Merge origin/master (v0.5.91) into gitea/new_feature

Resolve conflicts:
- streamingHandler.js: adopt upstreamResponseHeaders while keeping 0-token detail row avoidance
- capabilities.js: preserve user-asserted caps and globalThis slots without local caching of catalogSource
- AddCustomModelModal.js & providers/[id]/page.js: wire STT transport marker with custom model edits/assertions
- models/custom/route.js & aliasRepo.js: persist custom model transport and invalidate user caps
- usageRepo.js: key byApiKey live stats by full API key and keep tail in maskApiKey
- UsageStats.js: lazy load charts dynamically

open-sse/providers/capabilities.js

@@ -430,21 +430,29 @@ const TRUST_UPSTREAM_VISION = new Set(["openrouter"]);
  *
  * @param {string[]} comboModels
  * @param {Object|null} [comboLookup] optional map of combo name → models array for nested resolution
+ * @param {Function|null} [resolveCaps] optional (fullId) → caps override. The synced model
+ *   catalog is server-only (it reads a file), so a browser-side resolution cannot see the
+ *   limits it supplies and silently falls back to the generic patterns below. Callers that
+ *   have the server's answer (/api/models, via useModelCaps) pass it here; it is merged over
+ *   the local tables, so fields it does not carry (tools, pdf, audio/video, thinking*) survive.
  * @param {number} [_depth] internal recursion depth guard
  * @returns {object|null} full capabilities object, or null for empty input
  */
-export function aggregateComboCapabilities(comboModels, comboLookup = null, _depth = 0) {
+export function aggregateComboCapabilities(comboModels, comboLookup = null, resolveCaps = null, _depth = 0) {
   if (!comboModels?.length || _depth > 6) return null;
   const allCaps = comboModels.map((fullId) => {
     // Nested combo: bare name (no slash) that exists in the lookup — recurse
     if (!fullId.includes("/") && comboLookup?.[fullId]) {
-      return aggregateComboCapabilities(comboLookup[fullId], comboLookup, _depth + 1)
+      return aggregateComboCapabilities(comboLookup[fullId], comboLookup, resolveCaps, _depth + 1)
+          ?? resolveCaps?.(fullId)
           ?? getCapabilitiesForModel(null, fullId);
     }
     const slash = fullId.indexOf("/");
     const provider = slash === -1 ? null : fullId.slice(0, slash);
     const model = slash === -1 ? fullId : fullId.slice(slash + 1);
-    return getCapabilitiesForModel(provider, model);
+    const local = getCapabilitiesForModel(provider, model);
+    const override = resolveCaps?.(fullId);
+    return override ? { ...local, ...override } : local;
   });
   const first = allCaps[0];
   return {
@@ -482,7 +490,9 @@ const MODALITY_KEYS = ["vision", "pdf", "audioInput", "videoInput"];
 // handlers (silently: the setters still "succeed"). The slots therefore live on
 // 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.
+// stay empty and every consumer below short-circuits. Every read goes through
+// globalThis: caching it locally would keep a reader alive in other copies after
+// setCatalogSource(null).
 let catalogSource = null;
 const SOURCE_SLOTS = (globalThis.__9R_CAPABILITY_SOURCES ||= {
   catalog: null, // { getModalities, getLimits } — synced models.dev catalog
@@ -496,15 +506,13 @@ const SOURCE_SLOTS = (globalThis.__9R_CAPABILITY_SOURCES ||= {
  */
 export function setCatalogSource(source) {
   catalogSource = source || null;
-  SOURCE_SLOTS.catalog = source || null;
+  if (SOURCE_SLOTS) 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);
+  if (typeof globalThis === "undefined") return catalogSource;
+  return SOURCE_SLOTS?.catalog || globalThis.__9rCatalogSource || null;
 }
 
 // Capabilities the user asserted per provider+model (dashboard "Add/Edit Model"

open-sse/providers/models/helpers.js

@@ -1,3 +1,5 @@
+import { FORMATS } from "../../translator/formats.js";
+
 // Codex auto-generates a "-review" variant for each llm model (review quota family)
 export const CODEX_REVIEW_SUFFIX = "-review";
 
