Merge origin/master (v0.5.81) into gitea/new_feature

Resolve conflicts: - streamingHandler.js: merge buildStreamErrorBytes onAbortTerminal + local shouldPersistRequestDetail & streamStatusForContent - capabilities.js: preserve server-injected user-asserted caps and models.dev catalog lookup; wire CommandCode /alpha/generate caps inside resolve() - commandcode.js (services/usage): adopt upstream whoami + billing credits/subscriptions with 5h/weekly rate windows and plan caps - openai-to-commandcode.js: merge toNativeImageBlock (data-URI & http(s) support) and assistant reasoning_content preservation - commandcode-to-openai.js: adopt upstream mid-stream error throw for clean retry and abortion - tests: sync commandcode test suite and exclude .next from vitest config
2026-09-22 10:08:58 +07:00
parent 28e26f4295 a8c9d3802c
commit 11089ab137
95 changed files with 4059 additions and 1016 deletions
--- a/open-sse/providers/capabilities.js
+++ b/open-sse/providers/capabilities.js
@@ -116,6 +116,16 @@ export const MODEL_CAPABILITIES = {
  // DeepSeek's first V4 model with image input; text limits match V4-Flash.
  "deepseek-v4-flash-vision-exp": { vision: true, reasoning: true, thinkingFormat: "deepseek", contextWindow: 1000000, maxOutput: 384000 },

+  // DeepSeek V4.1-Flash is natively multimodal — models.dev lists
+  // opencode-go/deepseek-v4.1-flash with modalities.input ["text","image"] — and upstream
+  // the retired v4-flash / vision-exp ids route to it, so the live V4.1 ids carry the
+  // same image capability as the exp id above. "deepseek-flash" is the GA id on the
+  // DeepSeek API; it previously fell through to the generic *deepseek* pattern, whose
+  // 128K/64K limits are kept here. The repeated fields are deliberate: an exact entry
+  // short-circuits the pattern table, so a vision-only delta would drop them.
+  "deepseek-v4.1-flash": { vision: true, reasoning: true, thinkingFormat: "deepseek", contextWindow: 1000000, maxOutput: 384000 },
+  "deepseek-flash":      { vision: true, reasoning: true, thinkingFormat: "deepseek", contextWindow: 128000, maxOutput: 64000 },
+
  // Qwen plain coder/text (no vision) — registry "vision-model" / "coder-model" aliases
  "vision-model":      { vision: true, reasoning: true, thinkingFormat: "qwen", contextWindow: 1000000 },
  "coder-model":       { reasoning: true, thinkingFormat: "qwen", contextWindow: 1000000 },
@@ -131,6 +141,8 @@ export const MODEL_CAPABILITIES = {
  // via OpenAI Responses input_image; reasoning supports up to xhigh.
  "muse-spark-1.2-contributor-free": { vision: true, reasoning: true, thinkingFormat: "openai", contextWindow: 1048576, maxOutput: 131072 },
  "muse-spark-1.3-contributor-free": { vision: true, reasoning: true, thinkingFormat: "openai", contextWindow: 1048576, maxOutput: 131072 },
+  // OpenCode Free Union Alpha — multimodal (text+vision), 262K context, 131K max output
+  "union-alpha": { vision: true, contextWindow: 262144, maxOutput: 131072 },
 };

