@yanlinglabs/winter-provider-catalog 0.0.22 → 0.0.23

package/overlay/providers.json

@@ -849,11 +849,11 @@
     },
     {
       "id": "kimi-coding",
-      "displayName": "Kimi Code (Anthropic dialect)",
+      "displayName": "Kimi Code China (Anthropic dialect)",
       "protocols": [
         "anthropic-messages"
       ],
-      "$comment": "R6b-5 (WS-13b §2): a DIALECT SIBLING. The vendor documents two wire dialects at two different base URLs, and a provider row carries exactly one `adapterId`, so two dialects are two rows — each with its own `defaultEndpoints.api`, its own model keys, and the dialect in its `displayName`. `winter.anthropic-messages` is multi-provider in fact, measured rather than assumed: `runtime/src/provider/catalog-endpoint-shape.test.ts` drives two sibling rows against a loopback fake and each reaches ITS OWN `<root>/v1/messages`. A WINTER-OWNED row (`upstream.project: \"winter\"`, empty commit): it is not extracted, so nothing here came from the pinned tree's executable half. The endpoint and dialect are the VENDOR'S OWN documented values, fetched and read on 2026-09-06 — the `fetched-document` citation tier (PROVENANCE.md). SUBSCRIPTION-priced (WS-13b §1): `priceUsage` returns nothing for this row, so a Kimi Code session never reports `total_cost_usd` — a membership seat has no per-token price and reporting one would be a fabricated number. The OAuth half of this id is DROPPED per audit §5.2 (no third-party OAuth is documented); the key row is what ships. Endpoint is the ROOT: the adapter appends `/v1/messages`, reproducing the documented `https://api.kimi.com/coding/v1/messages`.",
+      "$comment": "R6b-5 (WS-13b §2): a DIALECT SIBLING. The vendor documents two wire dialects at two different base URLs, and a provider row carries exactly one `adapterId`, so two dialects are two rows — each with its own `defaultEndpoints.api`, its own model keys, and the dialect in its `displayName`. `winter.anthropic-messages` is multi-provider in fact, measured rather than assumed: `runtime/src/provider/catalog-endpoint-shape.test.ts` drives two sibling rows against a loopback fake and each reaches ITS OWN `<root>/v1/messages`. A WINTER-OWNED row (`upstream.project: \"winter\"`, empty commit): it is not extracted, so nothing here came from the pinned tree's executable half. The endpoint and dialect are the VENDOR'S OWN documented values, fetched and read on 2026-09-06 — the `fetched-document` citation tier (PROVENANCE.md). SUBSCRIPTION-priced (WS-13b §1): `priceUsage` returns nothing for this row, so a Kimi Code session never reports `total_cost_usd` — a membership seat has no per-token price and reporting one would be a fabricated number. The OAuth half of this id is DROPPED per audit §5.2 (no third-party OAuth is documented); the key row is what ships. Endpoint is the ROOT: the adapter appends `/v1/messages`, reproducing the documented `https://api.kimi.com/coding/v1/messages`. 2026-09-25 refresh: Kimi Code's docs (https://www.kimi.com/code/docs/en/, retrieved 2026-09-25) now label `api.kimi.com/coding[/v1]` as the CHINA endpoint and `api.kimi.ai/coding[/v1]` as the OVERSEAS one, with region-bound keys. This row has always pointed at api.kimi.com, so any key stored against it is a China key — the id keeps its host and is relabelled; the overseas plan is the separate `kimi-coding-intl` pair.",
       "authKinds": [
         "api-key"
       ],
@@ -886,8 +886,8 @@
     },
     {
       "id": "kimi-coding-openai",
-      "displayName": "Kimi Code (OpenAI dialect)",
-      "$comment": "The Kimi Code membership endpoint supports both documented dialects. This separate subscription row is the OpenAI-compatible sibling of `kimi-coding`, not Moonshot's token-billed platform API.",
+      "displayName": "Kimi Code China (OpenAI dialect)",
+      "$comment": "The Kimi Code membership endpoint supports both documented dialects. This separate subscription row is the OpenAI-compatible sibling of `kimi-coding`, not Moonshot's token-billed platform API. 2026-09-25 refresh: Kimi Code's docs (https://www.kimi.com/code/docs/en/, retrieved 2026-09-25) now label `api.kimi.com/coding[/v1]` as the CHINA endpoint and `api.kimi.ai/coding[/v1]` as the OVERSEAS one, with region-bound keys. This row has always pointed at api.kimi.com, so any key stored against it is a China key — the id keeps its host and is relabelled; the overseas plan is the separate `kimi-coding-intl` pair.",
       "protocols": [
         "openai-chat-completions"
       ],
@@ -1377,9 +1377,10 @@
       "id": "moonshot",
       "displayName": "Moonshot AI (Kimi platform)",
       "protocols": [
-        "openai-chat-completions"
+        "openai-chat-completions",
+        "openai-responses"
       ],
-      "$comment": "A WINTER-OWNED row (`upstream.project: \"winter\"`, empty commit): it is not extracted, so nothing here came from the pinned tree's executable half. The endpoint and dialect are the VENDOR'S OWN documented values, fetched and read on 2026-09-06 — the `fetched-document` citation tier (PROVENANCE.md). Upstream's `moonshot` entry states this same base URL but drives it through a bespoke `moonshot` executor, which Winter does not represent — an unknown executor FAILS the extraction run rather than defaulting, so the id is in `blocked` and the row is authored here. This is the TOKEN-PRICED platform API; `kimi-coding` is the separate SUBSCRIPTION plan, and audit §5.6(3) is explicit that conflating the two silently misreports cost.",
+      "$comment": "A WINTER-OWNED row (`upstream.project: \"winter\"`, empty commit): it is not extracted, so nothing here came from the pinned tree's executable half. The endpoint and dialect are the VENDOR'S OWN documented values, fetched and read on 2026-09-06 — the `fetched-document` citation tier (PROVENANCE.md). Upstream's `moonshot` entry states this same base URL but drives it through a bespoke `moonshot` executor, which Winter does not represent — an unknown executor FAILS the extraction run rather than defaulting, so the id is in `blocked` and the row is authored here. This is the TOKEN-PRICED platform API; `kimi-coding` is the separate SUBSCRIPTION plan, and audit §5.6(3) is explicit that conflating the two silently misreports cost. 2026-09-25 refresh: Moonshot's /v1/responses (https://platform.kimi.ai/docs/api/responses) is served on this same host and key — currently for kimi-k3 only — so it is a protocol of this row, not a separate provider (user ruling 2026-09-25).",
       "authKinds": [
         "api-key"
       ],
@@ -2597,6 +2598,2053 @@
         "tier": "fetched-document"
       },
       "scope": "llm"
+    },
+    {
+      "id": "google-openai",
+      "displayName": "Google Gemini API (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "Dialect sibling of `google` (native generateContent): the same Gemini API key and per-token billing, reached through Google's OpenAI-compatibility layer. A separate provider because a different wire dialect is a different provider in this catalog (user ruling 2026-09-25). The layer is documented as beta, so its parameter coverage is narrower than the native surface.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://generativelanguage.googleapis.com/v1beta/openai"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Google documents the OpenAI-compatibility layer as beta; parameter coverage (reasoning_effort mapping, unsupported fields) is narrower than the native generateContent surface."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://ai.google.dev/gemini-api/docs/openai — Google's Gemini API OpenAI-compatibility page, retrieved 2026-09-25: Chat Completions at `https://generativelanguage.googleapis.com/v1beta/openai/` with the ordinary Gemini API key as the bearer token; the page describes the layer as still in beta.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "moonshot-anthropic",
+      "displayName": "Moonshot AI (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `moonshot` on the international Kimi platform (per-token). Distinct from `kimi-coding`, which is the Kimi Code membership plan.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.moonshot.ai/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://platform.kimi.ai/docs/api/overview — Moonshot's (Kimi platform) API overview, retrieved 2026-09-25: the international platform serves Anthropic Messages at `https://api.moonshot.ai/anthropic` beside OpenAI Chat Completions/Responses at `https://api.moonshot.ai/v1`, with the same platform API key.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "moonshot-cn",