@@ -25,3 +27,20 @@ export function isMuseSparkModel(modelId) {
   const base = clean.includes("/") ? clean.split("/").pop() : clean;
   return /^muse[-_]?spark(?:$|[-_:.\s])/i.test(base);
 }
+
+// Endpoint families for OpenCode models outside the curated registry (modelsFetcher /
+// passthrough ids) — regex keeps auto-fetched models on the right endpoint:
+// /responses (gpt/grok/muse-spark), /messages (minimax/qwen), /chat/completions (rest).
+// Curated registry entries always win; this is the unknown-id fallback only.
+const OPENCODE_FAMILIES = [
+  { match: /^(grok|gpt|muse[-_]?spark)/i, supportedFormats: [FORMATS.OPENAI_RESPONSES], targetFormat: FORMATS.OPENAI_RESPONSES },
+  { match: /^deepseek-v4-(pro|flash)/, supportedFormats: [FORMATS.OPENAI, FORMATS.CLAUDE, FORMATS.OPENAI_RESPONSES] },
+  { match: /^(minimax|qwen)/, supportedFormats: [FORMATS.OPENAI, FORMATS.CLAUDE] },
+  { match: /^claude-/i, supportedFormats: [FORMATS.CLAUDE] },
+];
+
+export function opencodeFamilyFormats(modelId) {
+  if (!modelId || typeof modelId !== "string") return null;
+  const base = modelId.replace(/\([^()]+\)\s*$/, "").trim();
+  return OPENCODE_FAMILIES.find((f) => f.match.test(base)) || null;
+}

open-sse/providers/pricing.js

@@ -2,8 +2,28 @@
 //
 // Fallback order (first match wins):
 //   1. PROVIDER_PRICING[provider][model]  — provider-specific override
-//   2. MODEL_PRICING[model]               — canonical model price (provider-agnostic)
-//   3. PATTERN_PRICING                    — glob pattern match (e.g. "codex-*")
+//   2. FREE_MODEL_NAMESPACES               — upstream bills these at $0
+//   3. MODEL_PRICING[model]               — canonical model price (provider-agnostic)
+//   4. PATTERN_PRICING                    — glob pattern match (e.g. "codex-*")
+
+/**
+ * Namespaces upstream meters at $0. A free model must never inherit a paid
+ * rate: the vendor-prefix strip in getPricingForModel() would turn
+ * "cline-free/deepseek-v4.1-flash" into "deepseek-v4.1-flash" and match
+ * MODEL_PRICING, so the namespace is checked before both fallbacks.
+ */
+export const FREE_MODEL_NAMESPACES = ["cline-free/"];
+
+export const ZERO_PRICING = {
+  input: 0, output: 0, cached: 0, reasoning: 0, cache_creation: 0,
+};
+
+/** True when the model id sits in a namespace upstream bills at $0. */
+export function isFreeModel(model) {
+  if (!model) return false;
+  const lower = String(model).toLowerCase();
+  return FREE_MODEL_NAMESPACES.some((ns) => lower.startsWith(ns));
+}
 
 /**
  * Canonical model pricing — provider-agnostic.
@@ -361,10 +381,11 @@ export function matchPattern(pattern, model) {
 }
 
 /**
- * Resolve pricing for a model using the 3-step fallback chain:
+ * Resolve pricing for a model using the 4-step fallback chain:
  *   1. PROVIDER_PRICING[provider][model]
- *   2. MODEL_PRICING[model]
- *   3. PATTERN_PRICING (glob match)
+ *   2. free namespace (upstream bills $0)
+ *   3. MODEL_PRICING[model]
+ *   4. PATTERN_PRICING (glob match)
  *
  * @param {string} provider
  * @param {string} model
@@ -378,12 +399,15 @@ export function getPricingForModel(provider, model) {
     return PROVIDER_PRICING[provider][model];
   }
 
-  // 2. Canonical model pricing (strip vendor prefix if needed: "deepseek/deepseek-chat" → "deepseek-chat")
+  // 2. Free namespaces bill $0 regardless of the model name behind them.
+  if (isFreeModel(model)) return ZERO_PRICING;
+
+  // 3. Canonical model pricing (strip vendor prefix if needed: "deepseek/deepseek-chat" → "deepseek-chat")
   const baseModel = model.includes("/") ? model.split("/").pop() : model;
   if (MODEL_PRICING[baseModel]) return MODEL_PRICING[baseModel];
   if (MODEL_PRICING[model]) return MODEL_PRICING[model];
 
