[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-72297":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":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":25,"hasPages":25,"topics":26,"createdAt":10,"pushedAt":10,"updatedAt":31,"readmeContent":32,"aiSummary":33,"trendingCount":16,"starSnapshotCount":16,"syncStatus":34,"lastSyncTime":35,"discoverSource":36},72297,"DeepResearchAgent","SkyworkAI\u002FDeepResearchAgent","SkyworkAI","DeepResearchAgent is a hierarchical multi-agent system designed not only for deep research tasks but also for general-purpose task solving. The framework leverages a top-level planning agent to coordinate multiple specialized lower-level agents, enabling automated task decomposition and efficient execution across diverse and complex domains.","https:\u002F\u002Fskyworkai.github.io\u002FDeepResearchAgent\u002F",null,"Python",3449,444,24,5,0,13,40,66,39,106.55,"MIT License",false,"main",true,[27,28,29,30],"gaia","general-purpose","multiagent-systems","multimodel","2026-06-12 04:01:04","# Deep Research Agent\n\nEnglish | [中文说明](README_zh.md)\n\nDeep Research Agent is a self-evolution protocol and runtime for LLM-based agent systems.\n\nRecent agent protocols often under-specify cross-entity **lifecycle\u002Fcontext management**, **version tracking**, and **safe evolution update interfaces**, which encourages monolithic compositions and brittle glue code. Deep Research Agent addresses this by decoupling **what evolves** from **how evolution occurs**:\n\n- **RSPL (Resource Substrate Protocol Layer)**: models *prompts, agents, tools, environments, and memory* as protocol-registered resources with explicit **state**, **lifecycle**, and **versioned** interfaces.\n- **SEPL (Self Evolution Protocol Layer)**: specifies a closed-loop operator interface to **propose**, **assess**, and **commit** improvements with auditable lineage and **rollback**.\n\nBuilt on Deep Research Agent, the system includes a **Deep Research Agent** style tool-calling agent that can dynamically instantiate\u002Fretrieve\u002Frefine resources and improve during execution.\n\n## Self-evolution at a glance\n\nAt a high level, Deep Research Agent supports an iterative loop:\n\n- **Act**: an agent produces actions\u002Foutputs using an LLM and the available tools\u002Fenvironments.\n- **Observe**: capture outcomes, traces, intermediate reasoning, and environment feedback.\n- **Optimize**: update prompts\u002Fsolutions\u002Fvariables using an optimizer (e.g., reflection or RL-style methods).\n- **Remember**: persist summaries\u002Finsights\u002Frecords to memory for later steps and sessions.\n\n## Core building blocks\n\n- **Agents (`src\u002Fagent\u002F`)**: runtime logic that decides *what to do next* (planning, tool-calling, domain agents, etc.).\n- **Tools (`src\u002Ftool\u002F`)**: callable capabilities exposed to agents (workflow tools + default tools).\n- **Environments (`src\u002Fenvironment\u002F`)**: stateful interfaces that tools\u002Fagents can interact with (filesystem, trading backtest envs, browser\u002Fmobile envs, etc.).\n- **Memory (`src\u002Fmemory\u002F`)**: session\u002Fevent memory systems for summarization, insights, and long-term state.\n- **Optimizers (`src\u002Foptimizer\u002F`)**: self-improvement algorithms that turn feedback into updated prompts\u002Fsolutions\u002Fvariables (reflection, GRPO, Reinforce++, etc.).\n- **Tracing & versioning (`src\u002Ftracer\u002F`, `src\u002Fversion\u002F`)**: record trajectories and manage iterative artifacts across runs.\n- **Config system (`configs\u002F`, `src\u002Fconfig\u002F`)**: MMEngine-style configs to compose agents\u002Ftools\u002Fenvs\u002Fmemory\u002Fmodels consistently.\n\n## Design goals\n\n- **Composable**: add\u002Freplace agents, tools, environments, memory systems, and optimizers without rewriting the whole stack.\n- **Inspectable**: structured traces and memory events make it easier to analyze failures and improvement steps.\n- **Evolvable**: explicit optimizers + persistent memory enable iterative refinement rather than one-shot inference.\n\n## Repository layout\n\n```\nDeepResearchAgent\u002F\n  configs\u002F                 # config composition (agents\u002Ftools\u002Fenvs\u002Fmemory\u002Fmodels)\n  src\u002F\n    agent\u002F                 # agents\n    environment\u002F           # environments\n    tool\u002F                  # tools\n    memory\u002F                # memory systems\n    optimizer\u002F             # self-evolution optimizers\n    model\u002F                 # model manager + provider backends\n    prompt\u002F                # prompt templates \u002F prompt manager\n    tracer\u002F                # tracing\n    version\u002F               # versioning\n  libs\u002F                    # vendored libraries\n  workdir\u002F                 # runtime artifacts (logs, traces, results, etc.)\n```\n\n## Optional: run a Tool-Calling Agent\n\nPrerequisites:\n- Install dependencies in your environment\n- Copy `.env.template` to `.env` and set a model API key (e.g. `OPENROUTER_API_KEY=...`)\n\nExample:\n\n```bash\npython examples\u002Frun_tool_calling_agent.py --config configs\u002Ftool_calling_agent.py\n```\n\nOverride model\u002Fworkdir:\n\n```bash\npython examples\u002Frun_tool_calling_agent.py \\\n  --config configs\u002Ftool_calling_agent.py \\\n  --cfg-options model_name=openrouter\u002Fgpt-4o workdir=workdir\u002Fdemo tag=demo\n```\n","DeepResearchAgent 是一个面向深度研究任务及通用任务解决的多层次多智能体系统。它通过顶层规划智能体协调多个专门化下层智能体，实现自动化任务分解与高效执行，适用于多样化和复杂的领域。项目采用Python开发，核心功能包括资源基底协议层（RSPL）和自我进化协议层（SEPL），前者将提示、智能体、工具、环境和记忆抽象为具有明确状态、生命周期和版本接口的注册资源；后者则提供了一个闭环操作界面，用于提出、评估并提交改进，并支持回滚。该系统特别适合需要灵活组合不同组件以完成复杂任务的研究或生产场景，如文件系统操作、交易回测、网页\u002F移动端环境交互等。",2,"2026-06-11 03:41:14","high_star"]