | 目录\n\n| Core Sections                                                                | 核心章节      |\n| :--------------------------------------------------------------------------- | :-------- |\n| [🚀 Quick Start](#-quick-start--快速启动)                                        | 快速启动      |\n| [⚡ Why Integrate WisePick](#-why-integrate-wisepick--为什么接入智选)                | 为什么接入智选   |\n| [📜 Performance Benchmarks](#-performance--cost-benchmarks--性能与成本报告)         | 性能与成本报告   |\n| [🔌 Integration Specification](#-integration--router-specification--集成与路由规范) | 集成与路由规范   |\n| [🧠 How It Works](#-how-it-works--工作原理)                                      | 工作原理      |\n| [🏗️ Architectural Paradigm](#%EF%B8%8F-architectural-paradigm--架构范式演进)      | 架构范式演进    |\n| [🧪 Agent Workflow](#-agent-workflow--agent-工作流)                             | Agent 工作流 |\n| [🗺️ Roadmap](#%EF%B8%8F-roadmap--路线图)                                       | 路线图       |\n\n---\n\n## 🚀 Quick Start | 快速启动\n\n\u003e WisePick Decision API (WPDA) is routing infrastructure for agent runtimes.\n\u003e Deploy it as an independent API service; integrate it through the SDK.\n\u003e\n\u003e 智选 Decision API（WPDA）是面向 Agent Runtime 的路由基础设施。\n\u003e 请将其作为独立 API 服务部署，并通过 SDK 接入。\n\n### 1. Start Server | 启动服务\n\n```bash\ngit clone https://github.com/w2jmoe/WisePick.git\ncd WisePick\n\ncp .env.example .env\n# Configure DATABASE_URL\n\npip install -r requirements.txt\n\nuvicorn app.main:app --reload --host 0.0.0.0 --port 8000\n```\n\n### 2. Integrate SDK | 接入 SDK\n\nSee the [15-Minute Integration Guide](./README_API.md%2315-minute-integration-checklist).\n\n查看 [15分钟接入指南](./README_API.md%2315-minute-integration-checklist)。\n\n**Help us refine:** If you find the docs confusing, please [open an issue](https://github.com/w2jmoe/WisePick/issues) — we are aggressively refining our integration flow.\n**反馈建议：** 若您在接入中遇到任何困惑，请 [提交 Issue](https://github.com/w2jmoe/WisePick/issues)，我们正在全力优化文档与接入体验。\n\n---\n\n## ⚡ Why Integrate WisePick | 为什么接入智选\n\n* Lower cost and latency by cutting trial-and-error.\n  更低成本与延迟：减少无效试错与 Token 浪费。\n* One best capability per decision.\n  每次决策只选一个最合适的能力。\n* Feedback makes the next decision better.\n  执行反馈让下一次选择更好。\n\n---\n\n## 📜 Performance \u0026 Cost Benchmarks | 性能与成本报告\n\nProduction-oriented deterministic routing vs. native LLM tool-calling. Tested inside a Hermes-style agent runtime.\n面向生产的确定性路由与原生 LLM 工具调用的对比测试。已在 Hermes 类 Agent 运行时中完成验证。\n\n### Runtime Efficiency | 运行时效率提升\n\n| Metrics               | Native LLM | WisePick         | Optimization                     |\n| --------------------- | ---------- | ---------------- | -------------------------------- |\n| **🚀 Path Speed**     | Baseline   | **~31% Faster**  | Shrunk from 6.33 to 4.33 steps   |\n| **⏱️ Time Saved**     | Baseline   | **~62% Saved**   | Benchmark cut from 12m to 4m30s  |\n| **💵 Cost Cut**       | High       | **~33% Reduced** | $0.15 → $0.10 per session        |\n| **🎯 First-Hit Rate** | Exp.       | **100% Locked**  | Zero hallucinated tool-selection |\n\n### Key Capabilities | 核心优势\n\n* **Zero-Latency Gatekeeping**: Sub-millisecond average latency under isolated routing-core stress testing.