@bitkyc08/opencodex 1.9.5 → 2.0.1

package/src/providers/derive.ts

@@ -8,8 +8,17 @@ export interface DerivedKeyLoginProvider {
   dashboardUrl: string;
   models?: string[];
   defaultModel?: string;
+  reasoningEfforts?: string[];
+  modelReasoningEfforts?: Record<string, string[]>;
+  reasoningEffortMap?: Record<string, string>;
+  modelReasoningEffortMap?: Record<string, Record<string, string>>;
   noVisionModels?: string[];
   noReasoningModels?: string[];
+  noTemperatureModels?: string[];
+  noTopPModels?: string[];
+  noPenaltyModels?: string[];
+  autoToolChoiceOnlyModels?: string[];
+  preserveReasoningContentModels?: string[];
 }
 
 export interface DerivedInitProvider {
@@ -38,6 +47,14 @@ export function listRegistryEntries(): readonly ProviderRegistryEntry[] {
   return PROVIDER_REGISTRY;
 }
 
+function cloneRecordOfArrays(input: Record<string, string[]>): Record<string, string[]> {
+  return Object.fromEntries(Object.entries(input).map(([key, value]) => [key, [...value]]));
+}
+
+function cloneNestedRecord(input: Record<string, Record<string, string>>): Record<string, Record<string, string>> {
+  return Object.fromEntries(Object.entries(input).map(([key, value]) => [key, { ...value }]));
+}
+
 export function providerConfigSeed(entry: ProviderRegistryEntry): OcxProviderConfig {
   return {
     adapter: entry.adapter,
@@ -45,8 +62,17 @@ export function providerConfigSeed(entry: ProviderRegistryEntry): OcxProviderCon
     authMode: entry.authKind === "local" ? undefined : entry.authKind,
     ...(entry.defaultModel ? { defaultModel: entry.defaultModel } : {}),
     ...(entry.models ? { models: [...entry.models] } : {}),
+    ...(entry.reasoningEfforts ? { reasoningEfforts: [...entry.reasoningEfforts] } : {}),
+    ...(entry.modelReasoningEfforts ? { modelReasoningEfforts: cloneRecordOfArrays(entry.modelReasoningEfforts) } : {}),
+    ...(entry.reasoningEffortMap ? { reasoningEffortMap: { ...entry.reasoningEffortMap } } : {}),
+    ...(entry.modelReasoningEffortMap ? { modelReasoningEffortMap: cloneNestedRecord(entry.modelReasoningEffortMap) } : {}),
     ...(entry.noVisionModels ? { noVisionModels: [...entry.noVisionModels] } : {}),
     ...(entry.noReasoningModels ? { noReasoningModels: [...entry.noReasoningModels] } : {}),
+    ...(entry.noTemperatureModels ? { noTemperatureModels: [...entry.noTemperatureModels] } : {}),
+    ...(entry.noTopPModels ? { noTopPModels: [...entry.noTopPModels] } : {}),
+    ...(entry.noPenaltyModels ? { noPenaltyModels: [...entry.noPenaltyModels] } : {}),
+    ...(entry.autoToolChoiceOnlyModels ? { autoToolChoiceOnlyModels: [...entry.autoToolChoiceOnlyModels] } : {}),
+    ...(entry.preserveReasoningContentModels ? { preserveReasoningContentModels: [...entry.preserveReasoningContentModels] } : {}),
   };
 }
 
@@ -62,8 +88,17 @@ export function deriveKeyLoginMap(): Record<string, DerivedKeyLoginProvider> {
       dashboardUrl: entry.dashboardUrl,
       ...(entry.models ? { models: [...entry.models] } : {}),
       ...(entry.defaultModel ? { defaultModel: entry.defaultModel } : {}),
+      ...(entry.reasoningEfforts ? { reasoningEfforts: [...entry.reasoningEfforts] } : {}),
+      ...(entry.modelReasoningEfforts ? { modelReasoningEfforts: cloneRecordOfArrays(entry.modelReasoningEfforts) } : {}),
+      ...(entry.reasoningEffortMap ? { reasoningEffortMap: { ...entry.reasoningEffortMap } } : {}),
+      ...(entry.modelReasoningEffortMap ? { modelReasoningEffortMap: cloneNestedRecord(entry.modelReasoningEffortMap) } : {}),
       ...(entry.noVisionModels ? { noVisionModels: [...entry.noVisionModels] } : {}),
       ...(entry.noReasoningModels ? { noReasoningModels: [...entry.noReasoningModels] } : {}),
+      ...(entry.noTemperatureModels ? { noTemperatureModels: [...entry.noTemperatureModels] } : {}),
+      ...(entry.noTopPModels ? { noTopPModels: [...entry.noTopPModels] } : {}),
+      ...(entry.noPenaltyModels ? { noPenaltyModels: [...entry.noPenaltyModels] } : {}),
+      ...(entry.autoToolChoiceOnlyModels ? { autoToolChoiceOnlyModels: [...entry.autoToolChoiceOnlyModels] } : {}),
+      ...(entry.preserveReasoningContentModels ? { preserveReasoningContentModels: [...entry.preserveReasoningContentModels] } : {}),
     };
   }
   return out;

