My Quant — Editorial Discovery
我的量化系统不会说英语
Liquidity Lab 系列产品的公共编辑与广告门户: Seiche、LiquiLens 和 LiquiLens—Undertow。
该网站在构建时消费每个产品的结构化编辑源。其
SOURCE_PUBLISHED 网站模式为每条来源标记为已发布的记录生成一个
MyQuant 阅读页面,同时保留专家的规范 URL、精确
声明、证据时钟、指纹和限制。移动应用信息流
独立地处于 SUSPENDED 状态,且包含零条故事。
公共网站有两条明确的通道:
Interpreted:对 Seiche、LiquiLens 和 Undertow 工作的通俗英语解读;MyQuant Analysis:独立来源的关于重要新闻的内部报道。
Interpreted 通道包含经过评审的长文和 LiquiLens 历史案例
档案。案例档案分别显示每个透镜的 HIT、MISS 或 VOID 状态,并
说明回放是否被重建。它们不会被静默地
提升为实时调用或经过验证的回测证据。
内部报道存放在 content/ 中,且必须声明其来源、
贡献、限制和发布状态。
npm test
npm run sync
npm run dev构建过程没有运行时依赖。如果某个来源暂时不可用,
它会使用该来源最后一次成功的 data/cache.json 切片,并在
生成的页面中打印覆盖状态。LiquiLens 桌面频道额外
固定到 liquilens.desk-bit-feed.v1;如果根模式缺失或变更,
则视为上游契约失败,因此构建会保留最后接受的频道,并将
LiquiLens 标记为 cached/degraded,而不是对未知结构进行规范化。
公共 API 和 MCP
该产品暴露只读的发现能力,不涉及账户或用户状态:
GET /api/v1/capabilities描述编辑、信息流和证据表面;GET /api/v1/health返回非敏感的兼容性心跳;POST /mcp提供无状态的 MCP 发现和健康工具。GET /openapi.json是精选的 REST 契约;GET /.well-known/ai-catalog.json和GET /.well-known/mcp.json暴露 公共 AI/MCP 发现,注册表清单位于GET /server.json。
产品版本(0.1.0)、REST 契约(myquant.editorial/1.1)和
MCP 服务器(2.0.0)保持独立。公共健康检查还会在可用时报告 Vercel 的精确
部署 Git SHA,从而允许定时冒烟测试和注册表
发布将实时端点绑定到经过评审的提交。
注册表验证使用 MYQUANT_USE_REVIEWED_CACHE=1 构建;这会使用
已提交的证据收据,而不是在发布 SHA 已被选定后
拉取不同的编辑线路。
MCP 端点支持当前的每请求协议元数据以及
列出的遗留协议握手。遗留的 Mcp-Session-Id 头
出于兼容性而被接受,但绝不会创建服务器端会话。跨源
MCP 请求仅限于配置的 MyQuant 站点源;REST
元数据保持公开且可缓存。
Related MCP server: Mavryn
评审后的影子副本
模型输出绝不会进入网络或定时同步路径。操作员可以 使用以下方式导入一个不可变的、已经过验证的候选:
npm run app-copy:import -- \
--packet /absolute/path/evidence-packet.json \
--candidate /absolute/path/draft.json \
--validator /absolute/path/validation.json \
--review /absolute/path/public-copy-review.json导入器检查精确的文件哈希、来源 ID 和指纹、证据
数据包和验证器绑定、支持指针、仅影子权限、不可变的
教师/模型修订,以及双人评审的裁决。来源修订会使
副本立即失去资格:网站使用其确定性的
SOURCE_GROUNDED 说明,而应用版本则省略该记录。参见
docs/APP-COPY-IMPORT.md 了解评审收据契约。
定时 sync evidence 工作流检查七个发布
频道,时间分别为
02:17、08:17、14:17 和 20:17 UTC。证据指纹发生变化会产生一个
小的内容提交,这将成为部署触发器;如果网络没有变化,
则不会产生提交,也不会重新部署。频率是一种检查节奏,而不是配额。
编辑规则
自动化可以收集并规范化已发布的记录。它不得制造
事实、从缺失的信息流中静默推断,或发布未经
content/README.md 中描述的来源和证据字段的内部报道。
网站归档模式是一种操作员分发指令,而不是声称
每个来源记录都获得了法律或监管许可。更正、
反向链接、规范链接、内容边界和紧急停止功能保持
有效。首页使用由
assets/media/ 中记录的已批准应用创意衍生出的第一方静默剪辑。
Available Tools
5 toolsget_healthCheck the My Quant public surfacesARead-onlyIdempotentInspect
Check that this endpoint responds with product.health.v1 metadata: product, API/MCP versions, deployed source SHA when available, and checked_at. Use for connection and version diagnosis; it does not fetch the editorial feed or prove source freshness. Public and read-only; no credentials or arguments. Use list_capabilities for supported surfaces and content limits.
get_storyRead one story by stable source IDARead-onlyIdempotentInspect
Retrieve one published story using its exact stable ID from latest_stories or search_stories. Returns story, feed receipt, and release identity while retaining canonical sources, evidence clocks, publication state, release IDs, and limitations. Matching is exact and case-sensitive; an unknown ID returns a tool error. Public and read-only; no credentials. The canonical feed is cached for up to 60 seconds; unavailable or invalid feeds return a tool error. Use search_stories first when you only know a title or topic.