 const KIRO_GPT_5_6_CAPABILITIES = { vision: true, reasoning: true, search: true, thinkingFormat: "openai", contextWindow: 272000, maxOutput: 128000 };
@@ -214,6 +226,13 @@ export const PROVIDER_CAPABILITIES = {
    // contract). maxOutput 128000 per the server's product-config payload.
    "deepseek-v4.1-flash": { vision: true, reasoning: true, thinkingFormat: "openai", thinkingCanDisable: true, contextWindow: 1000000, maxOutput: 128000 },
  },
+  // CodeBuddy intl — same gateway catalog as CN, so deepseek-v4.1-flash mirrors
+  // the codebuddy-cn entry (the openai-style reasoning_effort format matters:
+  // the generic *deepseek-v4* pattern would otherwise pick the vendor-native
+  // "deepseek" thinking shape, which the CodeBuddy gateway does not accept).
+  "codebuddy-intl": {
+    "deepseek-v4.1-flash": { vision: true, reasoning: true, thinkingFormat: "openai", thinkingCanDisable: true, contextWindow: 1000000, maxOutput: 128000 },
+  },
  // Qoder — upstream exposes opaque internal ids (dfmodel, kmodel, …); the
  // registry `name` is display-only and capability lookup matches on the raw
  // id, so every qoder model would fall through to DEFAULT_CAPABILITIES
@@ -251,6 +270,16 @@ export const PROVIDER_CAPABILITIES = {
    "laguna-s-2.1":  { reasoning: true, thinkingFormat: "openai", contextWindow: 1000000, maxOutput: 32000 },
    "laguna-xs-2.1": { reasoning: true, thinkingFormat: "openai", contextWindow: 200000, maxOutput: 32000 },
  },
+  // Ollama Cloud — the generic *deepseek-v4* pattern misses the vision badge
+  // the library page publishes for this model (text+image in, 1M context).
+  // ponytail: thinkingFormat stays "deepseek" to preserve today's body shape;
+  // Ollama's native toggle is the top-level `think` field (bool or
+  // low/medium/high/max), which no format in thinkingUnified.js emits yet —
+  // openai-to-ollama.js drops it. Wire a "think" format when thinking on
+  // Ollama Cloud is actually needed.
+  "ollama": {
+    "deepseek-v4.1-flash:cloud": { vision: true, reasoning: true, thinkingFormat: "deepseek", contextWindow: 1000000, maxOutput: 384000 },
+  },
 };

 /**
@@ -349,7 +378,11 @@ export const PATTERN_CAPABILITIES = [
  { pattern: "*glm*",           caps: { reasoning: true, thinkingFormat: "zai", contextWindow: 200000 } },

  // ── DeepSeek (thinking.enabled + reasoning_effort; r1 = thinking-only) ─
-  { pattern: "*deepseek-v4*",   caps: { reasoning: true, thinkingFormat: "deepseek", contextWindow: 1000000, maxOutput: 384000 } },
+  // v4.1+ has real image input (probed live on Alibaba MaaS: correct color
+  // read from a PNG). v4-pro / v4-flash-0731 accept image blocks but ignore
+  // them (answered "Unknown"), so vision stays scoped to v4.* dotted releases.
+  { pattern: "*deepseek-v4.*",  caps: { vision: true, reasoning: true, thinkingFormat: "deepseek", thinkingEffortSupported: true, contextWindow: 1000000, maxOutput: 128000 } },
+  { pattern: "*deepseek-v4*",   caps: { reasoning: true, thinkingFormat: "deepseek", thinkingEffortSupported: true, contextWindow: 1000000, maxOutput: 384000 } },
  { pattern: "*reasoner*",      caps: { reasoning: true, thinkingFormat: "deepseek", thinkingCanDisable: false, contextWindow: 128000 } },
  { pattern: "*deepseek-r*",    caps: { reasoning: true, thinkingFormat: "deepseek", thinkingCanDisable: false, contextWindow: 128000 } },
  { pattern: "*deepseek-chat*", caps: { contextWindow: 128000 } },
@@ -425,6 +458,7 @@ const MODALITY_KEYS = ["vision", "pdf", "audioInput", "videoInput"];
 // globalThis, which IS shared across server bundles in the same process.
 // Same reason the browser bundle is safe: it never calls a setter, so the slots
 // stay empty and every consumer below short-circuits.
+let catalogSource = null;
 const SOURCE_SLOTS = (globalThis.__9R_CAPABILITY_SOURCES ||= {
  catalog: null, // { getModalities, getLimits } — synced models.dev catalog
  userCaps: null, // (provider, model) => asserted caps — dashboard toggles
@@ -432,10 +466,20 @@ const SOURCE_SLOTS = (globalThis.__9R_CAPABILITY_SOURCES ||= {

 /**
 * Install the synced catalog reader (server only).
- * @param {{ getModalities: Function, getLimits: Function } | null} source
+ * @param {{ getModalities: (provider: string, model: string) => object|null,
+ *           getLimits: (provider: string, model: string) => object|null } | null} source
 */
 export function setCatalogSource(source) {
+  catalogSource = source || null;
  SOURCE_SLOTS.catalog = source || null;
+  if (typeof globalThis !== "undefined") globalThis.__9rCatalogSource = source || null;
+}
+
+function getCatalogSource() {
+  if (catalogSource) return catalogSource;
+  if (SOURCE_SLOTS.catalog) return (catalogSource = SOURCE_SLOTS.catalog);
+  if (typeof globalThis === "undefined") return null;
+  return (catalogSource = globalThis.__9rCatalogSource || null);
 }

