{"name":"Helen","kind":"llm_discovery","version":"1","about":"从论文理出课题。组内论文拆解与研究报告已打成可逐条继续分析的 JSON。","urls":{"discover":"https://helen.chenxofhit.xyz/llm","catalog":"https://helen.chenxofhit.xyz/llm/catalog","llms_txt":"https://helen.chenxofhit.xyz/llms.txt","mcp":"https://helen.chenxofhit.xyz/mcp","login":"https://helen.chenxofhit.xyz/auth/login","me":"https://helen.chenxofhit.xyz/auth/me","page":"https://helen.chenxofhit.xyz/"},"mcp":{"url":"https://helen.chenxofhit.xyz/mcp","transport":"streamable-http","auth":"本地 Agent：Authorization: Bearer <HELEN_MCP_TOKEN>。ChatGPT：只填 URL，走 OAuth，登录页是 Helen 账号。","oauth":{"resource_metadata":"https://helen.chenxofhit.xyz/.well-known/oauth-protected-resource/mcp","authorization_server":"https://helen.chenxofhit.xyz/.well-known/oauth-authorization-server","register":"https://helen.chenxofhit.xyz/oauth/register","authorize":"https://helen.chenxofhit.xyz/oauth/authorize","token":"https://helen.chenxofhit.xyz/oauth/token"},"token_configured":true,"tools":["list_catalog","get_paper","get_paper_pdf","get_paper_excerpt","get_idea","get_direction","get_report","get_combination","search_knowledge"]},"auth":{"required_for_data":true,"methods":["session_cookie","bearer"],"cookie":"raos_session","bearer":"Authorization: Bearer <HELEN_MCP_TOKEN 或 OAuth access_token>","bearer_get":["https://helen.chenxofhit.xyz/llm","https://helen.chenxofhit.xyz/llm/catalog","https://helen.chenxofhit.xyz/papers/{paper_id}/llm","https://helen.chenxofhit.xyz/papers/{paper_id}/llm/excerpt","https://helen.chenxofhit.xyz/papers/{paper_id}/llm/ideas/{idea_id}","https://helen.chenxofhit.xyz/papers/{paper_id}/llm/directions/{direction_id}","https://helen.chenxofhit.xyz/papers/{paper_id}/file","https://helen.chenxofhit.xyz/reports/{report_id}/llm","https://helen.chenxofhit.xyz/reports/{report_id}/combinations/{rank}","https://helen.chenxofhit.xyz/cognition/search"],"login":{"method":"POST","url":"https://helen.chenxofhit.xyz/auth/login","body":{"username":"","password":""}},"note":"浏览器登录用 cookie。带 HTTP 工具的模型对 bearer_get 里的 GET 带 Bearer，不必先登录。写操作不接受 Bearer。未登录且无 Bearer 时本文件只含约定，不含目录。"},"conventions":["本地 Agent 连 MCP：url 为本站 /mcp，Header 带 Bearer token。ChatGPT 只填这个 URL，按 OAuth 跳到 Helen 登录。","不会 MCP 时：先读本发现文档，再带 Bearer GET /llm/catalog（或读本文件 catalog）拿活目录，再按 urls.json 下钻。","目录优先：先读 */llm 或调用 list_catalog / get_paper / get_report，再读子资源。一次只分析一条创意、方向或报告组合。","核对原句用 GET /papers/{id}/llm/excerpt?query= 或 MCP get_paper_excerpt，返回有字数上限的摘录。整份 PDF 是 GET /papers/{id}/file，留给人下载。","禁止请求 /papers/{id}/details（可能含 parsed_text）。论文拆解用 /papers/{id}/llm 及其子资源，或 MCP get_paper / get_idea / get_direction。","禁止把 /reports/{id} 的 markdown 整篇塞进上下文。报告用 /reports/{id}/llm 与 /combinations/{rank}，或 MCP get_report / get_combination。","只使用 JSON 里出现的 id、Issue 编号、论文标题；禁止编造。","继续分析时严格按该资源 expected_output_schema 输出。"],"loop":["GET https://helen.chenxofhit.xyz/llm  或连接 MCP https://helen.chenxofhit.xyz/mcp","POST https://helen.chenxofhit.xyz/auth/login  （浏览器）或对后续 GET 带 Authorization: Bearer","GET https://helen.chenxofhit.xyz/llm/catalog  或 MCP list_catalog","GET https://helen.chenxofhit.xyz/papers/{paper_id}/llm  或  GET https://helen.chenxofhit.xyz/reports/{report_id}/llm；MCP 则 get_paper / get_report","再 GET 子资源或调用 get_idea / get_direction / get_combination；原句用 get_paper_excerpt，按 expected_output_schema 回复"],"resources":[{"kind":"paper_breakdown","index":"GET https://helen.chenxofhit.xyz/papers/{paper_id}/llm","pdf":"GET https://helen.chenxofhit.xyz/papers/{paper_id}/file","excerpt":"GET https://helen.chenxofhit.xyz/papers/{paper_id}/llm/excerpt?query={q}","mcp_pdf":"get_paper_pdf","mcp_excerpt":"get_paper_excerpt","mcp_resource":"helen://papers/{paper_id}/file","children":["GET https://helen.chenxofhit.xyz/papers/{paper_id}/llm/ideas/{idea_id}","GET https://helen.chenxofhit.xyz/papers/{paper_id}/llm/directions/{direction_id}"],"note":"解读在 index 的 interpretation。interpretation.source.pdf 与 urls.pdf 是原文，interpretation.source.breakdown 是同一篇拆解。","avoid":["GET https://helen.chenxofhit.xyz/papers/{paper_id}/details"]},{"kind":"report_index","index":"GET https://helen.chenxofhit.xyz/reports/{report_id}/llm","children":["GET https://helen.chenxofhit.xyz/reports/{report_id}/combinations/{rank}","GET https://helen.chenxofhit.xyz/reports/{report_id}?compact=true"],"avoid":["GET https://helen.chenxofhit.xyz/reports/{report_id}  （含长 markdown，除非只要 compact=true）"]},{"kind":"cognition_search","index":"GET https://helen.chenxofhit.xyz/cognition/search?query={q}&k=8","note":"按主题检索组内论文/创意/方向，再回到对应 /papers/{{id}}/llm。"}],"output_schemas":{"paper_idea":{"idea_id":"","verdict":"推进 | 暂缓 | 放弃","mechanism":"可对接的机制一句话","what_to_reuse":"相对原文真正可迁移的部分","next_experiment":"最小验证与指标","risks":["风险"]},"paper_direction":{"direction_id":"","verdict":"推进 | 暂缓 | 放弃","theme_stack":["模块"],"linked_idea_ids":[],"six_week_plan":[{"week":"1-2","task":"","metric":""}],"kill_criteria":""},"report_combination":{"rank":1,"verdict":"推进 | 暂缓 | 放弃","mechanism_stack":["模块A","模块B"],"what_is_new":"相对已给 Issues，真正可立项或可发表的增量；禁止空话","experiment_plan":[{"week":"1-2","task":"可执行任务","metric":"必须报告的指标"}],"must_report_metrics":["指标1"],"kill_criteria":"何种结果应停止堆叠或换方向","open_risks":["风险与缓解"],"followup_issue_numbers":[0],"notes":"不得编造未出现的 Issue 编号或论文标题"}},"authenticated":false,"catalog":null}