-  // 3. Pattern match
+  // 4. Pattern match
   for (const { pattern, pricing } of PATTERN_PRICING) {
     if (matchPattern(pattern, baseModel) || matchPattern(pattern, model)) {
       return pricing;

open-sse/providers/registry/agnes.js

@@ -0,0 +1,29 @@
+export default {
+  id: "agnes",
+  priority: 120,
+  alias: "agnes",
+  aliases: [
+    "agnes-ai",
+  ],
+  uiAlias: "agnes",
+  display: {
+    name: "Agnes AI",
+    icon: "auto_awesome",
+    color: "#7C3AED",
+    textIcon: "AG",
+    website: "https://agnes-ai.com",
+    notice: {
+      text: "OpenAI-compatible gateway from Agnes AI, offering free API credits on sign-up. Accepts a bearer token or an x-api-key header.",
+      apiKeyUrl: "https://platform.agnes-ai.com",
+    },
+  },
+  category: "freeTier",
+  authType: "apikey",
+  transport: {
+    baseUrl: "https://apihub.agnes-ai.com/v1/chat/completions",
+    validateUrl: "https://apihub.agnes-ai.com/v1/models",
+  },
+  // No model ids could be verified without a key, so discovery is left to the
+  // live endpoint and any id is accepted through passthroughModels.
+  passthroughModels: true,
+};

open-sse/providers/registry/atria.js

@@ -0,0 +1,32 @@
+export default {
+  id: "atria",
+  priority: 120,
+  alias: "atria",
+  aliases: [
+    "atria-asi",
+  ],
+  uiAlias: "atria",
+  display: {
+    name: "Atria Dawn",
+    icon: "flare",
+    color: "#C2410C",
+    textIcon: "AD",
+    website: "https://atria-asi.ai",
+    notice: {
+      text: "OpenAI-compatible endpoint from Atria Dawn (AtomInnoLab). Currently a research preview offering a single text model, Atria-Dawn-Preview.",
+      apiKeyUrl: "https://api.atria-asi.ai/dashboard",
+    },
+  },
+  category: "apikey",
+  authType: "apikey",
+  transport: {
+    baseUrl: "https://api.atria-asi.ai/v1/chat/completions",
+    validateUrl: "https://api.atria-asi.ai/v1/models",
+  },
+  // Docs pin the model field to one case-sensitive id. Text-only for now: the
+  // service ships a hook that blocks image/PDF input, so no vision is claimed.
+  models: [
+    { id: "Atria-Dawn-Preview", name: "Atria Dawn Preview" },
+  ],
+  passthroughModels: true,
+};

open-sse/providers/registry/bai.js

@@ -0,0 +1,30 @@
+export default {
+  id: "bai",
+  priority: 120,
+  alias: "bai",
+  aliases: [
+    "b-ai",
+  ],
+  uiAlias: "bai",
+  display: {
+    name: "B.AI",
+    icon: "account_balance",
+    color: "#0369A1",
+    textIcon: "BA",
+    website: "https://b.ai",
+    notice: {
+      text: "OpenAI-compatible gateway with one of the larger catalogues here. Accepts a bearer token or an x-api-key header. Model ids are fetched live from the provider.",
+      apiKeyUrl: "https://b.ai",
+    },
+  },
+  category: "apikey",
+  authType: "apikey",
+  transport: {
+    baseUrl: "https://api.b.ai/v1/chat/completions",
+    validateUrl: "https://api.b.ai/v1/models",
+  },
+  // No ids hardcoded: the catalogue is large and rotates, so the live endpoint
+  // is the source of truth and any id is accepted via passthroughModels.
+  modelsFetcher: { url: "https://api.b.ai/v1/models", type: "openai" },
+  passthroughModels: true,
+};

open-sse/providers/registry/claude.js

@@ -54,6 +54,8 @@ export default {
       oauthUrl: "https://api.anthropic.com/api/oauth/usage",
       orgUrl: "https://api.anthropic.com/v1/organizations/{org_id}/usage",
       settingsUrl: "https://api.anthropic.com/v1/settings",
+      profileUrl: "https://api.anthropic.com/api/oauth/profile",
+      resetUrl: "https://api.anthropic.com/api/organizations/{org_id}/reset_rate_limits",
     },
   },
   models: [

open-sse/providers/registry/codex.js

@@ -2,7 +2,8 @@ import { withCodexReviewModels } from "../models/helpers.js";
 