package/src/providers/registry.ts

@@ -14,8 +14,17 @@ export interface ProviderRegistryEntry {
   dashboardUrl?: string;
   defaultModel?: string;
   models?: string[];
+  reasoningEfforts?: string[];
+  modelReasoningEfforts?: Record<string, string[]>;
+  reasoningEffortMap?: Record<string, string>;
+  modelReasoningEffortMap?: Record<string, Record<string, string>>;
   noVisionModels?: string[];
   noReasoningModels?: string[];
+  noTemperatureModels?: string[];
+  noTopPModels?: string[];
+  noPenaltyModels?: string[];
+  autoToolChoiceOnlyModels?: string[];
+  preserveReasoningContentModels?: string[];
   oauthId?: string;
   jawcodeBundle?: string;
   extraMetadataAliases?: string[];
@@ -24,9 +33,32 @@ export interface ProviderRegistryEntry {
 
 export type ProviderConfigSeed = Pick<
   OcxProviderConfig,
-  "adapter" | "baseUrl" | "authMode" | "defaultModel" | "models" | "noVisionModels" | "noReasoningModels"
+  "adapter" | "baseUrl" | "authMode" | "defaultModel" | "models"
+  | "reasoningEfforts" | "modelReasoningEfforts" | "reasoningEffortMap" | "modelReasoningEffortMap"
+  | "noVisionModels" | "noReasoningModels" | "noTemperatureModels" | "noTopPModels" | "noPenaltyModels"
+  | "autoToolChoiceOnlyModels" | "preserveReasoningContentModels"
 >;
 