+      "displayName": "Moonshot AI China (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions",
+        "openai-responses"
+      ],
+      "$comment": "China-region twin of `moonshot`: own host, own account system, CNY billing.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.moonshot.cn/v1",
+        "models": "https://api.moonshot.cn/v1/models"
+      },
+      "modelDiscovery": "openai-models",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://platform.kimi.com/docs/api/overview — Moonshot's China platform API overview, retrieved 2026-09-25: OpenAI Chat Completions and Responses at `https://api.moonshot.cn/v1`, Anthropic Messages at `https://api.moonshot.cn/anthropic`; billed in CNY (https://platform.kimi.com/docs/pricing/chat).",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "moonshot-cn-anthropic",
+      "displayName": "Moonshot AI China (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `moonshot-cn`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.moonshot.cn/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://platform.kimi.com/docs/api/overview — Moonshot's China platform API overview, retrieved 2026-09-25: Anthropic Messages at `https://api.moonshot.cn/anthropic` with the China platform API key.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "zhipu",
+      "displayName": "Zhipu BigModel API (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "China-region twin of `zai` (Z.AI is Zhipu's international brand): own host and account system, CNY per-token billing.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://open.bigmodel.cn/api/paas/v4"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://docs.bigmodel.cn/cn/guide/start/model-overview and https://docs.bigmodel.cn/cn/guide/start/pricing — Zhipu BigModel (China) model overview and pricing, retrieved 2026-09-25: OpenAI-compatible Chat at `https://open.bigmodel.cn/api/paas/v4` with a BigModel API key; CNY per-token billing.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "zhipu-anthropic",
+      "displayName": "Zhipu BigModel API (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `zhipu`, sharing its URL with `zhipu-coding-anthropic` (credential-distinguished). Unlike the international `zai-anthropic`, the China docs explicitly document METERED keys on this root. Caution from the glm-5.3 model page: an account that has subscribed to the Coding Plan (even an expired one) is temporarily limited to OpenAI Chat for metered model calls.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://open.bigmodel.cn/api/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe.",
+          "Per the glm-5.3 model page, accounts that have ever subscribed to the GLM Coding Plan are temporarily limited to OpenAI Chat for metered calls, so this root may refuse such accounts."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://docs.bigmodel.cn/cn/guide/develop/claude/introduction — Zhipu BigModel's Claude-API compatibility guide, retrieved 2026-09-25: Anthropic Messages at `https://open.bigmodel.cn/api/anthropic` with an ordinary (metered) BigModel API key.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "zhipu-coding",
+      "displayName": "Zhipu GLM Coding Plan China (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "China-region twin of `zai-coding`: subscription plan, no per-token prices.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://open.bigmodel.cn/api/coding/paas/v4"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://docs.bigmodel.cn/cn/coding-plan/latest-model — Zhipu GLM Coding Plan (编程套餐) documentation, retrieved 2026-09-25: OpenAI-compatible `https://open.bigmodel.cn/api/coding/paas/v4` and Anthropic `https://open.bigmodel.cn/api/anthropic` with the plan key; the plan serves glm-5.3 and glm-5.3-flash (older ids reroute); billing is the subscription quota.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "zhipu-coding-anthropic",
+      "displayName": "Zhipu GLM Coding Plan China (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `zhipu-coding`; shares its URL with metered `zhipu-anthropic` (credential-distinguished).",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://open.bigmodel.cn/api/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://docs.bigmodel.cn/cn/coding-plan/latest-model — Zhipu GLM Coding Plan (编程套餐) documentation, retrieved 2026-09-25: OpenAI-compatible `https://open.bigmodel.cn/api/coding/paas/v4` and Anthropic `https://open.bigmodel.cn/api/anthropic` with the plan key; the plan serves glm-5.3 and glm-5.3-flash (older ids reroute); billing is the subscription quota.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "minimax-token-plan-anthropic",
+      "displayName": "MiniMax Token Plan (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `minimax-token-plan`. Shares its URL with metered `minimax-anthropic`: the `sk-cp-` plan key vs the `sk-api-` metered key decides billing.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.minimax.io/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://platform.minimax.io/docs/token-plan/other-tools — MiniMax Token Plan tool guide, retrieved 2026-09-25: the plan's `sk-cp-` key works against the Anthropic-compatible root `https://api.minimax.io/anthropic` (recommended there for prompt caching) as well as the OpenAI root.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "alibaba-anthropic",
+      "displayName": "Alibaba Cloud Model Studio (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `alibaba` (Model Studio international, per-token).",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://dashscope-intl.aliyuncs.com/apps/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://help.aliyun.com/en/model-studio/base-url — Alibaba Cloud Model Studio base-URL reference, retrieved 2026-09-25: the Singapore (international) region serves Anthropic Messages at `https://dashscope-intl.aliyuncs.com/apps/anthropic` with the same Model Studio API key as the OpenAI-compatible mode.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "alibaba-cn-anthropic",
+      "displayName": "Alibaba Cloud Bailian China (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `alibaba-cn` (Bailian China, per-token, CNY).",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://dashscope.aliyuncs.com/apps/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://help.aliyun.com/en/model-studio/base-url — Alibaba Cloud Model Studio base-URL reference, retrieved 2026-09-25: the Beijing (China) region serves Anthropic Messages at `https://dashscope.aliyuncs.com/apps/anthropic`.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "qwen-token-plan-anthropic",
+      "displayName": "Alibaba Cloud Token Plan (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `qwen-cloud-token-plan` (international Token Plan).",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://token-plan.ap-southeast-1.maas.aliyuncs.com/apps/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://www.alibabacloud.com/help/en/model-studio/token-plan-personal-overview and https://www.alibabacloud.com/help/en/model-studio/token-plan-personal-quick-start — Alibaba Cloud Token Plan (Personal) documentation, retrieved 2026-09-25: Anthropic base `https://token-plan.ap-southeast-1.maas.aliyuncs.com/apps/anthropic` with the plan key; Singapore-only; billing is plan credits.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "qwen-coding",
+      "displayName": "Alibaba Cloud Coding Plan (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "Alibaba Cloud Coding Plan: subscription with a request quota, separate host and `sk-sp-` key from metered Model Studio.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://coding-intl.dashscope.aliyuncs.com/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://help.aliyun.com/en/model-studio/coding-plan and https://help.aliyun.com/en/model-studio/coding-plan-faq — Alibaba Cloud Coding Plan documentation, retrieved 2026-09-25: the plan's `sk-sp-` key is used against `https://coding-intl.dashscope.aliyuncs.com/v1` (OpenAI) and `https://coding-intl.dashscope.aliyuncs.com/apps/anthropic` (Anthropic); billing is the plan's request quota.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "qwen-coding-anthropic",
