[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-92921":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":8,"language":10,"languages":8,"totalLinesOfCode":8,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":14,"stars7d":14,"stars30d":15,"stars90d":14,"forks30d":14,"starsTrendScore":14,"compositeScore":16,"rankGlobal":8,"rankLanguage":8,"license":8,"archived":17,"fork":17,"defaultBranch":18,"hasWiki":19,"hasPages":19,"topics":20,"createdAt":8,"pushedAt":8,"updatedAt":21,"readmeContent":22,"aiSummary":23,"trendingCount":14,"starSnapshotCount":14,"syncStatus":13,"lastSyncTime":24,"discoverSource":25},92921,"Pekingman","osama-fawad\u002FPekingman","osama-fawad",null,"https:\u002F\u002Fosama-fawad.github.io\u002FPekingman\u002F","HTML",275,7,2,0,146,46.71,false,"main",true,[],"2026-07-22 04:02:07","\u003Ch1 align=\"center\">Pekingman v1.0.0: A Multimodal AGI-Oriented Agent System\u003C\u002Fh1>\n\n\u003Cp align=\"center\">\u003Csmall>\nTingYun Yin, Massachusetts Institute of Technology (MIT)    Peixu Cai, University of Southern California (USC)    Shahzaib Saqib Warraich, University of Southern California (USC)    Osama Fawad, Université Bourgogne Europe (UBE)\n\u003C\u002Fsmall>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fosama-fawad.github.io\u002FPekingman\u002F\">Project Page\u003C\u002Fa> |\n  \u003Ca href=\"https:\u002F\u002Fosama-fawad.github.io\u002FPekingman\u002F\">Demo Video\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Cvideo src=\"github-page\u002Fmedia\u002Fpekingman-demo.mp4\" controls poster=\"github-page\u002Fmedia\u002Fposter.jpg\">\u003C\u002Fvideo>\n\u003C\u002Fp>\n\nPekingman v1.0.0 is a multimodal AGI-oriented agent system designed for environmental perception, long-term memory, human-like reasoning, emotional consistency, and real-time behavioral response. Equipped with a high-capacity persistent memory system, it can store, retrieve, and revisit historical events, giving NPCs authentic human-like behavioral logic.\n\n## Abstract\n\nPekingman connects perception, memory, reasoning, emotion, and action into one continuous agent loop. The system is designed for virtual characters that need to observe changing environments, remember past interactions, make context-aware decisions, and respond with consistent behavior.\n\nThe current implementation focuses on:\n\n- ingesting multi-source environmental signals;\n- extracting emotional and semantic context;\n- storing and revisiting historical interactions;\n- selecting behavior from live context;\n- producing responsive actions for interactive NPC scenarios.\n\n## Human-Equivalent Capabilities\n\nPekingman is designed to replicate a broad set of core cognitive and behavioral capabilities associated with human-like agents. It continuously senses surrounding environmental changes, generates context-matched behavioral feedback to external stimuli, and maintains long-term memory of past experiences. It also supports humanized logical reasoning, sequential decision-making, and consistent emotional tendency expression, enabling NPCs to exhibit coherent, lifelike behaviors across long-running interactions.\n\nThe core capability set includes:\n\n1. **Environmental sensing:** continuously observe multi-dimensional scene changes through multimodal input.\n2. **Real-time response logic:** generate natural feedback and action plans matched to current external conditions.\n3. **Long-term memory iteration:** persist, retrieve, and reference historical interactions to maintain sustained character consistency.\n4. **Humanized decision and emotion output:** simulate human-style reasoning, preference shifts, task choices, and emotional expression.\n\nThis makes Pekingman useful for virtual NPCs, digital human projects, metaverse platforms, simulation games, and other commercial interactive content where characters need continuity, responsiveness, and believable behavior.\n\nLightweight games with low world complexity can often rely on basic trigger logic for NPC behavior. In contrast, large open-world titles and high-complexity simulation environments require agents with multimodal perception, real-time response, and long-term memory to produce natural and consistent virtual human behavior at scale.\n\n## News\n- **August 2026:** Release full code.\n- **July 2026:** Added a GitHub Pages demo page with a compressed web copy of the project video.