 // Capabilities the user asserted per provider+model (dashboard "Add/Edit Model"
@@ -478,16 +522,17 @@ function applyUserCaps(result, provider, model) {
 // flips when an outside source positively declares support.
 function refine(base, provider, model) {
  const result = { ...DEFAULT_CAPABILITIES, ...base };
-  const catalogSource = SOURCE_SLOTS.catalog;
-  if (catalogSource) {
-    const modalities = catalogSource.getModalities(model);
+
+  const source = getCatalogSource();
+  if (source) {
+    const modalities = source.getModalities(provider, model);
    if (modalities) {
      for (const key of MODALITY_KEYS) {
        if (modalities[key] === true) result[key] = true;
      }
    }

-    const limits = catalogSource.getLimits(provider, model);
+    const limits = source.getLimits(provider, model);
    if (limits) {
      if (limits.contextWindow > 0) result.contextWindow = limits.contextWindow;
      if (limits.maxOutput > 0) result.maxOutput = limits.maxOutput;
@@ -499,12 +544,67 @@ function refine(base, provider, model) {
  return result;
 }

+// Mirrors Command Code CLI `isKnownTextOnlyModel` (no image input). New models
+// default to vision; only this denylist stays text-only.
+const COMMANDCODE_TEXT_ONLY = new Set([
+  "deepseek/deepseek-v4-pro",
+  "deepseek/deepseek-v4-flash",
+  "deepseek/deepseek-v4-flash-fast",
+  "zai-org/glm-5.3",
+  "zai-org/glm-5.2",
+  "zai-org/glm-5.2-fast",
+  "zai-org/glm-5.1",
+  "zai-org/glm-5",
+  "minimaxai/minimax-m2.7",
+  "minimax/minimax-m2.7-free",
+  "minimaxai/minimax-m2.5",
+  "xiaomi/mimo-v2.5-pro",
+  "qwen/qwen3.6-max-preview",
+  "qwen/qwen3.7-max",
+  "meituan/longcat-2.0:free",
+  "stepfun/step-3.5-flash",
+  "tencent/hy4-preview",
+  "tencent/hy3",
+  "tencent/hy3-paid",
+  "nvidia/nemotron-3-ultra-550b-a55b",
+  "poolside/laguna-s-2.1-free",
+  "inclusionai/ling-3.0-flash-free",
+  "inclusionai/ling-3.0-flash-sante:free",
+]);
+
+function isCommandCodeTextOnly(model) {
+  const key = String(model || "").toLowerCase();
+  if (COMMANDCODE_TEXT_ONLY.has(key)) return true;
+  for (const id of COMMANDCODE_TEXT_ONLY) {
+    const base = id.includes("/") ? id.slice(id.lastIndexOf("/") + 1) : id;
+    if (key === base || key.endsWith("/" + base)) return true;
+  }
+  return false;
+}
+
 export function getCapabilitiesForModel(provider, model) {
  if (!model) return { ...DEFAULT_CAPABILITIES };

  // Canonical exact lookup strips vendor prefix: "anthropic/claude-opus-4.7" -> "claude-opus-4.7".
  const baseModel = model.includes("/") ? model.split("/").pop() : model;
  const resolve = () => {
+    // CommandCode wire is /alpha/generate for every model. Family patterns
+    // (deepseek-v4 → thinkingFormat:deepseek, vision:false) must not win here.
+    if (provider === "commandcode" || provider === "cmc") {
+      const providerCaps = PROVIDER_CAPABILITIES.commandcode;
+      if (providerCaps?.[model]) return { ...DEFAULT_CAPABILITIES, ...providerCaps[model] };
+      if (providerCaps?.[baseModel]) return { ...DEFAULT_CAPABILITIES, ...providerCaps[baseModel] };
+      return {
+        ...DEFAULT_CAPABILITIES,
+        reasoning: true,
+        thinkingFormat: "commandcode",
+        thinkingEffortSupported: true,
+        vision: !isCommandCodeTextOnly(model),
+        contextWindow: 1000000,
+        maxOutput: 384000,
+      };
+    }
+
    // 1. Provider-specific override
    if (provider) {
      const providerCaps = PROVIDER_CAPABILITIES[provider];
--- a/open-sse/providers/catalogOverride.js
+++ b/open-sse/providers/catalogOverride.js
@@ -13,6 +13,11 @@ export const CATALOG_FILE = path.join(DATA_DIR, "model-catalog.json");
 // Trimmed upstream catalog, read by the add-models skill (not by the router).
 export const CATALOG_RAW_FILE = path.join(DATA_DIR, "model-catalog-raw.json");