+      "displayName": "Alibaba Cloud Coding Plan (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `qwen-coding`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://coding-intl.dashscope.aliyuncs.com/apps/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://help.aliyun.com/en/model-studio/coding-plan — Alibaba Cloud Coding Plan documentation, retrieved 2026-09-25: Anthropic Messages at `https://coding-intl.dashscope.aliyuncs.com/apps/anthropic` with the plan key.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "qwen-coding-cn",
+      "displayName": "Alibaba Cloud Coding Plan China (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "Alibaba Cloud Coding Plan: subscription with a request quota, separate host and `sk-sp-` key from metered Model Studio.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://coding.dashscope.aliyuncs.com/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://help.aliyun.com/en/model-studio/coding-plan and https://help.aliyun.com/en/model-studio/coding-plan-faq — Alibaba Cloud Coding Plan documentation, retrieved 2026-09-25: the plan's `sk-sp-` key is used against `https://coding.dashscope.aliyuncs.com/v1` (OpenAI) and `https://coding.dashscope.aliyuncs.com/apps/anthropic` (Anthropic); billing is the plan's request quota.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "qwen-coding-cn-anthropic",
+      "displayName": "Alibaba Cloud Coding Plan China (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `qwen-coding-cn`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://coding.dashscope.aliyuncs.com/apps/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://help.aliyun.com/en/model-studio/coding-plan — Alibaba Cloud Coding Plan documentation, retrieved 2026-09-25: Anthropic Messages at `https://coding.dashscope.aliyuncs.com/apps/anthropic` with the plan key.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "xiaomi",
+      "displayName": "Xiaomi MiMo API (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "Xiaomi's first-party MiMo API (pay-as-you-go).",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.xiaomimimo.com/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://mimo.mi.com/docs/zh-CN/quick-start/summary/first-api-call — Xiaomi MiMo open platform quick start, retrieved 2026-09-25: OpenAI-compatible Chat Completions at `https://api.xiaomimimo.com/v1` and Anthropic Messages at `https://api.xiaomimimo.com/anthropic`, both with the platform's `sk-` key, per-token billing.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "xiaomi-anthropic",
+      "displayName": "Xiaomi MiMo API (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `xiaomi`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.xiaomimimo.com/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://mimo.mi.com/docs/zh-CN/quick-start/summary/first-api-call — Xiaomi MiMo open platform quick start, retrieved 2026-09-25: Anthropic Messages at `https://api.xiaomimimo.com/anthropic` (x-api-key).",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "qianfan-anthropic",
+      "displayName": "Baidu Qianfan (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `qianfan`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://qianfan.baidubce.com/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://cloud.baidu.com/doc/qianfan-docs/s/6mh3e6gjp — Baidu Qianfan Anthropic-compatibility documentation, retrieved 2026-09-25: every chat model is reachable through the Anthropic Messages specification at `https://qianfan.baidubce.com/anthropic` (x-api-key), per-token CNY billing.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "qianfan-coding",
+      "displayName": "Baidu Qianfan Coding Plan (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "Baidu Qianfan Coding Plan (subscription). Its roster is largely third-party models resold by Baidu.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://qianfan.baidubce.com/v2/coding"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "Baidu's Coding Plan page (https://cloud.baidu.com/doc/qianfan/s/imlg0beiu) states the plan is for interactive coding tools and NOT for scripts or backend/batch calls; its function-calling list (https://cloud.baidu.com/doc/qianfan-docs/s/xm95lyys5) names none of the plan's models, so every row here is tool-less (fail-closed) until Baidu documents it."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://cloud.baidu.com/doc/qianfan/s/imlg0beiu — Baidu Qianfan Coding Plan documentation, retrieved 2026-09-25: the plan's dedicated key is used against `https://qianfan.baidubce.com/v2/coding` (OpenAI) and `https://qianfan.baidubce.com/anthropic/coding` (Anthropic); billing is the plan's request quota.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "qianfan-coding-anthropic",
+      "displayName": "Baidu Qianfan Coding Plan (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `qianfan-coding`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://qianfan.baidubce.com/anthropic/coding"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe.",
+          "Baidu's Coding Plan page (https://cloud.baidu.com/doc/qianfan/s/imlg0beiu) states the plan is for interactive coding tools and NOT for scripts or backend/batch calls; its function-calling list (https://cloud.baidu.com/doc/qianfan-docs/s/xm95lyys5) names none of the plan's models, so every row here is tool-less (fail-closed) until Baidu documents it."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://cloud.baidu.com/doc/qianfan/s/imlg0beiu — Baidu Qianfan Coding Plan documentation, retrieved 2026-09-25: Anthropic root `https://qianfan.baidubce.com/anthropic/coding`.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "meta",
+      "displayName": "Meta Model API (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions",
+        "openai-responses"
+      ],
+      "$comment": "Meta's first-party hosted API since the Llama API public preview was retired on 2026-07-06 (see the `meta-llama` row). Serves the Muse family. The `-contributor` model variants are catalogued at the user's request (2026-09-25): they cost less because Meta may train on their prompts and completions (https://dev.meta.ai/docs/models), which their display names state.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.meta.ai/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://dev.meta.ai/docs/models and https://dev.meta.ai/docs/pricing-rate-limits — Meta Model API documentation, retrieved 2026-09-25: OpenAI-compatible Chat Completions and Responses at `https://api.meta.ai/v1` with a Model API key; pay-as-you-go per-token billing; serves the Muse Spark models.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "kimi-coding-intl",
+      "displayName": "Kimi Code Overseas (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Overseas twin of `kimi-coding` (which is the China plan at api.kimi.com). Endpoint is the ROOT: the adapter appends `/v1/messages`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.kimi.ai/coding"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://www.kimi.com/code/docs/en/ — Kimi Code's own docs, retrieved 2026-09-25: the overseas plan's Anthropic-compatible base is `https://api.kimi.ai/coding/` and its OpenAI-compatible base `https://api.kimi.ai/coding/v1` (the China plan uses api.kimi.com); membership API keys are region-bound; billing is the membership subscription.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "kimi-coding-intl-openai",
+      "displayName": "Kimi Code Overseas (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "Dialect sibling of `kimi-coding-intl`; overseas twin of `kimi-coding-openai`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.kimi.ai/coding/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://www.kimi.com/code/docs/en/ — Kimi Code's own docs, retrieved 2026-09-25: the overseas plan's Anthropic-compatible base is `https://api.kimi.ai/coding/` and its OpenAI-compatible base `https://api.kimi.ai/coding/v1` (the China plan uses api.kimi.com); membership API keys are region-bound; billing is the membership subscription.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "xiaomi-token-plan",