Inputs: id (required). Returns: feed, story, schema, release
latest_storiesRead the latest evidence-bounded storiesARead-onlyIdempotentInspect
Read recent published archive records when you want a chronological overview without a search term. Returns count, stories, the feed receipt, and release identity; each story retains sources, evidence clocks, publication state, release IDs, and limitations. Public and read-only; no credentials. Lists default to 10 records, at most 20, without pagination. The canonical feed is cached for up to 60 seconds; unavailable or invalid feeds return a tool error. Use search_stories for text matching or get_story for an exact returned ID.
Inputs: limit. Returns: Published archive records with a feed receipt and deployed MCP release identity…
list_capabilitiesDiscover the My Quant editorial productARead-onlyIdempotentInspect
Discover available product features, API/MCP URLs, versions, content limits, and interpretation boundaries before choosing a content tool. Returns product.capabilities.v1 metadata, without fetching stories or checking upstream freshness. Public and read-only; no credentials or arguments. Use latest_stories for recent records or get_health for endpoint compatibility. The mobile app distribution channel remains separately suspended.
search_storiesSearch evidence-bounded storiesARead-onlyIdempotentInspect
Find published archive records by text when you do not know an exact story ID. All space-separated terms must match as case-insensitive substrings across IDs, titles, summaries, tags, products, beats, evidence status, contributions, or source labels/release IDs. Returns newest-first stories, normalized query, count, feed receipt, and release identity, preserving evidence clocks and limitations. No relevance ranking or pagination; zero matches returns an empty list. Public and read-only; no credentials. The feed is cached for up to 60 seconds; invalid queries or unavailable/invalid feeds return a tool error. Use get_story for an exact ID or latest_stories for an unfiltered overview.
Inputs: query (required), limit. Returns: Published archive records with a feed receipt and deployed MCP release identity…
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 5 tools
Each tool targets a clearly distinct capability: metadata discovery (list_capabilities), health/version diagnosis (get_health), chronological listing (latest_stories), exact-ID retrieval (get_story), and text search (search_stories). The descriptions explicitly cross-reference sibling tools ('Use search_stories for...', 'Use get_story for an exact ID'), removing ambiguity between the two metadata tools and between listing vs. searching. An agent can reliably select the right tool for a given intent.
Most tools follow a predictable verb_noun snake_case pattern (list_capabilities, get_health, get_story, search_stories). The outlier is latest_stories, an adjective_noun phrase with no verb, which slightly breaks the pattern. Otherwise the convention is consistent and readable.
Five tools is well-scoped for a read-only editorial discovery surface, with each tool earning its place (capabilities, health, list, get, search). No redundant or trivially thin endpoints, and no bloat. The set is tight and purposeful.
The surface covers the core read-only discovery lifecycle: discover capabilities, verify health, list recent stories, fetch by ID, and search by text. Minor gaps remain, such as no pagination beyond 20 records and no tag/beat browsing distinct from free-text search, but agents can work around these. Coverage is solid for the stated public, read-only purpose.
Maintenance
Related MCP Connectors
Trust, freshness, policy, and discovery layer for public MCP servers.
Agent-native registry to discover APIs, MCP servers and CLIs, with live health checks.
Guarded MCP server for agent-readable business truth, provenance, readiness, and discovery.
Live public MCP Index. Search, recommend, and call healthy MCP servers.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceAggregate, route, and orchestrate multiple MCP backend servers behind a single MCP endpoint.-
- AlicenseNot gradedqualityDmaintenanceCentralized MCP control plane that proxies multiple upstream MCP servers with tool namespacing, filtering, policy enforcement, audit logging, and health checks.2 npmMIT
- AlicenseAqualityBmaintenancePublic MCP research for AI finance and trading agents: funding, bank risk and liquidity evidence. Companion SDK: persistent Python MCP sessions and Python/JS source-status streams. Setup: https://beepboop2025.github.io/financial-evidence-skills/distribution/ai-trading-tools.html. No data API key. Keep source dates, rights and gaps. No live quotes or execution.340 PyPI1MIT
- AlicenseNot gradedqualityCmaintenanceProvides MCP clients with a searchable registry of MCP servers, enabling discovery and publishing of servers.47 npmMIT