+
+const ZAI_GLM_52_MODELS = ["glm-5.2", "glm-5.2[1m]"];
+const ZAI_GLM_52_REASONING_EFFORTS = ["low", "medium", "high", "xhigh"];
+const ZAI_GLM_52_REASONING_MAP: Record<string, string> = {
+  none: "none",
+  minimal: "none",
+  low: "high",
+  medium: "high",
+  high: "high",
+  xhigh: "max",
+  max: "max",
+};
+const KIMI_THINKING_MODELS = ["kimi-k2.7-code", "kimi-k2.7-code-highspeed", "kimi-k2.6", "kimi-k2.5", "kimi-k2-0905-preview"];
+const KIMI_LOCKED_PARAMETER_MODELS = ["kimi-k2.7-code", "kimi-k2.7-code-highspeed", "kimi-k2.6", "kimi-k2.5"];
+const NEURALWATT_REASONING_HISTORY_MODELS = [
+  "glm-5.2",
+  "moonshotai/Kimi-K2.5", "kimi-k2.6", "kimi-k2.7-code",
+  "qwen3.5-397b", "qwen3.6-35b",
+];
+
 export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
   {
     id: "openai",
@@ -75,11 +107,72 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     oauthId: "kimi",
     jawcodeBundle: "moonshot",
     note: "Log in with your Kimi account",
-    models: ["kimi-k2.6", "kimi-k2.5"],
-    defaultModel: "kimi-k2.6",
+    models: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed", "kimi-k2.6", "kimi-k2.5"],
+    defaultModel: "kimi-k2.7-code",
+    // Kimi thinking is controlled by Kimi's `thinking` extension, not OpenAI `reasoning_effort`.
+    noReasoningModels: KIMI_THINKING_MODELS,
+    modelReasoningEfforts: Object.fromEntries(KIMI_THINKING_MODELS.map(id => [id, []])),
+    noTemperatureModels: KIMI_LOCKED_PARAMETER_MODELS,
+    noTopPModels: KIMI_LOCKED_PARAMETER_MODELS,
+    noPenaltyModels: KIMI_LOCKED_PARAMETER_MODELS,
+    autoToolChoiceOnlyModels: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed"],
+    preserveReasoningContentModels: KIMI_THINKING_MODELS,
   },
   { id: "openai-apikey", label: "OpenAI (API key)", adapter: "openai-responses", baseUrl: "https://api.openai.com/v1", authKind: "key", featured: true, dashboardUrl: "https://platform.openai.com/api-keys", defaultModel: "gpt-5.5" },
-  { id: "opencode-go", label: "opencode go", adapter: "openai-chat", baseUrl: "https://opencode.ai/zen/go/v1", authKind: "key", featured: true, dashboardUrl: "https://opencode.ai/auth", defaultModel: "kimi-k2.6", jawcodeBundle: "opencode-go", note: "GLM, DeepSeek, Kimi, Qwen, MiMo…" },
+  {
+    id: "opencode-go", label: "opencode go", adapter: "openai-chat", baseUrl: "https://opencode.ai/zen/go/v1",
+    authKind: "key", featured: true, dashboardUrl: "https://opencode.ai/auth", defaultModel: "kimi-k2.7-code",
+    jawcodeBundle: "opencode-go", note: "GLM, DeepSeek, Kimi, Qwen, MiMo…",
+    modelReasoningEfforts: {
+      "glm-5.2": ZAI_GLM_52_REASONING_EFFORTS,
+      "kimi-k2.7-code": [],
+      "kimi-k2.7-code-highspeed": [],
+    },
+    modelReasoningEffortMap: { "glm-5.2": ZAI_GLM_52_REASONING_MAP },
+    noReasoningModels: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed"],
+    noTemperatureModels: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed"],
+    noTopPModels: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed"],
+    noPenaltyModels: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed"],
+    autoToolChoiceOnlyModels: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed"],
+    preserveReasoningContentModels: ["glm-5.2", "kimi-k2.7-code", "kimi-k2.7-code-highspeed"],
+  },
+  {
+    id: "neuralwatt",
+    label: "Neuralwatt Cloud",
+    adapter: "openai-chat",
+    baseUrl: "https://api.neuralwatt.com/v1",
+    authKind: "key",
+    dashboardUrl: "https://portal.neuralwatt.com",
+    defaultModel: "glm-5.2",
+    models: [
+      "glm-5.2", "glm-5.2-fast",
+      "moonshotai/Kimi-K2.5", "kimi-k2.5-fast", "kimi-k2.6", "kimi-k2.6-fast",
+      "kimi-k2.7-code",
+      "qwen3.5-397b", "qwen3.5-397b-fast", "qwen3.6-35b", "qwen3.6-35b-fast",
+    ],
+    // Neuralwatt's /v1/models metadata is authoritative; these static hints are the offline fallback.
+    modelReasoningEfforts: {
+      "glm-5.2": ZAI_GLM_52_REASONING_EFFORTS,
+      "glm-5.2-fast": [],
+      "moonshotai/Kimi-K2.5": [],
+      "kimi-k2.5-fast": [],
+      "kimi-k2.6": [],
+      "kimi-k2.6-fast": [],
+      "kimi-k2.7-code": [],
+      "qwen3.5-397b": ["low", "medium", "high"],
+      "qwen3.5-397b-fast": [],
+      "qwen3.6-35b": ["low", "medium", "high"],
+      "qwen3.6-35b-fast": [],
+    },
+    modelReasoningEffortMap: { "glm-5.2": ZAI_GLM_52_REASONING_MAP },
+    noReasoningModels: ["glm-5.2-fast", "kimi-k2.5-fast", "kimi-k2.6-fast", "qwen3.5-397b-fast", "qwen3.6-35b-fast"],
+    noVisionModels: ["glm-5.2", "glm-5.2-fast", "qwen3.5-397b", "qwen3.5-397b-fast"],
+    noTemperatureModels: ["kimi-k2.7-code"],
+    noTopPModels: ["kimi-k2.7-code"],
+    noPenaltyModels: ["kimi-k2.7-code"],
+    autoToolChoiceOnlyModels: ["kimi-k2.7-code"],
+    preserveReasoningContentModels: NEURALWATT_REASONING_HISTORY_MODELS,
+  },
   { id: "openrouter", label: "OpenRouter", adapter: "openai-chat", baseUrl: "https://openrouter.ai/api/v1", authKind: "key", featured: true, dashboardUrl: "https://openrouter.ai/keys", jawcodeBundle: "openrouter" },
   { id: "groq", label: "Groq", adapter: "openai-chat", baseUrl: "https://api.groq.com/openai/v1", authKind: "key", featured: true, dashboardUrl: "https://console.groq.com/keys" },
   { id: "google", label: "Google Gemini", adapter: "google", baseUrl: "https://generativelanguage.googleapis.com", authKind: "key", featured: true, dashboardUrl: "https://aistudio.google.com/apikey", defaultModel: "gemini-3-pro", jawcodeBundle: "google", extraMetadataAliases: ["gemini"] },
@@ -92,11 +185,31 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
   { id: "together", label: "Together", baseUrl: "https://api.together.xyz/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://api.together.xyz/settings/api-keys" },
   { id: "fireworks", label: "Fireworks", baseUrl: "https://api.fireworks.ai/inference/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://fireworks.ai/account/api-keys" },