+      "displayName": "Xiaomi MiMo Token Plan Singapore (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "Xiaomi MiMo Token Plan, the Singapore cluster (international default). One row per cluster because the plan key is issued with its cluster's Base URL.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://token-plan-sgp.xiaomimimo.com/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The key and Base URL are issued per subscription and the account region must match the cluster (https://mimo.mi.com/docs/en-US/quick-start/faq/token-plan/Team-Benefits); a key is not documented to work on another cluster's host."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://mimo.mi.com/docs/en-US/tokenplan/Token%20Plan/quick-access — Xiaomi MiMo Token Plan quick access, retrieved 2026-09-25: the Singapore cluster (international default) serves OpenAI-compatible `https://token-plan-sgp.xiaomimimo.com/v1` and Anthropic-compatible `https://token-plan-sgp.xiaomimimo.com/anthropic`; after subscribing the user receives an exclusive `tp-` key and Base URL (\"the Base URL shall be subject to what is displayed on the Token Plan page\"); billing is the subscription plan.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "xiaomi-token-plan-anthropic",
+      "displayName": "Xiaomi MiMo Token Plan Singapore (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `xiaomi-token-plan`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://token-plan-sgp.xiaomimimo.com/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The key and Base URL are issued per subscription and the account region must match the cluster (https://mimo.mi.com/docs/en-US/quick-start/faq/token-plan/Team-Benefits); a key is not documented to work on another cluster's host.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://mimo.mi.com/docs/en-US/tokenplan/Token%20Plan/quick-access — Xiaomi MiMo Token Plan quick access, retrieved 2026-09-25: the Singapore cluster (international default) serves OpenAI-compatible `https://token-plan-sgp.xiaomimimo.com/v1` and Anthropic-compatible `https://token-plan-sgp.xiaomimimo.com/anthropic`; after subscribing the user receives an exclusive `tp-` key and Base URL (\"the Base URL shall be subject to what is displayed on the Token Plan page\"); billing is the subscription plan.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "xiaomi-token-plan-cn",
+      "displayName": "Xiaomi MiMo Token Plan China (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "Xiaomi MiMo Token Plan, the China cluster. One row per cluster because the plan key is issued with its cluster's Base URL.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://token-plan-cn.xiaomimimo.com/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The key and Base URL are issued per subscription and the account region must match the cluster (https://mimo.mi.com/docs/en-US/quick-start/faq/token-plan/Team-Benefits); a key is not documented to work on another cluster's host."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://mimo.mi.com/docs/en-US/tokenplan/Token%20Plan/quick-access — Xiaomi MiMo Token Plan quick access, retrieved 2026-09-25: the China cluster serves OpenAI-compatible `https://token-plan-cn.xiaomimimo.com/v1` and Anthropic-compatible `https://token-plan-cn.xiaomimimo.com/anthropic`; after subscribing the user receives an exclusive `tp-` key and Base URL (\"the Base URL shall be subject to what is displayed on the Token Plan page\"); billing is the subscription plan.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "xiaomi-token-plan-cn-anthropic",
+      "displayName": "Xiaomi MiMo Token Plan China (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `xiaomi-token-plan-cn`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://token-plan-cn.xiaomimimo.com/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The key and Base URL are issued per subscription and the account region must match the cluster (https://mimo.mi.com/docs/en-US/quick-start/faq/token-plan/Team-Benefits); a key is not documented to work on another cluster's host.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://mimo.mi.com/docs/en-US/tokenplan/Token%20Plan/quick-access — Xiaomi MiMo Token Plan quick access, retrieved 2026-09-25: the China cluster serves OpenAI-compatible `https://token-plan-cn.xiaomimimo.com/v1` and Anthropic-compatible `https://token-plan-cn.xiaomimimo.com/anthropic`; after subscribing the user receives an exclusive `tp-` key and Base URL (\"the Base URL shall be subject to what is displayed on the Token Plan page\"); billing is the subscription plan.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "xiaomi-token-plan-eu",
+      "displayName": "Xiaomi MiMo Token Plan Europe (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "Xiaomi MiMo Token Plan, the Amsterdam (Europe) cluster. One row per cluster because the plan key is issued with its cluster's Base URL.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://token-plan-ams.xiaomimimo.com/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The key and Base URL are issued per subscription and the account region must match the cluster (https://mimo.mi.com/docs/en-US/quick-start/faq/token-plan/Team-Benefits); a key is not documented to work on another cluster's host."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://mimo.mi.com/docs/en-US/tokenplan/Token%20Plan/quick-access — Xiaomi MiMo Token Plan quick access, retrieved 2026-09-25: the Amsterdam (Europe) cluster serves OpenAI-compatible `https://token-plan-ams.xiaomimimo.com/v1` and Anthropic-compatible `https://token-plan-ams.xiaomimimo.com/anthropic`; after subscribing the user receives an exclusive `tp-` key and Base URL (\"the Base URL shall be subject to what is displayed on the Token Plan page\"); billing is the subscription plan.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "xiaomi-token-plan-eu-anthropic",
+      "displayName": "Xiaomi MiMo Token Plan Europe (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `xiaomi-token-plan-eu`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://token-plan-ams.xiaomimimo.com/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The key and Base URL are issued per subscription and the account region must match the cluster (https://mimo.mi.com/docs/en-US/quick-start/faq/token-plan/Team-Benefits); a key is not documented to work on another cluster's host.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://mimo.mi.com/docs/en-US/tokenplan/Token%20Plan/quick-access — Xiaomi MiMo Token Plan quick access, retrieved 2026-09-25: the Amsterdam (Europe) cluster serves OpenAI-compatible `https://token-plan-ams.xiaomimimo.com/v1` and Anthropic-compatible `https://token-plan-ams.xiaomimimo.com/anthropic`; after subscribing the user receives an exclusive `tp-` key and Base URL (\"the Base URL shall be subject to what is displayed on the Token Plan page\"); billing is the subscription plan.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "tencent-tokenhub",
+      "displayName": "Tencent TokenHub China (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions",
+        "openai-responses"
+      ],
+      "$comment": "Tencent's current first-party Hunyuan (Hy) surface, China site — the successor of the old Hunyuan platform (`tencent`, shutting down 2026-09-30).",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://tokenhub.tencentmaas.com/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://cloud.tencent.com/document/product/1823/130079 — Tencent Cloud TokenHub API documentation, retrieved 2026-09-25: one base `https://tokenhub.tencentmaas.com/v1` serves OpenAI Chat Completions (`/chat/completions`), Responses (`/responses`) and Anthropic Messages (`/messages`) with the TokenHub API key; per-token billing. The China site issues its own keys (China `tokenhub.tencentmaas.com`, Singapore `tokenhub-intl.tencentmaas.com`).",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "tencent-tokenhub-anthropic",
+      "displayName": "Tencent TokenHub China (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `tencent-tokenhub`. Endpoint is the ROOT: the adapter appends `/v1/messages`, reproducing the documented `https://tokenhub.tencentmaas.com/v1/messages`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://tokenhub.tencentmaas.com"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://cloud.tencent.com/document/product/1823/130079 — Tencent Cloud TokenHub API documentation, retrieved 2026-09-25: one base `https://tokenhub.tencentmaas.com/v1` serves OpenAI Chat Completions (`/chat/completions`), Responses (`/responses`) and Anthropic Messages (`/messages`) with the TokenHub API key; per-token billing. The China site issues its own keys (China `tokenhub.tencentmaas.com`, Singapore `tokenhub-intl.tencentmaas.com`).",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "tencent-tokenhub-intl",