 // Codex CLI version seen by OpenAI's backend — single source for the Version /
 // User-Agent identity headers. Bump when the installed codex CLI is upgraded.
-const CODEX_CLI_VERSION = "0.154.0";
+const CODEX_CLI_VERSION = "0.155.0";
+const GPT_6_LITE_THINKING_LEVELS = ["low", "medium", "high", "xhigh", "max"];
 
 export default {
   id: "codex",
@@ -42,6 +43,7 @@ export default {
     headers: {
       originator: "codex_cli_rs",
       "User-Agent": `codex_cli_rs/${CODEX_CLI_VERSION}`,
+      version: CODEX_CLI_VERSION,
     },
     usage: {
       url: "https://chatgpt.com/backend-api/wham/usage",
@@ -51,6 +53,8 @@ export default {
   },
   models: [
     { id: "gpt-6-astra", name: "GPT 6.0 Astra" },
+    { id: "gpt-6-sol", name: "GPT 6.0 Sol", responsesLite: true, thinkingLevels: GPT_6_LITE_THINKING_LEVELS },
+    { id: "gpt-6-luna", name: "GPT 6.0 Luna", responsesLite: true, thinkingLevels: GPT_6_LITE_THINKING_LEVELS },
     { id: "gpt-5.6-sol", name: "GPT 5.6 Sol" },
     { id: "gpt-5.6-sol-review", name: "GPT 5.6 Sol Review", upstreamModelId: "gpt-5.6-sol", quotaFamily: "review" },
     { id: "gpt-5.6-terra", name: "GPT 5.6 Terra" },

open-sse/providers/registry/dahl.js

@@ -0,0 +1,35 @@
+export default {
+  id: "dahl",
+  priority: 120,
+  alias: "dahl",
+  aliases: [
+    "dahl-inference",
+  ],
+  uiAlias: "dahl",
+  display: {
+    name: "Dahl Inference",
+    icon: "hub",
+    color: "#1E40AF",
+    textIcon: "DH",
+    website: "https://dahl.global",
+    notice: {
+      text: "OpenAI-compatible Gonka inference node. Small, fixed catalogue (GLM-5.3-Flash, DeepSeek-V4-Flash, MiniMax-M2.7) at a flat per-token rate.",
+      apiKeyUrl: "https://dahl.global/dashboard",
+    },
+  },
+  category: "apikey",
+  authType: "apikey",
+  transport: {
+    baseUrl: "https://inference.dahl.global/v1/chat/completions",
+    validateUrl: "https://inference.dahl.global/v1/models",
+  },
+  // The live catalogue is public (no auth), so modelsFetcher works without a key
+  // and the ids below are a convenience seed rather than an exhaustive list.
+  models: [
+    { id: "zai-org/GLM-5.3-Flash", name: "GLM-5.3 Flash" },
+    { id: "deepseek-ai/DeepSeek-V4-Flash-0731", name: "DeepSeek V4 Flash 0731" },
+    { id: "MiniMaxAI/MiniMax-M2.7", name: "MiniMax M2.7" },
+  ],
+  modelsFetcher: { url: "https://inference.dahl.global/v1/models", type: "openai" },
+  passthroughModels: true,
+};