   { id: "firepass", label: "Fire Pass (Fireworks Kimi)", baseUrl: "https://api.fireworks.ai/inference/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://fireworks.ai/account/api-keys" },
-  { id: "moonshot", label: "Moonshot (Kimi API)", baseUrl: "https://api.moonshot.ai/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://platform.moonshot.ai/console/api-keys", defaultModel: "kimi-k2-0905-preview", jawcodeBundle: "moonshot" },
+  {
+    id: "moonshot", label: "Moonshot (Kimi API)", baseUrl: "https://api.moonshot.ai/v1", adapter: "openai-chat", authKind: "key",
+    dashboardUrl: "https://platform.moonshot.ai/console/api-keys", defaultModel: "kimi-k2.7-code", jawcodeBundle: "moonshot",
+    models: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed", "kimi-k2.6", "kimi-k2.5", "kimi-k2-0905-preview"],
+    noReasoningModels: KIMI_THINKING_MODELS,
+    modelReasoningEfforts: Object.fromEntries(KIMI_THINKING_MODELS.map(id => [id, []])),
+    noTemperatureModels: KIMI_LOCKED_PARAMETER_MODELS,
+    noTopPModels: KIMI_LOCKED_PARAMETER_MODELS,
+    noPenaltyModels: KIMI_LOCKED_PARAMETER_MODELS,
+    autoToolChoiceOnlyModels: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed"],
+    preserveReasoningContentModels: KIMI_THINKING_MODELS,
+  },
   { id: "huggingface", label: "Hugging Face", baseUrl: "https://router.huggingface.co/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://huggingface.co/settings/tokens" },
   { id: "nvidia", label: "NVIDIA NIM", baseUrl: "https://integrate.api.nvidia.com/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://build.nvidia.com" },
   { id: "venice", label: "Venice", baseUrl: "https://api.venice.ai/api/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://venice.ai/settings/api" },
-  { id: "zai", label: "Z.AI (GLM Coding)", baseUrl: "https://api.z.ai/api/coding/paas/v4", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://z.ai/manage-apikey/apikey-list", defaultModel: "glm-4.6" },
+  {
+    id: "zai", label: "Z.AI — GLM Coding Plan", baseUrl: "https://api.z.ai/api/coding/paas/v4", adapter: "openai-chat", authKind: "key",
+    dashboardUrl: "https://z.ai/manage-apikey/apikey-list", defaultModel: "glm-5.2",
+    note: "GLM-5.2 coding subscription",
+    models: ["glm-5.2", "glm-5.2[1m]", "glm-5.1", "glm-5", "glm-4.6"],
+    noVisionModels: ZAI_GLM_52_MODELS,
+    modelReasoningEfforts: Object.fromEntries(ZAI_GLM_52_MODELS.map(id => [id, ZAI_GLM_52_REASONING_EFFORTS])),
+    modelReasoningEffortMap: Object.fromEntries(ZAI_GLM_52_MODELS.map(id => [id, ZAI_GLM_52_REASONING_MAP])),
+    preserveReasoningContentModels: ZAI_GLM_52_MODELS,
+  },
   { id: "nanogpt", label: "NanoGPT", baseUrl: "https://nano-gpt.com/api/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://nano-gpt.com/api" },
   { id: "synthetic", label: "Synthetic", baseUrl: "https://api.synthetic.new/openai/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://synthetic.new" },
   { id: "qwen-portal", label: "Qwen Portal", baseUrl: "https://portal.qwen.ai/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://portal.qwen.ai" },
@@ -123,9 +236,20 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     ],
   },
   { id: "mistral", label: "Mistral", baseUrl: "https://api.mistral.ai/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://console.mistral.ai/api-keys", defaultModel: "codestral-latest" },
-  { id: "minimax", label: "MiniMax", baseUrl: "https://api.minimax.io/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://platform.minimax.io", defaultModel: "MiniMax-M2.5", jawcodeBundle: "minimax", metadataModelIdNormalize: "case-insensitive" },
-  { id: "minimax-cn", label: "MiniMax (CN)", baseUrl: "https://api.minimaxi.com/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://platform.minimaxi.com", defaultModel: "MiniMax-M2.5", jawcodeBundle: "minimax", metadataModelIdNormalize: "case-insensitive" },
-  { id: "kimi-code", label: "Kimi (coding)", baseUrl: "https://api.kimi.com/coding/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://platform.moonshot.cn/console/api-keys", defaultModel: "kimi-k2.5" },
+  { id: "minimax", label: "MiniMax — Coding Plan", baseUrl: "https://api.minimax.io/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://platform.minimax.io", defaultModel: "MiniMax-M2.5", jawcodeBundle: "minimax", metadataModelIdNormalize: "case-insensitive", note: "Subscription Key or API Key" },
+  { id: "minimax-cn", label: "MiniMax — Coding Plan (CN)", baseUrl: "https://api.minimaxi.com/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://platform.minimaxi.com", defaultModel: "MiniMax-M2.5", jawcodeBundle: "minimax", metadataModelIdNormalize: "case-insensitive", note: "中国区 Subscription Key" },
+  {
+    id: "kimi-code", label: "Kimi (coding)", baseUrl: "https://api.kimi.com/coding/v1", adapter: "openai-chat", authKind: "key",
+    dashboardUrl: "https://platform.moonshot.cn/console/api-keys", defaultModel: "kimi-k2.7-code",
+    models: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed", "kimi-k2.6", "kimi-k2.5"],
+    noReasoningModels: KIMI_THINKING_MODELS,
+    modelReasoningEfforts: Object.fromEntries(KIMI_THINKING_MODELS.map(id => [id, []])),
+    noTemperatureModels: KIMI_LOCKED_PARAMETER_MODELS,
+    noTopPModels: KIMI_LOCKED_PARAMETER_MODELS,
+    noPenaltyModels: KIMI_LOCKED_PARAMETER_MODELS,
+    autoToolChoiceOnlyModels: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed"],
+    preserveReasoningContentModels: KIMI_THINKING_MODELS,
+  },
   { id: "opencode-zen", label: "opencode zen", baseUrl: "https://opencode.ai/zen/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://opencode.ai/auth" },
   { id: "vercel-ai-gateway", label: "Vercel AI Gateway", baseUrl: "https://ai-gateway.vercel.sh/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://vercel.com/dashboard" },
   { id: "xiaomi", label: "Xiaomi MiMo", baseUrl: "https://api.xiaomimimo.com/anthropic", adapter: "anthropic", authKind: "key", dashboardUrl: "https://xiaomimimo.com", defaultModel: "mimo-v2.5-pro" },