+      "displayName": "Tencent TokenHub International (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions",
+        "openai-responses"
+      ],
+      "$comment": "Tencent's current first-party Hunyuan (Hy) surface, International site — the successor of the old Hunyuan platform (`tencent`, shutting down 2026-09-30).",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://tokenhub-intl.tencentmaas.com/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://cloud.tencent.com/document/product/1823/130079 — Tencent Cloud TokenHub API documentation, retrieved 2026-09-25: one base `https://tokenhub-intl.tencentmaas.com/v1` serves OpenAI Chat Completions (`/chat/completions`), Responses (`/responses`) and Anthropic Messages (`/messages`) with the TokenHub API key; per-token billing. The International site issues its own keys (China `tokenhub.tencentmaas.com`, Singapore `tokenhub-intl.tencentmaas.com`).",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "tencent-tokenhub-intl-anthropic",
+      "displayName": "Tencent TokenHub International (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `tencent-tokenhub-intl`. Endpoint is the ROOT: the adapter appends `/v1/messages`, reproducing the documented `https://tokenhub-intl.tencentmaas.com/v1/messages`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://tokenhub-intl.tencentmaas.com"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://cloud.tencent.com/document/product/1823/130079 — Tencent Cloud TokenHub API documentation, retrieved 2026-09-25: one base `https://tokenhub-intl.tencentmaas.com/v1` serves OpenAI Chat Completions (`/chat/completions`), Responses (`/responses`) and Anthropic Messages (`/messages`) with the TokenHub API key; per-token billing. The International site issues its own keys (China `tokenhub.tencentmaas.com`, Singapore `tokenhub-intl.tencentmaas.com`).",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "tencent-coding",
+      "displayName": "Tencent Cloud Coding Plan (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "Tencent Cloud Coding Plan (subscription).",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.lkeap.cloud.tencent.com/coding/v3"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://cloud.tencent.com/document/product/1823/130092 — Tencent Cloud Coding Plan documentation, retrieved 2026-09-25: the plan key (`sk-sp-`) is used against `https://api.lkeap.cloud.tencent.com/coding/v3` (OpenAI) and `https://api.lkeap.cloud.tencent.com/coding/anthropic` (Anthropic); billing is the subscription plan.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "tencent-coding-anthropic",
+      "displayName": "Tencent Cloud Coding Plan (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `tencent-coding`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.lkeap.cloud.tencent.com/coding/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://cloud.tencent.com/document/product/1823/130092 — Tencent Cloud Coding Plan documentation, retrieved 2026-09-25: the plan key (`sk-sp-`) is used against `https://api.lkeap.cloud.tencent.com/coding/v3` (OpenAI) and `https://api.lkeap.cloud.tencent.com/coding/anthropic` (Anthropic); billing is the subscription plan.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "tencent-token-plan",
+      "displayName": "Tencent Cloud Token Plan (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "Tencent Cloud Token Plan (subscription).",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.lkeap.cloud.tencent.com/plan/v3"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://cloud.tencent.com/document/product/1823/130060 — Tencent Cloud Token Plan documentation, retrieved 2026-09-25: the plan key (`sk-sp-`) is used against `https://api.lkeap.cloud.tencent.com/plan/v3` (OpenAI) and `https://api.lkeap.cloud.tencent.com/plan/anthropic` (Anthropic); billing is the subscription plan.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "tencent-token-plan-anthropic",
+      "displayName": "Tencent Cloud Token Plan (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `tencent-token-plan`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.lkeap.cloud.tencent.com/plan/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://cloud.tencent.com/document/product/1823/130060 — Tencent Cloud Token Plan documentation, retrieved 2026-09-25: the plan key (`sk-sp-`) is used against `https://api.lkeap.cloud.tencent.com/plan/v3` (OpenAI) and `https://api.lkeap.cloud.tencent.com/plan/anthropic` (Anthropic); billing is the subscription plan.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "minimax-cn-anthropic",
+      "displayName": "MiniMax China (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `minimax-cn`.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.minimax.cn/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe.",
+          "The Anthropic root string was read by the research pass but not re-exposed on a later fetch; it follows the documented OpenAI host migration."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://platform.minimax.cn/docs/api-reference/text-anthropic-api — MiniMax China Anthropic-compatible API reference, retrieved 2026-09-25: Anthropic Messages at `https://api.minimax.cn/anthropic` for all eight text models, `sk-api-` key.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "minimax-cn-token-plan",
+      "displayName": "MiniMax China Token Plan (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "China twin of `minimax-token-plan`; shares its URL with metered `minimax-cn` (credential-distinguished).",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.minimax.cn/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://platform.minimax.cn/docs/token-plan/intro — MiniMax China Token Plan documentation, retrieved 2026-09-25: the China plan (sk-cp- key) is used against the China hosts `https://api.minimax.cn/v1` and `https://api.minimax.cn/anthropic`; billing is the subscription plan.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "minimax-cn-token-plan-anthropic",
+      "displayName": "MiniMax China Token Plan (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "China twin of `minimax-token-plan`; shares its URL with metered `minimax-cn` (credential-distinguished).",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.minimax.cn/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://platform.minimax.cn/docs/token-plan/intro — MiniMax China Token Plan documentation, retrieved 2026-09-25: the China plan (sk-cp- key) is used against the China hosts `https://api.minimax.cn/v1` and `https://api.minimax.cn/anthropic`; billing is the subscription plan.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "meta-anthropic",
+      "displayName": "Meta Model API (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `meta`. Endpoint is the ROOT: the adapter appends `/v1/messages`, exactly the documented path.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.meta.ai"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "The vendor's OpenAI-dialect reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://dev.meta.ai/docs/protocols/messages — Meta Model API Messages protocol, retrieved 2026-09-25: Anthropic SDK `base_url=\"https://api.meta.ai\"`, POST `/v1/messages` (and `/v1/messages/count_tokens`) with the same Model API key; per-token billing.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "zai-anthropic",
+      "displayName": "Z.AI API (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.z.ai/api/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "OmniRoute",
+        "commit": "5458026c216f77a3da68ea49152dc33470cfe2cb",
+        "sourcePaths": [
+          "open-sse/config/providers/registry/zai/index.ts",
+          "src/shared/constants/providers/apikey/index.ts"
+        ]
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "The cited documentation (docs.z.ai/devpack/tool/claude, and docs.z.ai/scenario-example/develop-tools/claude re-read 2026-09-25) describes `https://api.z.ai/api/anthropic` only for the GLM Coding Plan; no page read states that METERED (pay-as-you-go) API keys are accepted there. Needs a live metered-key probe before this row can be trusted for token billing.",
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint."