+// Schema of the file this module reads. The writer stamps it; a file carrying an
+// older value predates provider-scoped modality keys, and its flat keys are not
+// looked up here, so the sync rebuilds it instead of asking upstream for a 304.
+export const CATALOG_VERSION = 2;
+
 const EMPTY = { models: {}, providers: {} };
 let cache = EMPTY;
 let cachedMtime = -1;
@@ -45,14 +50,19 @@ function load() {
  return cache;
 }

-// Modality is a property of the model itself — any gateway serving it inherits
-// the same image/video/pdf support, so this is keyed by model id alone.
-export function getCatalogModalities(model) {
-  return load().models[baseId(model)] || null;
+// Modalities are recorded per gateway upstream, and gateways disagree about the
+// same weights — some do not proxy images at all — so the key is provider +
+// model, in the local provider id space, exactly like the limits below. Keying
+// by model id alone made short ids collide across vendors: "auto", "free" and
+// "efficient" are router modes in one catalog and model names in another, and a
+// request to the router mode inherited a stranger's vision.
+export function getCatalogModalities(provider, model) {
+  if (!provider) return null;
+  return load().models[`${provider}:${baseId(model)}`] || null;
 }

-// Context and output limits are a property of the gateway, not the model: each
-// one truncates differently, so these stay keyed by provider + model.
+// Context and output limits are a property of the gateway too: each one
+// truncates differently, so these stay keyed by provider + model.
 export function getCatalogLimits(provider, model) {
  const byProvider = provider && load().providers[provider];
  if (!byProvider) return null;
--- a/open-sse/providers/pricing.js
+++ b/open-sse/providers/pricing.js
@@ -111,6 +111,8 @@ export const MODEL_PRICING = {
  "deepseek-v3.2-chat":           { input: 0.14,  output: 0.28,  cached: 0.0028, reasoning: 0.28,   cache_creation: 0.14  },
  "deepseek-v3.2-reasoner":       { input: 0.14,  output: 0.28,  cached: 0.0028, reasoning: 0.28,   cache_creation: 0.14  },
  "deepseek-v4-flash":            { input: 0.14,  output: 0.28,  cached: 0.0028, reasoning: 0.28,   cache_creation: 0.14  },
+  "deepseek-v4.1-flash":          { input: 0.14,  output: 0.28,  cached: 0.0028, reasoning: 0.28,   cache_creation: 0.14  },
+  "deepseek-flash":               { input: 0.14,  output: 0.28,  cached: 0.0028, reasoning: 0.28,   cache_creation: 0.14  },
  "deepseek-v4-pro":              { input: 0.435, output: 0.87,  cached: 0.003625, reasoning: 0.87,  cache_creation: 0.435 },