\n  （隔离路由核心压测下平均亚毫秒级延迟）\n* **Anti-Loop Depth**: Stabilizes execution across 20+ mixed-tool tasks without infinite loops.\n  （在 20+ 混合工具任务下稳定运行，彻底消除无限循环路径）\n\nBenchmark scripts \u0026 instrumentation: [BENCHMARK](./benchmark/) | [STRESS_TEST_RESULTS.md](./docs/STRESS_TEST_RESULTS.md)\n\n---\n\n## 🔌 Integration \u0026 Router Specification | 集成与路由规范\n\nWisePick is stateless. You own execution; we provide routing.\n智选是无状态决策层：由你负责执行，我们负责路由。\n\n* **For Human Builders / 面向人类开发者 ([README_API.md](./README_API.md)):** SDK integration, programmatic turn interception, multi-turn lock release, and execution hook feedback closure.\n  SDK 代码集成、首轮意图拦截、多轮强制解锁以及执行钩子反馈闭环。\n* **For AI Agents \u0026 Automation / 面向智能体与自动配置 ([AGENTS.md](./AGENTS.md)):** Machine-readable `wisepick.agent.v1` manifest contract, protocol state machine mapping, and runtime environment declaration.\n  机器可读的声明式配置清单契约、协议状态机映射以及运行时环境依赖声明。\n\n---\n\n## 🧠 How It Works | 工作原理\n\n### Capability Matching | 能力匹配\n\nTask text → capability labels derived from bootstrap rules.\n任务文本 → 由引导规则得到能力标签。\n\n### Capability Scoring | 能力评分\n\n```text\nscore =\ncapability_match       * 0.40  (语义匹配度 - 核心逻辑)\nexecution_success_rate * 0.20  (历史可靠性)\nefficiency_factor      * 0.20  (执行效率 - 基于 avg_latency_ms)\neconomy_factor         * 0.10  (成本性价比 - 基于 avg_token_cost)\nbootstrap_weight       * 0.10  (初始冷启动权重)\n```\n\n### Feedback Loop | 反馈闭环\n\n```text\ndecision → execution → feedback → capability_stats → next decision\n\nRouting updates from real execution outcomes.\n路由统计随真实执行结果更新。\n```\n\n### Components | 核心组件\n\n```text\nRouting Core (decision_engine)\nConverts incoming tasks into ECU scores and selects the best capability.\n将输入任务转换为 ECU（执行单元）评分并进行路由决策。\n\nCapability Registry (api_tool_specs)\nMaintains available providers, capability labels, and bootstrap weights.\n管理可用 Provider、能力标签及冷启动权重分配。\n\nExecution Memory (tool_stats, feedback)\nStores execution feedback reported by your runtime and continuously improves routing quality.\n存储由运行时上报的执行反馈，并持续优化路由质量。\n```\n\n---\n\n## 🏗️ Architectural Paradigm | 架构范式演进\n\nWisePick unifies both hard-coded and dynamic tool discovery under a deterministic routing layer.\n\n智选将硬编码与动态发现路线统一于确定性路由层。\n\n| Paradigm    | Discovery      | Runtime Pain                                               | WisePick Value                                      |\n| ----------- | -------------- | ---------------------------------------------------------- | --------------------------------------------------- |\n| **Static**  | Manual config  | Brittle scaling, zero runtime flexibility                  | Centralized ECU registry, cleaner code              |\n| **Dynamic** | Auto-discovery | **Tool Anxiety:** Context explosion, loops, hallucinations | **Deterministic Filter:** Cuts 95% noise, 100% lock |\n\n## 🔬 ECU Response (with ROI Metrics) | 带有 ROI 指标的 ECU 响应\n\n\u003e **ECU (Executable Capability Unit)**\n\u003e\n\u003e A standardized executable capability an agent can route, invoke, and learn from.\n\u003e\n\u003e 可执行能力单元：可被路由、调用并通过反馈学习的标准化能力抽象。