[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-92702":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":15,"stars7d":15,"stars30d":16,"stars90d":15,"forks30d":15,"starsTrendScore":15,"compositeScore":17,"rankGlobal":10,"rankLanguage":10,"license":18,"archived":19,"fork":19,"defaultBranch":20,"hasWiki":21,"hasPages":19,"topics":22,"createdAt":10,"pushedAt":10,"updatedAt":33,"readmeContent":34,"aiSummary":35,"trendingCount":15,"starSnapshotCount":15,"syncStatus":14,"lastSyncTime":36,"discoverSource":37},92702,"EVA-CLIENT","Noietch\u002FEVA-CLIENT","Noietch","EVA-Client: A Unified Framework for Deployment, Evaluation, and Data Collection on Real Robots","https:\u002F\u002Fcolalab.net\u002Fprojects\u002Feva-client\u002F",null,"Python",55,5,2,0,1,42.43,"Apache License 2.0",false,"main",true,[23,24,25,26,27,28,29,30,31,32],"agilex-piper","deployment","embodied-ai","franka","infernece","robotics","teleoperation","ur5e","vla","wam","2026-07-22 04:02:06","\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fcolalab.net\u002Fprojects\u002Feva-client\u002F\">\u003Cimg src=\"assets\u002Feva-logo.svg\" alt=\"EVA-Client Logo\" width=\"34%\">\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Ch1 align=\"center\">EVA-Client: A Unified Framework for Deployment, Evaluation, and Data Collection on Real Robots\u003C\u002Fh1>\n\n\u003Cp align=\"center\">One policy, any robot — the smooth all-in-one real-robot stack. Debug, record, evaluate, visualize, all in the browser.\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n\u003Ca href=\"https:\u002F\u002Fcolalab.net\u002Fprojects\u002Feva-client\u002F\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FProject%20Page-colalab.net-blue?style=for-the-badge&logo=googlechrome&logoColor=white\" alt=\"Project Page\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fcolalab.net\u002Fprojects\u002Feva-client\u002Fpaper\u002FEVA_Client_Report.pdf\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FTechnical%20Report-PDF-red?style=for-the-badge&logo=adobeacrobatreader&logoColor=white\" alt=\"Technical Report\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fcolalab.net\u002Fprojects\u002Feva-client\u002Fdocs\u002Fintroduction.html\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDocs-English-2ea44f?style=for-the-badge&logo=readthedocs&logoColor=white\" alt=\"Documentation (English)\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fcolalab.net\u002Fprojects\u002Feva-client\u002Fdocs\u002Fintroduction.zh.html\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F文档-中文-2ea44f?style=for-the-badge&logo=readthedocs&logoColor=white\" alt=\"Documentation (中文)\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FNoietch\u002FEVA-CLIENT\u002Fstargazers\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNoietch\u002FEVA-CLIENT?style=for-the-badge&logo=github&logoColor=white&color=0a0a0a&cacheSeconds=60\" alt=\"GitHub Stars\">\u003C\u002Fa>\n\u003Cp align=\"center\">\n  \u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F09cf8c98-396d-45d0-bd38-8603412ec3c2\" controls muted>\u003C\u002Fvideo>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\u003Cem>EVA-Client driving an AgileX bimanual arm end-to-end from the browser — teleop → record → π₀ checkpoint → smooth async deploy. Real hardware, not a rendering.\u003C\u002Fem>\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Cb>Jump to:\u003C\u002Fb> \n  \u003Ca href=\"#-what-you-get\">What you get\u003C\u002Fa> ·\n  \u003Ca href=\"#-robot-zoo--compatibility\">Robot zoo\u003C\u002Fa> ·\n  \u003Ca href=\"#-protocols--middleware\">Protocols\u003C\u002Fa> ·\n  \u003Ca href=\"#%EF%B8%8F-architecture\">Architecture\u003C\u002Fa> ·\n  \u003Ca href=\"#-documentation\">Documentation\u003C\u002Fa> ·\n  \u003Ca href=\"#%EF%B8%8F-roadmap\">Roadmap\u003C\u002Fa> ·\n  \u003Ca href=\"#-citation\">Cite\u003C\u002Fa>\n\u003C\u002Fp>\n\n---\n\n## ✨ What you get\n\n* **🚀 Deployment.