open-sse/providers/registry/gemini.js

@@ -58,6 +58,7 @@ export default {
     { id: "gemini-2.5-flash", name: "Gemini 2.5 Flash", params: ["language","prompt"], kind: "stt" },
     { id: "gemini-2.5-flash-lite", name: "Gemini 2.5 Flash Lite (Cheapest)", params: ["language","prompt"], kind: "stt" },
     { id: "gemini-2.0-flash", name: "Gemini 2.0 Flash", params: ["language","prompt"], kind: "stt" },
+    { id: "gemini-2.5-flash-native-audio-preview-09-17", name: "Gemini Live Transcription (Realtime)", params: ["language","prompt","system_instruction","setup_timeout_ms","turn_timeout_ms"], kind: "stt", transport: "gemini-live" },
     { id: "gemini-3.1-flash-tts-preview", name: "Gemini 3.1 Flash TTS", kind: "tts" },
     { id: "gemini-2.5-flash-preview-tts", name: "Gemini 2.5 Flash TTS", kind: "tts" },
     { id: "gemini-2.5-pro-preview-tts", name: "Gemini 2.5 Pro TTS", kind: "tts" },

open-sse/providers/registry/index.js

@@ -125,6 +125,11 @@ import p119 from "./selfhosted-embedding.js";
 import p120 from "./fish-audio.js";
 import p121 from "./alitp-intl.js";
 import p122 from "./xquik.js";
+import p125 from "./tokenharbor.js";
+import p126 from "./dahl.js";
+import p127 from "./atria.js";
+import p129 from "./agnes.js";
+import p130 from "./bai.js";
 export default [
   p0,
   p1,
@@ -250,4 +255,9 @@ export default [
   p120,
   p121,
   p122,
+  p125,
+  p126,
+  p127,
+  p129,
+  p130,
 ];