package/src/reasoning-effort.ts

@@ -0,0 +1,102 @@
+import type { OcxProviderConfig } from "./types";
+import { modelInList } from "./types";
+
+export const CODEX_REASONING_LEVELS: { effort: string; description: string }[] = [
+  { effort: "low", description: "Fast responses with lighter reasoning" },
+  { effort: "medium", description: "Balances speed and reasoning depth" },
+  { effort: "high", description: "Greater reasoning depth for complex problems" },
+  { effort: "xhigh", description: "Extended reasoning for the hardest problems" },
+];
+
+const CODEX_REASONING_ORDER = CODEX_REASONING_LEVELS.map(l => l.effort);
+const CODEX_REASONING_SET = new Set(CODEX_REASONING_ORDER);
+
+export function modelRecordValue<T>(record: Record<string, T> | undefined, modelId: string): T | undefined {
+  if (!record) return undefined;
+  if (Object.prototype.hasOwnProperty.call(record, modelId)) return record[modelId];
+  const colon = modelId.indexOf(":");
+  if (colon > 0) {
+    const family = modelId.slice(0, colon);
+    if (Object.prototype.hasOwnProperty.call(record, family)) return record[family];
+  }
+  const folded = modelId.toLowerCase();
+  for (const [key, value] of Object.entries(record)) {
+    if (key.toLowerCase() === folded) return value;
+  }
+  return undefined;
+}
+
+export function sanitizeCodexReasoningEfforts(efforts: readonly string[] | undefined): string[] | undefined {
+  if (efforts === undefined) return undefined;
+  const seen = new Set<string>();
+  const out: string[] = [];
+  for (const effort of efforts) {
+    if (!CODEX_REASONING_SET.has(effort) || seen.has(effort)) continue;
+    seen.add(effort);
+    out.push(effort);
+  }
+  return out.sort((a, b) => CODEX_REASONING_ORDER.indexOf(a) - CODEX_REASONING_ORDER.indexOf(b));
+}
+
+/**
+ * Provider/model configured reasoning levels for the Codex catalog. `undefined` means “no override”,
+ * while an empty array means “intentionally expose no effort control for this model”.
+ */
+export function configuredReasoningEfforts(provider: OcxProviderConfig, modelId: string): string[] | undefined {
+  if (modelInList(provider.noReasoningModels, modelId)) return [];
+  const modelEfforts = modelRecordValue(provider.modelReasoningEfforts, modelId);
+  if (modelEfforts !== undefined) return sanitizeCodexReasoningEfforts(modelEfforts) ?? [];
+  if (provider.reasoningEfforts !== undefined) return sanitizeCodexReasoningEfforts(provider.reasoningEfforts) ?? [];
+  return undefined;
+}
+
+function requestToCodexEffort(requested: string): string | undefined {
+  if (requested === "none") return undefined;
+  if (requested === "minimal") return "low";
+  if (requested === "max") return "xhigh";
+  return CODEX_REASONING_SET.has(requested) ? requested : undefined;
+}
+
+function clampToSupportedCodexEffort(requested: string, supported: readonly string[]): string | undefined {
+  if (supported.length === 0) return undefined;
+  const codex = requestToCodexEffort(requested);
+  if (!codex) return undefined;
+  if (supported.includes(codex)) return codex;
+
+  const requestedRank = CODEX_REASONING_ORDER.indexOf(codex);
+  let best = supported[0];
+  let bestRank = CODEX_REASONING_ORDER.indexOf(best);
+  for (const effort of supported) {
+    const rank = CODEX_REASONING_ORDER.indexOf(effort);
+    if (rank <= requestedRank && rank >= bestRank) {
+      best = effort;
+      bestRank = rank;
+    }
+  }
+  // If every supported tier is above the requested tier, choose the lowest supported tier.
+  return best;
+}
+
+export function reasoningEffortMapFor(provider: OcxProviderConfig, modelId: string): Record<string, string> | undefined {
+  return modelRecordValue(provider.modelReasoningEffortMap, modelId) ?? provider.reasoningEffortMap;
+}
+
+/**
+ * Translate Codex's reasoning label into the provider's real wire value. The Codex catalog must only
+ * advertise labels Codex itself accepts (`low`/`medium`/`high`/`xhigh`), but some upstreams use
+ * different values (`max`) or a smaller subset (`low`/`medium`/`high`).
+ */
+export function mapReasoningEffort(provider: OcxProviderConfig, modelId: string, requested: string | undefined): string | undefined {
+  if (!requested) return undefined;
+  if (modelInList(provider.noReasoningModels, modelId)) return undefined;
+
+  const wireMap = reasoningEffortMapFor(provider, modelId);
+  if (wireMap && Object.prototype.hasOwnProperty.call(wireMap, requested)) return wireMap[requested];
+
+  const supported = configuredReasoningEfforts(provider, modelId);
+  const codexEffort = supported !== undefined ? clampToSupportedCodexEffort(requested, supported) : requestToCodexEffort(requested);
+  if (!codexEffort) return undefined;
+
+  if (wireMap && Object.prototype.hasOwnProperty.call(wireMap, codexEffort)) return wireMap[codexEffort];
+  return codexEffort;
+}

package/src/responses/parser.ts

@@ -188,7 +188,7 @@ function findToolNameById(messages: OcxMessage[], callId: string): string {
   return "";
 }
 