** One command brings up a real-robot closed loop:\n  `.py` config → transport ([ROS1](https:\u002F\u002Fgithub.com\u002Fros\u002Fros) \u002F [ROS2](https:\u002F\u002Fgithub.com\u002Fros2\u002Fros2) \u002F [ZeroMQ](https:\u002F\u002Fgithub.com\u002Fzeromq\u002Fpyzmq) \u002F offline dataset) → policy\n  backend ([OpenPI](https:\u002F\u002Fgithub.com\u002FPhysical-Intelligence\u002Fopenpi), [OpenPI-RTC](https:\u002F\u002Fwww.pi.website\u002Fresearch\u002Freal_time_chunking), [StarVLA](https:\u002F\u002Fgithub.com\u002FstarVLA\u002FstarVLA), [GR00T](https:\u002F\u002Fgithub.com\u002FNvidia\u002FIsaac-GR00T), mock, replay) → inference\n  strategy (sync \u002F [async](https:\u002F\u002Fgithub.com\u002FOpenDriveLab\u002Fkai0#train-deploy-alignment) \u002F naive \u002F [ACT-ensemble](https:\u002F\u002Fgithub.com\u002Ftonyzhaozh\u002Fact) \u002F RTC) with live latency\n  compensation. **6 robots already** — joint-space or EEF-space (PyRoki IK),\n  all live-switchable from the DEBUG tab.\n* **📊 Evaluation.** Multi-checkpoint sweeps with per-trial records: every\n  rollout captures camera video, 3D URDF scene, per-dimension state charts,\n  and per-prompt milestone scores into dataset metadata, then replays\n  synchronously in the RESULT tab. Prompt shuffling, and remote\n  policy servers over SSH port-forward are first-class.\n* **🎥 Data collection.** Teleop capture straight into [LeRobot](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Flerobot) v2.1 episodes\n  from the COLLECT tab — background saver, in-tab QC PASS\u002FFAIL replay, camera\n  streams encoded to mp4, per-frame green\u002Fred quality flags. Teleop demos and\n  model rollouts share one on-disk layout.\n\n---\n\n## 🔥 What's NEW!\n\n* **[2026-07] EVA-Client is open-sourced!** \n* **[2026-07] Paper, docs, and project page are live!** Read the [Technical Report](https:\u002F\u002Fcolalab.net\u002Fprojects\u002Feva-client\u002Fpaper\u002FEVA_Client_Report.pdf), browse the [Documentation](https:\u002F\u002Fcolalab.net\u002Fprojects\u002Feva-client\u002Fdocs\u002Fintroduction.html) ([中文](https:\u002F\u002Fcolalab.net\u002Fprojects\u002Feva-client\u002Fdocs\u002Fintroduction.zh.html)), and visit the [Project Page](https:\u002F\u002Fcolalab.net\u002Fprojects\u002Feva-client\u002F).\n\n---\n\n## 🤖 Robot zoo & compatibility\n\n✅ Supported  ·  🚧 In development  ·  📦 To be added\n\n| Robot | Form factor | Supported |\n|-------|-------------|:---------:|\n| AgileX Piper | Dual 6-DoF arm + gripper | ✅ |\n| ARX R5 | Dual 6-DoF arm + gripper | ✅ |\n| Dual Franka Panda | Dual 7-DoF arm + gripper | ✅ |\n| Galaxea R1 Lite | Dual 6-DoF arm on torso | ✅ |\n| Universal Robots UR5e | Single 6-DoF arm + gripper | ✅ |\n| AgiBot G2 | Dual-arm humanoid (24-DoF body) | ✅ |\n| AgiBot G2 (mobile base) | Humanoid on mobile chassis | 🚧 |\n| YAM | Dual-arm manipulator | 🚧 |\n| Tianji | Dual-arm manipulator | 🚧 |\n| Unitree H1 \u002F G1 | Humanoid | 🚧 |\n| Fourier GR-1 | Humanoid | 📦 |\n| Booster T1 | Humanoid | 📦 |\n| Mobile ALOHA | Mobile dual-arm | 📦 |\n\n---\n\n## 🔌 Protocols & middleware\n\n| Layer | Name | Protocol \u002F wire format | Supported |\n|-------|------|------------------------|:---------:|\n| Transport | ROS 1 | ROS 1 stack | ✅ |\n| Transport | ROS 2 | ROS 2 stack | ✅ |\n| Transport | ZeroMQ | ZeroMQ execution node | ✅ |\n| Transport | Offline dataset | Offline LeRobot v2.x replay | ✅ |\n| Policy backend | OpenPI | WebSocket + msgpack (OpenPI, stateless) | ✅ |\n| Policy backend | OpenPI-RTC | WebSocket + msgpack (Real-Time Chunking) | ✅ |\n| Policy backend | StarVLA | WebSocket + msgpack (typed envelope) | ✅ |\n| Policy backend | GR00T | ZeroMQ REQ\u002FREP + msgpack-numpy (Isaac-GR00T) | ✅ |\n| Policy backend | Mock | Local (smooth random actions) | ✅ |\n| Policy backend | Replay | Local (recorded trajectory replay) | ✅ |\n\n---\n\n## 🏗️ Architecture\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"assets\u002Fworkflow.png\" width=\"100%\" alt=\"EVA-Client workflow\">\n\u003C\u002Fp>\n\n---\n\n## 📚 Documentation\n\nFull guides live in [`docs\u002F`](.