+        ]
+      },
+      "scope": "llm",
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://docs.z.ai/devpack/tool/claude — z.ai's own GLM Coding Plan documentation for Anthropic-compatible clients, retrieved 2026-09-06: `ANTHROPIC_BASE_URL` is `\"https://api.z.ai/api/anthropic\"`, which is byte-identical to the API root the mapper derives from this id's pinned RegistryEntry (`format: \"claude\"` at `https://api.z.ai/api/anthropic/v1/messages`). Two independent sources, one value. (Round-1 finding I-4: the previous citation named `open.bigmodel.cn`, the mainland product site, and sat in the pinned-upstream tier although the lane had fetched this document.)",
+        "tier": "fetched-document"
+      }
+    },
+    {
+      "id": "qwen-cloud-token-plan",
+      "displayName": "Alibaba Cloud Token Plan (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "OmniRoute",
+        "commit": "5458026c216f77a3da68ea49152dc33470cfe2cb",
+        "sourcePaths": [
+          "open-sse/config/providers/registry/qwen-cloud-token-plan/index.ts",
+          "src/shared/constants/providers/apikey/index.ts"
+        ]
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "WS-13b §1 admission evidence is PINNED-UPSTREAM, not a fetched vendor document: the citation names the vendor site OmniRoute records, and the api-key path is attested by the pinned RegistryEntry rather than by a doc this lane read",
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Subscription (plan) credential: usage is metered against plan quota, not per-token list prices, so models on this row carry no `pricing`."
+        ]
+      },
+      "scope": "llm",
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://www.qwencloud.com/pricing/token-plan — the vendor's own site as OmniRoute's product catalog records it at 5458026 (src/shared/constants/providers/apikey/*.ts, blob-pinned in extraction-manifest.json). The api-key path is attested by the pinned RegistryEntry: authType \"apikey\", OpenAI Chat Completions at https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1. EVIDENCE TIER: pinned-upstream, NOT a page this lane fetched and read — see PROVENANCE.md \"Two tiers of admission citation\".",
+        "tier": "pinned-upstream"
+      },
+      "$comment": "2026-09-25 refresh: re-admitted on fetched Alibaba documentation (tier fetched-document) and corrected to `subscription` — the Token Plan is a credits subscription, so its models carry no per-token price. Alibaba's Sep 23/25 pages state the plan is available ONLY in the Singapore region (the cn-beijing host on the base-url page is not a launched product). FIRST KEY ONLY (R-FW-3): the refresh fetched and read https://www.alibabacloud.com/help/en/model-studio/token-plan-personal-overview and https://www.alibabacloud.com/help/en/model-studio/token-plan-personal-quick-start — Alibaba Cloud Token Plan (Personal) documentation, retrieved 2026-09-25: Singapore-only subscription plan; OpenAI-compatible base `https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1` and Anthropic base `.../apps/anthropic` with the plan key; billing is plan credits. — the tier stays `pinned-upstream` until a live-gate pass supplies the second key."
+    },
+    {
+      "id": "alibaba",
+      "displayName": "Alibaba Cloud Model Studio (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions",
+        "openai-responses"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
+        "models": "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/models"
+      },
+      "modelDiscovery": "openai-models",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "OmniRoute",
+        "commit": "5458026c216f77a3da68ea49152dc33470cfe2cb",
+        "sourcePaths": [
+          "open-sse/config/providers/registry/alibaba/index.ts",
+          "src/shared/constants/providers/apikey/index.ts"
+        ]
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "WS-13b §1 admission evidence is PINNED-UPSTREAM, not a fetched vendor document: the citation names the vendor site OmniRoute records, and the api-key path is attested by the pinned RegistryEntry rather than by a doc this lane read",
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint."
+        ]
+      },
+      "scope": "llm",
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://bailian.console.alibabacloud.com/ — the vendor's own site as OmniRoute's product catalog records it at 5458026 (src/shared/constants/providers/apikey/*.ts, blob-pinned in extraction-manifest.json). The api-key path is attested by the pinned RegistryEntry: authType \"apikey\", OpenAI Chat Completions at https://dashscope-intl.aliyuncs.com/compatible-mode/v1. EVIDENCE TIER: pinned-upstream, NOT a page this lane fetched and read — see PROVENANCE.md \"Two tiers of admission citation\".",
+        "tier": "pinned-upstream"
+      },
+      "$comment": "2026-09-25 refresh: re-admitted on Alibaba's own base-url page (tier fetched-document). Model Studio also serves OpenAI Responses at `/compatible-mode/v1/responses` on this host and key (https://www.alibabacloud.com/help/en/model-studio/qwen-api-via-openai-responses) — listed as a protocol, but rows stay `chat` until adapter-level Responses support lands. FIRST KEY ONLY (R-FW-3): the refresh fetched and read https://help.aliyun.com/en/model-studio/base-url — Alibaba Cloud Model Studio base-URL reference, retrieved 2026-09-25: Singapore (international) OpenAI-compatible base `https://dashscope-intl.aliyuncs.com/compatible-mode/v1`, Anthropic base `.../apps/anthropic`; per-token billing (https://www.alibabacloud.com/help/en/model-studio/model-pricing). — the tier stays `pinned-upstream` until a live-gate pass supplies the second key."
+    },
+    {
+      "id": "alibaba-cn",
+      "displayName": "Alibaba Cloud Bailian China (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://dashscope.aliyuncs.com/compatible-mode/v1",
+        "models": "https://dashscope.aliyuncs.com/compatible-mode/v1/models"
+      },
+      "modelDiscovery": "openai-models",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "OmniRoute",
+        "commit": "5458026c216f77a3da68ea49152dc33470cfe2cb",
+        "sourcePaths": [
+          "open-sse/config/providers/registry/alibaba/cn/index.ts",
+          "src/shared/constants/providers/apikey/index.ts"
+        ]
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "WS-13b §1 admission evidence is PINNED-UPSTREAM, not a fetched vendor document: the citation names the vendor site OmniRoute records, and the api-key path is attested by the pinned RegistryEntry rather than by a doc this lane read",
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys."
+        ]
+      },
+      "scope": "llm",
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://dashscope.console.aliyun.com/ — the vendor's own site as OmniRoute's product catalog records it at 5458026 (src/shared/constants/providers/apikey/*.ts, blob-pinned in extraction-manifest.json). The api-key path is attested by the pinned RegistryEntry: authType \"apikey\", OpenAI Chat Completions at https://dashscope.aliyuncs.com/compatible-mode/v1. EVIDENCE TIER: pinned-upstream, NOT a page this lane fetched and read — see PROVENANCE.md \"Two tiers of admission citation\".",
+        "tier": "pinned-upstream"
+      },
+      "$comment": "2026-09-25 refresh: re-admitted on Alibaba's own base-url page (tier fetched-document); CNY per-token billing (https://help.aliyun.com/zh/model-studio/model-pricing). FIRST KEY ONLY (R-FW-3): the refresh fetched and read https://help.aliyun.com/en/model-studio/base-url — Alibaba Cloud Model Studio base-URL reference, retrieved 2026-09-25: Beijing (China) OpenAI-compatible base `https://dashscope.aliyuncs.com/compatible-mode/v1`, Anthropic base `.../apps/anthropic`. — the tier stays `pinned-upstream` until a live-gate pass supplies the second key."