open-sse/providers/registry/opencode-go.js

@@ -35,37 +35,56 @@ export default {
   ],
   // supportedFormats follow the endpoint table in https://opencode.ai/docs/go/
   models: [
-    { id: "deepseek-flash", name: "DeepSeek V4.1 Flash", supportedFormats: ["openai"] },
+    { id: "deepseek-flash", name: "DeepSeek Flash", supportedFormats: ["openai"] },
     { id: "glm-5.3-flash", name: "GLM 5.3 Flash (Vision)", supportedFormats: ["openai"] },
     { id: "glm-5.3", name: "GLM 5.3", supportedFormats: ["openai"] },
     { id: "glm-5.2", name: "GLM 5.2", supportedFormats: ["openai"] },
     { id: "glm-5.1", name: "GLM 5.1", supportedFormats: ["openai"] },
+    { id: "glm-5", name: "GLM 5", supportedFormats: ["openai"] },
     { id: "kimi-k2.7-code", name: "Kimi K2.7 Code", supportedFormats: ["openai"] },
     { id: "kimi-k2.6", name: "Kimi K2.6", supportedFormats: ["openai"] },
+    { id: "kimi-k2.5", name: "Kimi K2.5", supportedFormats: ["openai"] },
     { id: "kimi-k3", name: "Kimi K3", supportedFormats: ["openai"] },
     { id: "deepseek-v4-pro", name: "DeepSeek V4 Pro", supportedFormats: ["openai", "claude", "openai-responses"] },
     { id: "deepseek-v4-flash", name: "DeepSeek V4 Flash", supportedFormats: ["openai", "claude", "openai-responses"] },
     { id: "deepseek-v4-flash-vision-exp", name: "DeepSeek V4 Flash Vision (Exp)", supportedFormats: ["openai", "claude", "openai-responses"] },
+    { id: "deepseek-v4.1-flash", name: "DeepSeek V4.1 Flash", supportedFormats: ["openai", "claude", "openai-responses"] },
     { id: "longcat-2.0", name: "LongCat 2.0", supportedFormats: ["openai"] },
+    { id: "mimo-v2.6-flash", name: "MiMo V2.6 Flash", supportedFormats: ["openai"] },
+    { id: "mimo-v2.6-pro", name: "MiMo V2.6 Pro", supportedFormats: ["openai"] },
     { id: "mimo-v2.5", name: "MiMo V2.5", supportedFormats: ["openai"] },
     { id: "mimo-v2.5-pro", name: "MiMo V2.5 Pro", supportedFormats: ["openai"] },
+    { id: "mimo-v2-pro", name: "MiMo V2 Pro", supportedFormats: ["openai"] },
+    { id: "mimo-v2-omni", name: "MiMo V2 Omni", supportedFormats: ["openai"] },
     { id: "minimax-m3", name: "MiniMax M3", supportedFormats: ["openai", "claude"] },
     { id: "minimax-m2.7", name: "MiniMax M2.7", supportedFormats: ["openai", "claude"] },
     { id: "minimax-m2.5", name: "MiniMax M2.5", supportedFormats: ["openai", "claude"] },
+    { id: "space-bunny-free", name: "Space Bunny Free", supportedFormats: ["openai", "claude"] },
     { id: "qwen3.8-max", name: "Qwen 3.8 Max", supportedFormats: ["openai", "claude"] },
     { id: "qwen3.8-flash", name: "Qwen 3.8 Flash", supportedFormats: ["openai", "claude"] },
     { id: "qwen3.7-max", name: "Qwen 3.7 Max", supportedFormats: ["openai", "claude"] },
     { id: "qwen3.7-plus", name: "Qwen 3.7 Plus", supportedFormats: ["openai", "claude"] },
     { id: "qwen3.6-plus", name: "Qwen 3.6 Plus", supportedFormats: ["openai", "claude"] },
+    { id: "qwen3.5-plus", name: "Qwen 3.5 Plus", supportedFormats: ["openai", "claude"] },
     { id: "hy4-preview", name: "Hy4 Preview", supportedFormats: ["openai"] },
     { id: "hy3", name: "Hy3", supportedFormats: ["openai"] },
+    { id: "hy3-preview", name: "Hy3 Preview", supportedFormats: ["openai"] },
+    // In /zen/go/v1/models but absent from the docs endpoint table — chat lane is the fallback guess
+    { id: "omen-alpha", name: "Omen Alpha", supportedFormats: ["openai"] },
     // Served by /zen/go/v1/responses only — the responses-only entry forces chatCore
     // past the sourceFormat-matched transports into translation (see chatCore guard).
+    { id: "grok-4.7", name: "Grok 4.7", targetFormat: "openai-responses", supportedFormats: ["openai-responses"] },
     { id: "grok-4.6", name: "Grok 4.6", targetFormat: "openai-responses", supportedFormats: ["openai-responses"] },
+    { id: "grok-4.5", name: "Grok 4.5", targetFormat: "openai-responses", supportedFormats: ["openai-responses"] },
     { id: "gpt-5.6-luna", name: "GPT 5.6 Luna", targetFormat: "openai-responses", supportedFormats: ["openai-responses"] },
+    { id: "gpt-6-luna", name: "GPT 6 Luna", targetFormat: "openai-responses", supportedFormats: ["openai-responses"] },
     { id: "muse-spark-1.2-contributor", name: "Muse Spark 1.2 Contributor", targetFormat: "openai-responses", supportedFormats: ["openai-responses"] },
     { id: "muse-spark-1.3-contributor", name: "Muse Spark 1.3 Contributor", targetFormat: "openai-responses", supportedFormats: ["openai-responses"] },
   ],
+  // Live catalogue; ids outside this curated list get their endpoint lane from the
+  // family regex in providers/models/helpers.js (opencodeFamilyFormats).
+  modelsFetcher: { url: "https://opencode.ai/zen/go/v1/models", type: "opencode-go" },
+  passthroughModels: true,
   features: {
     usage: true,
     usageApikey: true,