-const REASONING_EFFORTS = new Set(["minimal", "low", "medium", "high", "xhigh", "max"]);
+const REASONING_EFFORTS = new Set(["none", "minimal", "low", "medium", "high", "xhigh", "max"]);
 
 export function parseRequest(body: unknown): OcxParsedRequest {
   const parsed = responsesRequestSchema.safeParse(body);

package/src/server.ts

@@ -1,4 +1,4 @@
-import { existsSync } from "node:fs";
+import { existsSync, readFileSync } from "node:fs";
 import { extname, join } from "node:path";
 import { createAnthropicAdapter } from "./adapters/anthropic";
 import { createAzureAdapter } from "./adapters/azure";
@@ -32,7 +32,15 @@ import { enrichProviderFromCatalog, listKeyLoginProviders } from "./oauth/key-pr
 import { deriveProviderPresets } from "./providers/derive";
 import type { OcxConfig, OcxProviderConfig } from "./types";
 
-const VERSION = "0.0.1";
+// Single source of truth = package.json (../ from src/), so /healthz + the GUI badge match the
+// installed npm version instead of a stale hardcode.
+const VERSION = (() => {
+  try {
+    return JSON.parse(readFileSync(new URL("../package.json", import.meta.url), "utf8")).version as string;
+  } catch {
+    return "0.0.0";
+  }
+})();
 
 const MIME_TYPES: Record<string, string> = {
   ".html": "text/html", ".js": "application/javascript", ".css": "text/css",
@@ -354,6 +362,15 @@ function jsonResponse(data: unknown, status = 200): Response {
 }
 
 async function handleManagementAPI(req: Request, url: URL, config: OcxConfig): Promise<Response | null> {
+  async function refreshCodexCatalogBestEffort(): Promise<void> {
+    try {
+      const { refreshCodexModelCatalog } = await import("./codex-refresh");
+      await refreshCodexModelCatalog(config);
+    } catch {
+      /* catalog absent */
+    }
+  }
+
   if (url.pathname === "/api/config" && req.method === "GET") {
     const safeConfig = JSON.parse(JSON.stringify(config));
     for (const prov of Object.values(safeConfig.providers as Record<string, OcxProviderConfig>)) {
@@ -398,6 +415,7 @@ async function handleManagementAPI(req: Request, url: URL, config: OcxConfig): P
     config.providers[name] = prov;
     if (body.setDefault) config.defaultProvider = name;
     save(config);
+    await refreshCodexCatalogBestEffort();
     return jsonResponse({ success: true, name });
   }
 
@@ -408,11 +426,7 @@ async function handleManagementAPI(req: Request, url: URL, config: OcxConfig): P
     delete config.providers[name];
     save(config);
     // Drop its models from Codex's catalog immediately (re-sync + cache bust) so removal is live.
-    try {
-      const { syncCatalogModels, invalidateCodexModelsCache } = await import("./codex-catalog");
-      await syncCatalogModels(config);
-      invalidateCodexModelsCache();
-    } catch { /* catalog absent */ }
+    await refreshCodexCatalogBestEffort();
     return jsonResponse({ success: true });
   }
 
@@ -434,11 +448,7 @@ async function handleManagementAPI(req: Request, url: URL, config: OcxConfig): P
     config.disabledModels = disabled;
     const { saveConfig: save } = await import("./config");
     save(config);
-    try {
-      const { syncCatalogModels, invalidateCodexModelsCache } = await import("./codex-catalog");
-      await syncCatalogModels(config);
-      invalidateCodexModelsCache();
-    } catch { /* catalog absent */ }
+    await refreshCodexCatalogBestEffort();
     return jsonResponse({ ok: true, disabled });
   }
 
@@ -479,11 +489,7 @@ async function handleManagementAPI(req: Request, url: URL, config: OcxConfig): P
     config.subagentModels = chosen;
     const { saveConfig: save } = await import("./config");
     save(config);
-    try {
-      const { syncCatalogModels, invalidateCodexModelsCache } = await import("./codex-catalog");
-      await syncCatalogModels(config);
-      invalidateCodexModelsCache();
-    } catch { /* catalog absent */ }
+    await refreshCodexCatalogBestEffort();
     return jsonResponse({ ok: true, applied: chosen });
   }
 
@@ -522,6 +528,15 @@ async function handleManagementAPI(req: Request, url: URL, config: OcxConfig): P
     return jsonResponse({ success: true });
   }
 