+    },
+    {
+      "id": "qianfan",
+      "displayName": "Baidu Qianfan (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://qianfan.baidubce.com/v2",
+        "models": "https://qianfan.baidubce.com/v2/models"
+      },
+      "modelDiscovery": "openai-models",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "OmniRoute",
+        "commit": "5458026c216f77a3da68ea49152dc33470cfe2cb",
+        "sourcePaths": [
+          "open-sse/config/providers/registry/qianfan/index.ts",
+          "src/shared/constants/providers/apikey/index.ts"
+        ]
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "WS-13b §1 admission evidence is PINNED-UPSTREAM, not a fetched vendor document: the citation names the vendor site OmniRoute records, and the api-key path is attested by the pinned RegistryEntry rather than by a doc this lane read",
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys."
+        ]
+      },
+      "scope": "llm",
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://cloud.baidu.com/product-s/qianfan_home — the vendor's own site as OmniRoute's product catalog records it at 5458026 (src/shared/constants/providers/apikey/*.ts, blob-pinned in extraction-manifest.json). The api-key path is attested by the pinned RegistryEntry: authType \"apikey\", OpenAI Chat Completions at https://qianfan.baidubce.com/v2. EVIDENCE TIER: pinned-upstream, NOT a page this lane fetched and read — see PROVENANCE.md \"Two tiers of admission citation\".",
+        "tier": "pinned-upstream"
+      },
+      "$comment": "2026-09-25 refresh: re-admitted on Baidu's fetched model list (tier fetched-document). The duplicate `baidu` row points at the same endpoint and is left untouched pending a consolidation decision. FIRST KEY ONLY (R-FW-3): the refresh fetched and read https://cloud.baidu.com/doc/qianfan-docs/s/7m95lyy43 — Baidu Qianfan's current served-model list (updated 2026-07-13), retrieved 2026-09-25: OpenAI-compatible base `https://qianfan.baidubce.com/v2`, Anthropic-compatible base `https://qianfan.baidubce.com/anthropic`; per-token CNY billing. — the tier stays `pinned-upstream` until a live-gate pass supplies the second key."
+    },
+    {
+      "id": "tencent",
+      "displayName": "Tencent Hunyuan (legacy platform)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.hunyuan.cloud.tencent.com/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "OmniRoute",
+        "commit": "5458026c216f77a3da68ea49152dc33470cfe2cb",
+        "sourcePaths": [
+          "open-sse/config/providers/registry/tencent/index.ts",
+          "src/shared/constants/providers/apikey/index.ts"
+        ]
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Platform shutting down 2026-09-30 per https://cloud.tencent.com/document/product/1729/131925; use `tencent-tokenhub`.",
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint."
+        ]
+      },
+      "scope": "llm",
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://hunyuan.tencent.com — the vendor's own site as OmniRoute's product catalog records it at 5458026 (src/shared/constants/providers/apikey/*.ts, blob-pinned in extraction-manifest.json). The api-key path is attested by the pinned RegistryEntry: authType \"apikey\", OpenAI Chat Completions at https://api.hunyuan.cloud.tencent.com/v1. EVIDENCE TIER: pinned-upstream, NOT a page this lane fetched and read — see PROVENANCE.md \"Two tiers of admission citation\".",
+        "tier": "pinned-upstream"
+      },
+      "$comment": "2026-09-25 refresh: Tencent announced (https://cloud.tencent.com/document/product/1729/131925) that the old Hunyuan platform at api.hunyuan.cloud.tencent.com shuts down on 2026-09-30 and that its main ids (hunyuan-turbos-latest, -t1-latest, -pro, -lite, -functioncall, the 2.0 dated ids) went offline on 2026-06-22. Successor: `tencent-tokenhub`."
+    },
+    {
+      "id": "minimax-cn",
+      "displayName": "MiniMax China (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.minimax.cn/v1",
+        "models": "https://api.minimax.cn/v1/models"
+      },
+      "modelDiscovery": "openai-models",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "OmniRoute",
+        "commit": "5458026c216f77a3da68ea49152dc33470cfe2cb",
+        "sourcePaths": [
+          "open-sse/config/providers/registry/minimax/cn/index.ts",
+          "src/shared/constants/providers/apikey/index.ts"
+        ]
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "WS-13b §1 admission evidence is PINNED-UPSTREAM, not a fetched vendor document: the citation names the vendor site OmniRoute records, and the api-key path is attested by the pinned RegistryEntry rather than by a doc this lane read",
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "China-region platform: keys are issued by and bound to the China account system and are not interchangeable with the international platform's keys.",
+          "Host migrated from api.minimaxi.com to api.minimax.cn per the current docs; a key issued before the migration has not been probed against the new host."
+        ]
+      },
+      "scope": "llm",
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://www.minimaxi.com — the vendor's own site as OmniRoute's product catalog records it at 5458026 (src/shared/constants/providers/apikey/*.ts, blob-pinned in extraction-manifest.json). The api-key path is attested by the pinned RegistryEntry: authType \"apikey\", OpenAI Chat Completions at https://api.minimaxi.com/v1. EVIDENCE TIER: pinned-upstream, NOT a page this lane fetched and read — see PROVENANCE.md \"Two tiers of admission citation\".",
+        "tier": "pinned-upstream"
+      },
+      "$comment": "2026-09-25 refresh: MiniMax's China docs moved from platform.minimaxi.com (now a 302 to platform.minimax.cn) and the current OpenAI reference names `https://api.minimax.cn/v1`; the old `api.minimaxi.com` host survives only in older material (the MiniMax CLI README). Endpoint switched to the documented host. FIRST KEY ONLY (R-FW-3): the refresh fetched and read https://platform.minimax.cn/docs/api-reference/text-openai-api — MiniMax China OpenAI-compatible API reference, retrieved 2026-09-25: base `https://api.minimax.cn/v1` with an `sk-api-` key; CNY per-token billing (https://platform.minimax.cn/docs/guides/pricing-paygo). — the tier stays `pinned-upstream` until a live-gate pass supplies the second key."
+    },
+    {
+      "id": "mistral",
+      "displayName": "Mistral",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.mistral.ai/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "OmniRoute",
+        "commit": "5458026c216f77a3da68ea49152dc33470cfe2cb",
+        "sourcePaths": [
+          "open-sse/config/providers/registry/mistral/index.ts",
+          "src/shared/constants/providers/apikey/index.ts"
+        ]
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "WS-13b §1 admission evidence is PINNED-UPSTREAM, not a fetched vendor document: the citation names the vendor site OmniRoute records, and the api-key path is attested by the pinned RegistryEntry rather than by a doc this lane read",
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint."