\n\n```json\n{\n  \"metadata\": {\n    \"schema_version\": \"mcp.route_decision.v1\",\n    \"decision_id\": \"dec_abc123def4567890\",\n    \"trace_id\": \"trace_9876543210abcdef\",\n    \"router_name\": \"wisepick\",\n    \"capability_id\": \"audio_transcription\",\n    \"provider\": \"feishu_minutes\",\n    \"execution_type\": \"api\",\n    \"callable\": true,\n    \"confidence\": 0.87,\n    \"latency_ms\": 450,\n    \"candidate_count\": 1,\n    \"top_candidates\": [\n      {\n        \"rank\": 1,\n        \"capability_id\": \"audio_transcription\",\n        \"score\": 0.87,\n        \"selected\": true\n      }\n    ],\n    \"reason_codes\": [\"capability_match\"]\n  }\n}\n```\n\nWisePick predicts performance before execution to ensure the best ROI.\n\nWisePick 在执行前预测性能，以确保最佳投资回报率 (ROI)。\n\n---\n\n## 🧪 Agent Workflow | Agent 工作流\n\n```mermaid\ngraph LR\n    A[Ask Routing] --\u003e B(Receive ECU)\n    B --\u003e C{Map to Local Implementation}\n    C --\u003e D[Execute]\n    D --\u003e E[Send Feedback]\n    E -.-\u003e|Updates ROI Models| A\n```\n\nWisePick provides decision intelligence and feedback loops—not task execution.\n\n智选提供决策智能与反馈闭环；**不替代**任务执行本身。\n\n---\n\n## 🔮 Vision | 愿景\n\n**Long-term: Execution Experience Network — collective decision memory across agent runtimes.**\n\n**长期愿景：执行经验网络（Execution Experience Network）— 跨运行时的集体决策记忆。**\n\n**Today:** [WisePick Decision API (WPDA)](#-wisepick-decision-api-wpda) delivers deterministic capability routing and execution feedback via SDK integration; your runtime owns invoke, policy, and retries.\n\n**当前：** [WisePick Decision API（WPDA）](#-wisepick-decision-api-wpda) 通过 SDK 提供确定性能力路由与执行反馈；工具调用、策略与重试由您的运行时负责。\n\nWisePick turns execution outcomes into reusable routing experience — the foundation for shared ECU feedback and feedback-driven execution optimization on the path to that network.\n\n智选将执行结果沉淀为可复用的路由经验，为 ECU 反馈共享与反馈驱动的执行优化奠定基础，并沿路线图演进至上述网络愿景。\n\n---\n\n## 🗺️ Roadmap | 路线图\n\n* **✅ v0.1**: Core Infrastructure · 核心路由与反馈闭环\n\n  * Deterministic ECU routing \u0026 feedback loop.\n  * Multi-dimensional ROI metrics aggregation (Latency / Cost / Quality).\n* **🔄 v0.2**: Runtime-Aware Optimization · Runtime 感知执行优化\n\n  * Task-level capability routing for multi-agent runtimes.\n  * Adaptive execution-path optimization based on latency / cost / quality feedback.\n  * Lightweight integration adapters for orchestration frameworks.\n* **🔄 v0.3**: Collective Decision Memory · 集体决策记忆\n\n  * Cross-agent experience sharing (Execution Experience Network direction).\n  * Global capability indexing \u0026 optimization.\n\n---\n\n## 🤗 Feedback \u0026 Integration | 反馈与集成\n\n* **Issues:** [GitHub Issues](https://github.com/w2jmoe/WisePick/issues)\n* **Email:** w2jmoe@gmail.com\n\n**Every routing decision feeds execution feedback today — and collective decision memory as the platform evolves toward the Execution Experience Network vision.**\n\n**每一次路由决策都会形成执行反馈；随平台演进，亦将沉淀为集体决策记忆，迈向执行经验网络愿景。˗ˋˏ( ´͈ ᗜ `͈ )ˎˊ˗**\n\n---\n\n## 🌸 License | 许可协议\n\nApache License 2.0 — see [LICENSE](./LICENSE).\n\n\u003cimg style=\"max-width: 100%; width: 500px;\" alt=\"WisePick✨️\" src=\"https://github.com/user-attachments/assets/fb3af8b7-eb36-425d-a266-0cced12cb592\" /\u003e\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fw2jmoe%2Fwisepick","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fw2jmoe%2Fwisepick","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fw2jmoe%2Fwisepick/lists"}