+  if (url.pathname === "/api/stop" && req.method === "POST") {
+    const { restoreNativeCodex } = await import("./codex-inject");
+    const { stopServiceIfInstalled } = await import("./service");
+    stopServiceIfInstalled();
+    restoreNativeCodex();
+    setTimeout(() => process.exit(0), 200);
+    return jsonResponse({ success: true, message: "Proxy stopping, native Codex restored." });
+  }
+
   return null;
 }
 
@@ -551,6 +566,7 @@ export function startServer(port?: number) {
 
   const server = Bun.serve<WsData>({
     port: listenPort,
+    idleTimeout: 255,
     async fetch(req) {
       const url = new URL(req.url);
 

package/src/service.ts

@@ -107,9 +107,13 @@ export function buildWindowsServiceScript(): string {
     windowsBatchSet("OCX_SERVICE", "1"),
     windowsBatchSet("PATH", path),
     windowsBatchSet("CODEX_HOME", process.env.CODEX_HOME?.trim()),
+    ":loop",
     `"${bun}" "${cli}" start`,
-    "set \"OCX_EXIT=%ERRORLEVEL%\"",
-    "endlocal & exit /b %OCX_EXIT%",
+    "if %ERRORLEVEL% NEQ 0 (",
+    "  timeout /t 5 /nobreak >nul",
+    "  goto loop",
+    ")",
+    "endlocal",
   ].filter((line): line is string => Boolean(line));
   return `${lines.join("\r\n")}\r\n`;
 }
@@ -232,6 +236,26 @@ function platformOps(): ServiceOps | null {
   return null;
 }
 
+/**
+ * If a service is installed, stop it so the process manager doesn't respawn after `ocx stop`.
+ * Returns true if a service was found and stopped.
+ */
+export function stopServiceIfInstalled(): boolean {
+  if (process.platform === "darwin") {
+    if (existsSync(plistPath())) {
+      try { stopLaunchd(); return true; } catch { return false; }
+    }
+  } else if (process.platform === "win32") {
+    try {
+      const q = sh(`schtasks /query /tn ${TASK} 2>nul`);
+      if (q.includes(TASK)) { stopWindows(); return true; }
+    } catch { /* task not found */ }
+  } else if (process.platform === "linux" && isSystemd() && existsSync(unitPath())) {
+    try { stopSystemd(); return true; } catch { return false; }
+  }
+  return false;
+}
+
 export function serviceCommand(sub?: string): void {
   const ops = platformOps();
   if (!ops) {

package/src/star-prompt.ts

@@ -5,7 +5,7 @@ import { createInterface } from "node:readline/promises";
 import { getConfigDir } from "./config";
 
 const REPO = "lidge-jun/opencodex";
-/** Shared with scripts/postinstall.mjs so the prompt fires exactly once across install + first start. */
+/** Fires exactly once from the first interactive `ocx start`. */
 const MARKER = ".star-prompted";
 
 function ghAvailable(): boolean {
@@ -22,9 +22,10 @@ function starRepo(): { ok: boolean; error?: string } {
 }
 
 /**
- * First interactive `ocx start`: a one-time `[Y/n]` "star on GitHub?" prompt. On yes, stars the repo
- * via the user's `gh` auth (same approach as the npm postinstall). No-op under the background service,
- * for non-TTY/piped runs, when already prompted, or when `gh` is unavailable. Never throws.
+ * First interactive `ocx start`: a one-time `[Y/n]` "star on GitHub?" prompt.
+ * On yes, stars the repo via the user's `gh` auth. No-op under the background
+ * service, for non-TTY/piped runs, when already prompted, or when `gh` is
+ * unavailable. Never throws.
  */
 export async function maybeShowStarPrompt(): Promise<void> {
   try {

package/src/types.ts

@@ -221,11 +221,33 @@ export interface OcxProviderConfig {
    * Only the openai-responses adapter implements "forward"; openai-chat uses its own key/token.
    */
   authMode?: "key" | "forward" | "oauth";
+  /**
+   * Provider-wide Codex-visible reasoning tiers for routed models. Use only Codex-supported labels
+   * here (`low`, `medium`, `high`, `xhigh`); translate to provider-specific wire values with
+   * `reasoningEffortMap` / `modelReasoningEffortMap` below.
+   */
+  reasoningEfforts?: string[];
+  /** Model-specific Codex-visible reasoning tiers. An empty array means “do not expose effort”. */
+  modelReasoningEfforts?: Record<string, string[]>;
+  /** Provider-wide mapping from Codex effort labels to upstream `reasoning_effort` values. */
+  reasoningEffortMap?: Record<string, string>;
+  /** Model-specific mapping from Codex effort labels to upstream `reasoning_effort` values. */
+  modelReasoningEffortMap?: Record<string, Record<string, string>>;
   /**
    * Model ids that do NOT support a reasoning/thinking parameter. The openai-chat adapter drops
    * reasoning_effort for these even when Codex selects a reasoning level (e.g. xAI grok-build-0.1).
    */
   noReasoningModels?: string[];
+  /** Model ids that reject caller-specified temperature. */
+  noTemperatureModels?: string[];
+  /** Model ids that reject caller-specified top_p. */
+  noTopPModels?: string[];
+  /** Model ids that reject caller-specified presence/frequency penalty values. */
+  noPenaltyModels?: string[];
+  /** Model ids whose tool_choice only accepts `auto` or `none`; forced/named choices are downgraded. */
+  autoToolChoiceOnlyModels?: string[];
+  /** Model ids that expect prior assistant `reasoning_content` to be preserved in chat history. */
+  preserveReasoningContentModels?: string[];
   /**
    * Model ids that do NOT accept image inputs. The proxy gives them "eyes" via the vision sidecar:
    * attached images are described by a gpt vision model and replaced with text before the call.