\n- **June 2026:** Expanded interactive character control, environment awareness, and live visualization.\n- **May 2026:** Added behavior selection, memory continuity, identity modeling, and end-to-end validation.\n\n## Overview\n\nThe project has four main capability areas:\n\n- **Perception:** understands environmental changes from multiple signal types.\n- **Memory:** stores, retrieves, and references long-running interaction history.\n- **Reasoning:** chooses context-aware behavior based on goals, state, and prior experience.\n- **Action:** produces responses suitable for interactive virtual characters and NPCs.\n\n## Hardware Requirements\n\nMinimum requirements depend on which capabilities you run:\n\n- CPU-only hardware is enough for lightweight demos and documentation preview.\n- A modern GPU is recommended for advanced perception, generation, and large-scale interactive scenarios.\n- Real-time 3D simulation experiments require a workstation capable of running the target simulation environment.\n\n## Software Requirements\n\n- Python 3.11\n- Conda or another Python environment manager\n- Optional GPU acceleration for advanced demos\n- Optional 3D engine support for simulation experiments\n\n## Cloning the Repository\n\n```bash\ngit clone \u003Cyour-pekingman-repository-url>\ncd pekingman_project\n```\n\nThis repository contains local data and experimental assets. Before publishing publicly, review ignored files and avoid committing secrets, generated caches, private documents, or local runtime data.\n\n## Environment Setup\n\nCreate the Python environment with the provided environment file:\n\n```bash\nconda env create -f environment.yml\nconda activate pekingman\n```\n\nInstall lightweight web preview requirements when needed:\n\n```bash\npip install -r requirements-web.txt\n```\n\nCopy and edit environment variables:\n\n```bash\ncp ..\u002Fapi-keys.env.example ..\u002Fapi-keys.env\n```\n\n`api-keys.env` is ignored by Git and should contain local database credentials and API keys.\n\n## Optional Services\n\nAdvanced deployments can connect persistent storage, model services, speech processing, and simulation runtime components. These services are optional for reading the documentation or viewing the demo page.\n\n## Running the Demo Page\n\nThe static project page lives in `github-page\u002F`:\n\n```bash\ncd github-page\npython3 -m http.server 8080\n```\n\nOpen:\n\n```text\nhttp:\u002F\u002F127.0.0.1:8080\u002F\n```\n\nThe page uses `github-page\u002Fmedia\u002Fpekingman-demo.mp4`, a web-optimized copy of `demo_video\u002F北京猿人.mp4`.\n\n## Tests\n\nRun local checks from the development environment:\n\n```bash\npython -m pytest\n```\n\nSome checks require optional services, local credentials, or simulation inputs. If a check fails because a service is missing, start the corresponding prerequisite or run a narrower smoke test.\n\n## Functionalities\n\n- **Observe:** collect multimodal environmental signals.\n- **Remember:** preserve working memory, long-term memory, and entity-linked context.\n- **Understand:** summarize current state from perception, memory, and emotional context.\n- **Plan:** select tasks from goals, context, health, and environment conditions.\n- **Act:** produce executable behavior plans for virtual characters.\n- **Monitor:** inspect live agent activity through the local visualization page.\n\n## Known Limitations\n\n- Several advanced capabilities are experimental and may fall back when optional services are unavailable.\n- Public release should exclude local credentials, generated outputs, private documents, and local runtime data.\n- Optional perception and generation dependencies may require platform-specific installation steps.\n\n## Citation\n\nIf you use this project as a research artifact, please cite the repository:\n\n```bibtex\n@software{pekingman2026,\n  title  = {Pekingman v1.0.0: A Multimodal AGI-Oriented Agent System},\n  year   = {2026}\n}\n```\n\nAdd the repository URL to the BibTeX entry after the project is published.\n\n## Acknowledgments\n\nThis project integrates ideas and tools from embodied agents, memory systems, multimodal AI, interactive simulation, and open-source Python infrastructure.\n","Pekingman 是一个面向通用人工智能（AGI）的多模态智能体系统，专注于构建具备环境感知、长时记忆、类人推理与情感一致性的虚拟角色。其核心功能包括多源环境信号输入、语义与情感上下文提取、高容量持久化记忆存储与检索、基于实时上下文的行为决策与响应生成。技术上采用端到端闭环架构，整合感知—记忆—推理—情感—动作模块，支持NPC在长时间交互中保持行为连贯性与人格一致性。适用于虚拟数字人、元宇宙社交场景、仿真训练系统及强调角色真实感的互动游戏开发。","2026-07-11 02:30:19","CREATED_QUERY"]