  // === GLM ===
--- a/open-sse/providers/registry/codebuddy-intl.js
+++ b/open-sse/providers/registry/codebuddy-intl.js
@@ -58,7 +58,9 @@ export default {
    { id: "kimi-k2.5", name: "Kimi-K2.5" },
    { id: "hy3-preview", name: "Hy3 Preview" },
    { id: "deepseek-v4-pro", name: "DeepSeek-V4-Pro" },
-    { id: "deepseek-v4-flash", name: "DeepSeek-V4-Flash" },
+    // deepseek-v4-flash replaced server-side by deepseek-v4.1-flash (same
+    // catalog as CN; the old endpoint still answers 200 but the list is the contract).
+    { id: "deepseek-v4.1-flash", name: "DeepSeek-V4.1-Flash" },
    { id: "deepseek-v3-2-volc", name: "DeepSeek-V3.2" },
  ],
  oauth: {
--- a/open-sse/providers/registry/codex.js
+++ b/open-sse/providers/registry/codex.js
@@ -65,6 +65,9 @@ export default {
    { id: "gpt-5.4-mini-review", name: "GPT 5.4 Mini Review", upstreamModelId: "gpt-5.4-mini", quotaFamily: "review" },
    { id: "gpt-5.3-codex-spark", name: "GPT 5.3 Codex Spark" },
    { id: "gpt-5.3-codex-spark-review", name: "GPT 5.3 Codex Spark Review", upstreamModelId: "gpt-5.3-codex-spark", quotaFamily: "review" },
+    // Codex CLI's auto-review virtual model. Unlike the "-review" variants above it is not derived
+    // from a base model, so it is forwarded verbatim instead of having "-review" stripped (#1398).
+    { id: "codex-auto-review", name: "Codex Auto Review", upstreamModelId: "codex-auto-review", quotaFamily: "review" },
    { id: "gpt-image-2.5", name: "GPT Image 2.5", capabilities: ["text2img","edit","multiImage"], params: ["size","quality","background","image_detail","output_format"], kind: "image" },
    { id: "gpt-image-2.5-flare", name: "GPT Image 2.5 Flare", capabilities: ["text2img","edit","multiImage"], params: ["size","quality","background","image_detail","output_format"], kind: "image" },
    { id: "gpt-image-2.5-sunburst", name: "GPT Image 2.5 Sunburst", capabilities: ["text2img","edit","multiImage"], params: ["size","quality","background","image_detail","output_format"], kind: "image" },
--- a/open-sse/providers/registry/commandcode.js
+++ b/open-sse/providers/registry/commandcode.js
@@ -74,4 +74,8 @@ export default {
    { id: "claude-opus-4-7", name: "Claude Opus 4.7" },
    { id: "claude-haiku-4-5", name: "Claude Haiku 4.5" },
  ],
+  features: {
+    usage: true,
+    usageApikey: true,
+  },
 };
--- a/open-sse/providers/registry/deepseek.js
+++ b/open-sse/providers/registry/deepseek.js
@@ -59,6 +59,7 @@ export default {
    { id: "deepseek-v4-pro", name: "DeepSeek V4 Pro" },
    { id: "deepseek-v4-pro-max", name: "DeepSeek V4 Pro Max", upstreamModelId: "deepseek-v4-pro" },
    { id: "deepseek-v4-pro-none", name: "DeepSeek V4 Pro No Thinking", upstreamModelId: "deepseek-v4-pro" },
+    { id: "deepseek-v4.1-flash", name: "DeepSeek V4.1 Flash" },
    { id: "deepseek-v4-flash", name: "DeepSeek V4 Flash" },
    { id: "deepseek-v4-flash-vision-exp", name: "DeepSeek V4 Flash Vision (Exp)" },
    { id: "deepseek-chat", name: "DeepSeek V3.2 Chat" },
--- a/open-sse/providers/registry/ollama.js
+++ b/open-sse/providers/registry/ollama.js
@@ -30,6 +30,7 @@ export default {
    { id: "glm-4.7-flash", name: "GLM 4.7 Flash" },
    { id: "qwen3.5", name: "Qwen3.5" },
    { id: "minimax-m3", name: "MiniMax M3" },
+    { id: "deepseek-v4.1-flash:cloud", name: "DeepSeek V4.1 Flash" },
  ],
  serviceKinds: ["llm", "webFetch"],
  fetchConfig: {
--- a/open-sse/providers/registry/opencode-go.js
+++ b/open-sse/providers/registry/opencode-go.js
@@ -13,7 +13,7 @@ export default {
    textIcon: "OC",
    website: "https://opencode.ai/auth",
    notice: {
-      text: "OpenCode Go subscription: $5/mo (then  0/mo). Access to Kimi, GLM, Qwen, MiMo, MiniMax models.",
+      text: "OpenCode Go subscription: $5/mo (then 10/mo). Access to Kimi, GLM, Qwen, MiMo, MiniMax models.",
      apiKeyUrl: "https://opencode.ai/auth",
    },
  },
--- a/open-sse/providers/registry/opencode.js
+++ b/open-sse/providers/registry/opencode.js
@@ -17,13 +17,17 @@ export default {
    headers: {
      "x-opencode-client": "desktop",
    },
+    forceStream: true,
    noAuth: true,
+    quirks: {
+      forceAutoToolChoiceModels: ["muse-spark-1.3-contributor-free"],
+    },
  },
  models: [
-    // Muse Spark models are served by /zen/v1/responses; the rest stay on
-    // /chat/completions, so the format is declared per-model, not per-provider.
+    // Endpoint formats differ per model, so declare non-chat models explicitly.
    { id: "muse-spark-1.2-contributor-free", name: "Muse Spark 1.2 Contributor Free", targetFormat: "openai-responses" },
    { id: "muse-spark-1.3-contributor-free", name: "Muse Spark 1.3 Contributor Free", targetFormat: "openai-responses" },
+    { id: "union-alpha", name: "Union Alpha Free", targetFormat: "claude" },
  ],
  modelsFetcher: { url: "https://opencode.ai/zen/v1/models", type: "opencode-free" },
  passthroughModels: true,
--- a/open-sse/providers/registry/zed.js
+++ b/open-sse/providers/registry/zed.js
@@ -1,10 +1,9 @@
 // Zed provider — RSA keypair callback auth (NOT standard OAuth).
 export default {
  id: "zed",
-  priority: 10,
+  priority: 999,
  alias: "zd",
  uiAlias: "zd",
-  hidden: true,
  display: {
    name: "Zed",
    icon: "code",
--- a/open-sse/providers/shared.js
+++ b/open-sse/providers/shared.js
@@ -62,8 +62,17 @@ const ANTHROPIC_BETA_BASE = [
 const ANTHROPIC_BETA_HEAVY_AGENT = ["advanced-tool-use-2025-11-20", "effort-2025-11-24"];