\u002Fdocs). Start here:\n\n| Guide | What's inside |\n|-------|---------------|\n| [📦 Installation](.\u002Fdocs\u002Finstallation.md) | Requirements, `uv` setup, hardware extras, verify |\n| [🚀 Quick start](.\u002Fdocs\u002Fquick-start.md) | Two-process bring-up, supported transports, deploy\u002Feval\u002Freplay presets |\n| [🧭 Web console](.\u002Fdocs\u002Fweb-console.md) | The six tabs — MANUAL, COLLECT, REPLAY, DEBUG, EVAL, RESULT |\n| [📚 Core concepts](.\u002Fdocs\u002Fconcepts.md) | Transports, policy backends, inference strategies, robots, action spaces |\n| [⚙️ Configuration](.\u002Fdocs\u002Fconfiguration.md) | `_base_` inheritance, deep merge, startup pipeline |\n| [🎞️ Recording](.\u002Fdocs\u002Frecording.md) | LeRobot v2.1 on-disk layout, QC flags, eval trials |\n| [🔬 Development](.\u002Fdocs\u002Fdevelopment.md) | Tests, lint, type-check, and how the robot side is faked |\n\n---\n\n## 🗺️ Roadmap\n\n- [ ] **More robots.** Extend the robot zoo to more embodiments — dual-arm\n      manipulators (YAM, Tianji, …), humanoids\n      (Unitree H1\u002FG1, Fourier GR-1, Booster T1, …) and mobile \u002F wheeled\n      platforms (mobile ALOHA, Galaxea R1 base, quadruped + arm).\n- [ ] **Human-in-the-loop data collection for RL.** Interventions during\n      policy rollout captured as preference \u002F correction data, DAgger-style\n      relabeling, and reward-model signals piped back through the LeRobot\n      episode format for online RL fine-tuning.\n- [ ] **Embodied agent.** Wrap the deployment + evaluation loop with a\n      language-driven planner (VLM \u002F VLA + tool use) so long-horizon tasks\n      can be decomposed, executed, verified, and re-planned end-to-end from\n      the same console.\n- [ ] **Data annotation.** Extend Collect mode with fine-grained task and\n      sub-task annotation, segmenting long-horizon episodes into labeled\n      sub-task units and milestones within the same LeRobot dataset. This\n      makes a collection reusable at the level of individual manipulation\n      phases.\n\n---\n\n## 📝 Citation\n\nIf EVA-Client is useful for your research or product, please cite:\n\n```bibtex\n@misc{yang2026evaclient,\n      title={EVA-Client: A Unified Data Collection, Inference, and Deployment Framework for Embodied Policies on Real Robots}, \n      author={Heqing Yang and Yang Yi and Liyao Wang and Linqing Zhong and Donglin Yang and Ruipu Wu and Zitong Bai and Fengjiao Chen and Manyuan Zhang and Linjiang Huang and Si Liu},\n      year={2026},\n      eprint={2607.02646},\n      archivePrefix={arXiv},\n      primaryClass={cs.RO},\n      url={https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.02646}, \n}\n```\n\n## 📬 Contact\n\nFor questions or collaboration, feel free to reach out via WeChat:\n\n\u003Cp align=\"left\">\n  \u003Cimg src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F203d7029-449d-4070-b9c4-e78863625e92\" alt=\"WeChat QR Code\" width=\"240\" \u002F>\n\u003C\u002Fp>\n\n---\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>License & Acknowledgements\u003C\u002Fb>\u003C\u002Fsummary>\n\nThis project is licensed under the Apache-2.0 License. See [LICENSE](.\u002FLICENSE)\nfor more information.\n\nBuilds upon several excellent open-source efforts, including\n[PyRoki](https:\u002F\u002Fgithub.com\u002Fchungmin99\u002Fpyroki) and\n[jaxls](https:\u002F\u002Fgithub.com\u002Fbrentyi\u002Fjaxls) for kinematics,\n[OpenPI](https:\u002F\u002Fgithub.com\u002FPhysical-Intelligence\u002Fopenpi) for policy serving,\nthe [LeRobot](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Flerobot) dataset format, and a\nvendored mmengine-style config system from\n[MMEngine](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmengine).\n\n\u003C\u002Fdetails>\n","EVA-Client 是一个面向具身智能的统一框架，用于在真实机器人上实现策略部署、性能评估与多模态数据采集。它支持 ROS1\u002FROS2、ZeroMQ 等通信协议，兼容 AgileX Piper、Franka、UR5e、WAM 等六类硬件平台，提供浏览器端可视化调试、远程遥操作、异步推理调度及延迟补偿机制，并集成 OpenPI、StarVLA、GR00T 等主流 VLA 模型后端。适用于机器人算法研发、真实场景闭环验证、人机协同数据收集等实验室及工程落地场景。","2026-07-10 02:30:12","CREATED_QUERY"]