open-sse/providers/registry/tokenharbor.js

@@ -0,0 +1,49 @@
+export default {
+  id: "tokenharbor",
+  priority: 120,
+  alias: "tokenharbor",
+  aliases: [
+    "th",
+    "thh",
+  ],
+  uiAlias: "tokenharbor",
+  display: {
+    name: "Token Harbor",
+    icon: "anchor",
+    color: "#0F766E",
+    textIcon: "TH",
+    website: "https://tokenharbor.ai",
+    notice: {
+      text: "OpenAI-compatible aggregator. One API key reaches every model, billed per-token from a prepaid wallet. Model ids are bare (e.g. claude-opus-5.5, gpt-6-astra, deepseek-v4.1-flash:free) and are fetched live from the provider.",
+      apiKeyUrl: "https://tokenharbor.ai/dashboard",
+    },
+  },
+  category: "apikey",
+  authType: "apikey",
+  transport: {
+    // OpenAI-compatible. `format` is left at the shared "openai" default and
+    // `thinkingFormat` is deliberately NOT declared: Token Harbor forwards
+    // requests verbatim, so each model must resolve its own thinking wire
+    // format through providers/capabilities.js. Setting a provider-wide value
+    // would force one format (e.g. claude-adaptive) onto every model.
+    baseUrl: "https://tokenharbor.ai/v1/chat/completions",
+    validateUrl: "https://tokenharbor.ai/v1/models",
+    retry: {
+      429: 2,
+    },
+  },
+  // Curated seed; the live catalogue is fetched via modelsFetcher and any other
+  // id is accepted via passthroughModels. Their catalogue rotates (the :free set
+  // in particular), so this stays deliberately small and is only the offline
+  // fallback. Ids are bare — Token Harbor does not prefix them by upstream vendor.
+  models: [
+    { id: "claude-opus-5.5", name: "Claude Opus 5.5" },
+    { id: "claude-sonnet-5", name: "Claude Sonnet 5" },
+    { id: "gpt-6-astra", name: "GPT-6 Astra" },
+    { id: "gpt-6-sol", name: "GPT-6 Sol" },
+    { id: "deepseek-v4.1-flash:free", name: "DeepSeek V4.1 Flash (Free)" },
+    { id: "grok-4.7", name: "Grok 4.7" },
+  ],
+  modelsFetcher: { url: "https://tokenharbor.ai/v1/models", type: "openai" },
+  passthroughModels: true,
+};

open-sse/providers/shared.js

@@ -77,6 +77,11 @@ export function selectAnthropicBeta(model = "", body = null) {
   return flags.join(",");
 }
 
+export function mergeAnthropicBeta(...values) {
+  const flags = values.flatMap((v) => (typeof v === "string" ? v.split(",") : [])).map((f) => f.trim()).filter(Boolean);
+  return [...new Set(flags)].join(",");
+}
+
 // Shared baseUrls
 export const KIMI_CODING_BASE_URL = "https://api.kimi.com/coding/v1/messages";
 

open-sse/providers/thinkingLevels.js

@@ -3,6 +3,7 @@
 import { getCapabilitiesForModel } from "./capabilities.js";
 import { matchPattern } from "./pricing.js";
 import { resolveKiroEffortPath } from "../config/kiroConstants.js";
+import { getProviderModels } from "../config/providerModels.js";
 
 // Shared level sets (deduped) — verified against provider docs + wire in thinkingUnified.applyFormat.
 const L = {
@@ -42,6 +43,8 @@ const PATTERN_THINKING = [
   { provider: "codex", pattern: "*gpt-5.6-luna*", levels: CODEX_GPT_5_6_LEVELS },
   { pattern: "*codex*", levels: ["low", "medium", "high", "xhigh"] }, // codex cannot disable thinking
   { pattern: "*mimo*v2.6*", levels: ["none", "low", "medium", "high", "xhigh"] },
+  // mimo-v2.5-pro on opencode-go rejects reasoning_effort "max" (probed live); v2.5 accepts it.
+  { pattern: "*mimo*v2.5-pro*", levels: ["none", "low", "medium", "high", "xhigh"] },
   // 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.
@@ -73,10 +76,14 @@ export function getThinkingLevels(provider, model) {
   if (provider === "kiro" && resolveKiroEffortPath(model) === null) return null;
   const caps = getCapabilitiesForModel(provider, model);
   if (!caps.reasoning) return null;
+  const baseId = String(model || "").replace(/\([^()]+\)\s*$/, "");
+  const modelLevels = provider === "codex"
+    ? getProviderModels("cx").find((entry) => entry.id === baseId)?.thinkingLevels
+    : null;
   const hit = PATTERN_THINKING.find((entry) =>
     (!entry.provider || entry.provider === provider) && matchPattern(entry.pattern, model)
   );
-  let levels = hit?.levels || FORMAT_LEVELS[caps.thinkingFormat] || L.base;
+  let levels = modelLevels || hit?.levels || FORMAT_LEVELS[caps.thinkingFormat] || L.base;
   if (caps.thinkingCanDisable === false) levels = levels.filter((l) => l !== "none");
   return levels;
 }