 // Heavy-agent beta flags are gated to opus/sonnet — cheaper models don't need them.
-export function selectAnthropicBeta(model = "") {
-  const flags = [...ANTHROPIC_BETA_BASE];
+// `redact-thinking` asks Anthropic to return signature-only thinking blocks, which
+// is right for clients that never render thinking but blanks the summaries a
+// client explicitly requested with `thinking.display: "summarized"`.
+const ANTHROPIC_BETA_REDACT_THINKING = "redact-thinking-2026-02-12";
+
+export function wantsThinkingSummaries(body) {
+  return body?.thinking?.display === "summarized";
+}
+
+export function selectAnthropicBeta(model = "", body = null) {
+  const flags = ANTHROPIC_BETA_BASE.filter((flag) => flag !== ANTHROPIC_BETA_REDACT_THINKING || !wantsThinkingSummaries(body));
  if (/^claude-(opus|sonnet)/.test(model)) flags.push(...ANTHROPIC_BETA_HEAVY_AGENT);
  return flags.join(",");
 }
--- a/open-sse/providers/thinkingLevels.js
+++ b/open-sse/providers/thinkingLevels.js
@@ -26,6 +26,7 @@ const FORMAT_LEVELS = {
  qwen: L.base,
  kimi: L.levelMax,
  deepseek: L.hiMax,
+  commandcode: ["none", "low", "medium", "high", "xhigh", "max"],
  minimax: L.onOff,
  hunyuan: L.base,
  step: L.base,
@@ -40,6 +41,10 @@ const PATTERN_THINKING = [
  { provider: "codex", pattern: "*gpt-5.6-terra*", levels: [...CODEX_GPT_5_6_LEVELS, "ultra"] },
  { provider: "codex", pattern: "*gpt-5.6-luna*", levels: CODEX_GPT_5_6_LEVELS },
  { pattern: "*codex*", levels: ["low", "medium", "high", "xhigh"] }, // codex cannot disable thinking
+  // DeepSeek v4.* (Alibaba MaaS, probed live): effort low|medium|high|xhigh|max
+  // all 200 via output_config.effort; "none" is a 400 on the anthropic route
+  // (disable thinking instead). none kept for the picker = disable.
+  { pattern: "*deepseek-v4.*", levels: ["none", "low", "medium", "high", "xhigh", "max"] },
  // codebuddy-cn per-model effort sets — the server's product-config payload
  // publishes `reasoning.supportedEfforts` per model. NOTE: the chat endpoint
  // accepts any level you send (probed none/minimal/low/medium/high/xhigh/max
@@ -52,6 +57,8 @@ const PATTERN_THINKING = [
  { provider: "codebuddy-cn", pattern: "deepseek-v4*", levels: ["low", "high", "xhigh"] },
  { provider: "codebuddy-cn", pattern: "hy3*",         levels: ["low", "high"] },
  { provider: "codebuddy-cn", pattern: "hy4*",         levels: ["high"] },
+  // codebuddy-intl rides the same gateway catalog, so its deepseek levels match.
+  { provider: "codebuddy-intl", pattern: "deepseek-v4*", levels: ["low", "high", "xhigh"] },
 ];

 // The generic level set used when a model's thinking format is unknown. Exported