+        ]
+      },
+      "scope": "llm",
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://mistral.ai — the vendor's own site as OmniRoute's product catalog records it at 5458026 (src/shared/constants/providers/apikey/*.ts, blob-pinned in extraction-manifest.json). The api-key path is attested by the pinned RegistryEntry: authType \"apikey\", OpenAI Chat Completions at https://api.mistral.ai/v1. EVIDENCE TIER: pinned-upstream, NOT a page this lane fetched and read — see PROVENANCE.md \"Two tiers of admission citation\".",
+        "tier": "pinned-upstream"
+      },
+      "$comment": "2026-09-25 refresh: re-admitted on Mistral's fetched docs (tier fetched-document). Devstral and Magistral are retired (folded into Mistral Medium 3.5 / Small 4). Codestral's old dedicated host (codestral.mistral.ai) still answers but current docs show only the metered La Plateforme FIM route, so it is not a separate provider. FIRST KEY ONLY (R-FW-3): the refresh fetched and read https://docs.mistral.ai/models and https://docs.mistral.ai/inference/pricing — Mistral La Plateforme model and pricing documentation, retrieved 2026-09-25: OpenAI-style Chat Completions at `https://api.mistral.ai/v1` with a La Plateforme API key; per-token billing. — the tier stays `pinned-upstream` until a live-gate pass supplies the second key."
+    },
+    {
+      "id": "meta-llama",
+      "displayName": "Meta Llama API",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.llama.com/compat/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "OmniRoute",
+        "commit": "5458026c216f77a3da68ea49152dc33470cfe2cb",
+        "sourcePaths": [
+          "open-sse/config/providers/registry/meta-llama/index.ts",
+          "src/shared/constants/providers/apikey/index.ts"
+        ]
+      },
+      "risk": {
+        "class": "blocked",
+        "reasons": [
+          "The Llama API public preview at api.llama.com was retired on 2026-07-06; Meta's first-party hosted API is now the Meta Model API (`meta`)."
+        ]
+      },
+      "scope": "llm",
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://llama.developer.meta.com — the vendor's own site as OmniRoute's product catalog records it at 5458026 (src/shared/constants/providers/apikey/*.ts, blob-pinned in extraction-manifest.json). The api-key path is attested by the pinned RegistryEntry: authType \"apikey\", OpenAI Chat Completions at https://api.llama.com/compat/v1. EVIDENCE TIER: pinned-upstream, NOT a page this lane fetched and read — see PROVENANCE.md \"Two tiers of admission citation\".",
+        "tier": "pinned-upstream"
+      },
+      "$comment": "2026-09-25 refresh: the Llama API public preview (api.llama.com) was retired on 2026-07-06 (Meta's notice route https://dev.meta.ai/docs/llama-api-deprecation/, corroborated by https://github.com/deepset-ai/haystack-core-integrations/issues/3544 and https://www.promptfoo.dev/docs/providers/llamaApi/). Blocked; successor is `meta` (Meta Model API)."
+    },
+    {
+      "id": "nvidia",
+      "displayName": "NVIDIA NIM",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://integrate.api.nvidia.com/v1"
+      },
+      "modelDiscovery": "openai-models",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "OmniRoute",
+        "commit": "5458026c216f77a3da68ea49152dc33470cfe2cb",
+        "sourcePaths": [
+          "open-sse/config/providers/registry/nvidia/index.ts",
+          "src/shared/constants/providers/apikey/index.ts"
+        ]
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "WS-13b §1 admission evidence is PINNED-UPSTREAM, not a fetched vendor document: the citation names the vendor site OmniRoute records, and the api-key path is attested by the pinned RegistryEntry rather than by a doc this lane read",
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint.",
+          "Hosted endpoints are trial/rate-limited developer access; NVIDIA marks some model endpoints deprecated without a date."
+        ]
+      },
+      "scope": "llm",
+      "pricingBasis": "free",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://build.nvidia.com — the vendor's own site as OmniRoute's product catalog records it at 5458026 (src/shared/constants/providers/apikey/*.ts, blob-pinned in extraction-manifest.json). The api-key path is attested by the pinned RegistryEntry: authType \"apikey\", OpenAI Chat Completions at https://integrate.api.nvidia.com/v1. EVIDENCE TIER: pinned-upstream, NOT a page this lane fetched and read — see PROVENANCE.md \"Two tiers of admission citation\".",
+        "tier": "pinned-upstream"
+      },
+      "$comment": "2026-09-25 refresh: re-admitted on NVIDIA's fetched build.nvidia.com model pages (tier fetched-document). `free` stays correct: NVIDIA retired the credit system for build.nvidia.com in favour of per-model trial rate limits (staff reply 2025-09-10, https://forums.developer.nvidia.com/t/request-more-4-000-credits-option-on-build-nvidia-com/344567) and publishes no per-token price table. Several Nemotron pages mark their Free Endpoint DEPRECATED with no shutdown date. FIRST KEY ONLY (R-FW-3): the refresh fetched and read https://build.nvidia.com/nvidia/nemotron-3-super-120b-a12b — NVIDIA build.nvidia.com model page, retrieved 2026-09-25: OpenAI-compatible Chat Completions at `https://integrate.api.nvidia.com/v1` with an NVIDIA API key; Free Endpoint (rate-limited trial access, no per-token billing). — the tier stays `pinned-upstream` until a live-gate pass supplies the second key."
+    },
+    {
+      "id": "cohere",
+      "displayName": "Cohere",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.cohere.ai/compatibility/v1",
+        "models": "https://api.cohere.ai/compatibility/v1/models"
+      },
+      "modelDiscovery": "openai-models",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "OmniRoute",
+        "commit": "5458026c216f77a3da68ea49152dc33470cfe2cb",
+        "sourcePaths": [
+          "open-sse/config/providers/registry/cohere/index.ts",
+          "src/shared/constants/providers/apikey/index.ts"
+        ]
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "WS-13b §1 admission evidence is PINNED-UPSTREAM, not a fetched vendor document: the citation names the vendor site OmniRoute records, and the api-key path is attested by the pinned RegistryEntry rather than by a doc this lane read",
+          "Admitted on the vendor's own fetched documentation; this application has not run a live key gate against this endpoint."
+        ]
+      },
+      "scope": "llm",
+      "pricingBasis": "token",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://cohere.com — the vendor's own site as OmniRoute's product catalog records it at 5458026 (src/shared/constants/providers/apikey/*.ts, blob-pinned in extraction-manifest.json). The api-key path is attested by the pinned RegistryEntry: authType \"apikey\", OpenAI Chat Completions at https://api.cohere.com/compatibility/v1. EVIDENCE TIER: pinned-upstream, NOT a page this lane fetched and read — see PROVENANCE.md \"Two tiers of admission citation\".",
+        "tier": "pinned-upstream"
+      },
+      "$comment": "2026-09-25 refresh: re-admitted on Cohere's fetched Compatibility API doc (tier fetched-document), whose documented base is `https://api.cohere.ai/compatibility/v1` (the extracted row used the api.cohere.com spelling of the same Cohere API). Cohere's native v2 Chat API is a different dialect and is not modelled. Command A+, A Reasoning and A Vision are documented as free up to rate limits, so they carry no per-token price. FIRST KEY ONLY (R-FW-3): the refresh fetched and read https://docs.cohere.com/v2/docs/compatibility-api — Cohere Compatibility API documentation, retrieved 2026-09-25: OpenAI-compatible Chat Completions at `https://api.cohere.ai/compatibility/v1` with a Cohere API key; per-token billing per https://cohere.com/pricing. — the tier stays `pinned-upstream` until a live-gate pass supplies the second key."
     }
   ]
 }