package/src/ws-bridge.ts

@@ -219,9 +219,12 @@ export function sendResponsesJsonAsEvents(
       item,
     });
   });
+  const finalStatus = response.status === "failed" || response.status === "incomplete"
+    ? response.status
+    : "completed";
   sendJsonFrame(ws, {
-    type: "response.completed",
-    response: { ...response, status: "completed" },
+    type: finalStatus === "failed" ? "response.failed" : "response.completed",
+    response: { ...response, status: finalStatus },
   });
 }
 

package/gui/dist/assets/index-C1wlp1SM.css

@@ -1 +0,0 @@
-:root{--bg:#0b0b0f;--surface:#14141a;--raised:#1c1c25;--raised-hover:#232330;--border:#2a2a35;--border-soft:#20202a;--text:#e9e9ee;--muted:#9a9aa6;--faint:#6a6a76;--accent:#7c5cff;--accent-hover:#9077ff;--accent-ink:#fff;--accent-soft:#7c5cff24;--accent-ring:#7c5cff73;--green:#34d399;--green-soft:#34d39921;--red:#f87171;--red-soft:#f8717121;--amber:#fbbf24;--amber-soft:#fbbf2421;--radius:12px;--radius-sm:8px;--radius-xs:6px;--font:-apple-system, BlinkMacSystemFont, "Segoe UI", Roboto, system-ui, "Helvetica Neue", sans-serif;--mono:ui-monospace, "SF Mono", "JetBrains Mono", "Cascadia Code", Menlo, Consolas, monospace;--shadow:0 1px 2px #00000080, 0 12px 32px #00000047;--shadow-sm:0 1px 2px #0006;--lightningcss-light: ;--lightningcss-dark:initial;color-scheme:dark}*{box-sizing:border-box}html,body,#root{height:100%}body{background:var(--bg);color:var(--text);font-family:var(--font);-webkit-font-smoothing:antialiased;text-rendering:optimizelegibility;background-image:radial-gradient(1200px 600px at 18% -10%,#7c5cff14,#0000 60%);margin:0;font-size:14px;line-height:1.5}a{color:var(--accent-hover);text-decoration:none}a:hover{text-decoration:underline}code,.mono{font-family:var(--mono);font-size:.92em}h1,h2,h3,h4{letter-spacing:-.01em;margin:0;font-weight:650}::selection{background:var(--accent-soft)}::-webkit-scrollbar{width:10px;height:10px}::-webkit-scrollbar-thumb{background:var(--border);border:2px solid var(--bg);border-radius:99px}::-webkit-scrollbar-thumb:hover{background:#353542}:focus-visible{outline:2px solid var(--accent-ring);outline-offset:2px;border-radius:4px}.app{grid-template-columns:232px 1fr;min-height:100dvh;display:grid}.sidebar{border-right:1px solid var(--border-soft);background:linear-gradient(#ffffff04,#0000);flex-direction:column;align-self:start;gap:4px;height:100dvh;padding:18px 14px;display:flex;position:sticky;top:0}.brand{align-items:center;gap:10px;padding:6px 8px 14px;display:flex}.brand img{width:26px;height:26px}.brand .name{letter-spacing:-.02em;font-size:15px;font-weight:700}.brand .ver{font-family:var(--mono);color:var(--muted);background:var(--raised);border:1px solid var(--border);border-radius:99px;padding:1px 6px;font-size:10px}.nav-item{border-radius:var(--radius-sm);text-align:left;cursor:pointer;width:100%;color:var(--muted);font:inherit;background:0 0;border:none;align-items:center;gap:10px;padding:8px 10px;font-size:13.5px;font-weight:500;transition:background .12s,color .12s;display:flex}.nav-item:hover{background:var(--raised);color:var(--text)}.nav-item.active{background:var(--accent-soft);color:var(--text)}.nav-item svg{width:17px;height:17px;color:var(--faint);flex-shrink:0}.nav-item.active svg{color:var(--accent)}.sidebar-foot{margin-top:auto;padding-top:12px}.sidebar-link{color:var(--muted);border-radius:var(--radius-sm);align-items:center;gap:9px;padding:8px 10px;font-size:13px;display:flex}.sidebar-link:hover{background:var(--raised);color:var(--text);text-decoration:none}.sidebar-link svg{width:16px;height:16px}.main{min-width:0}.main-inner{max-width:980px;margin:0 auto;padding:32px 36px 64px}.page-head{justify-content:space-between;align-items:center;gap:16px;margin-bottom:6px;display:flex}.page-head h2{font-size:19px}.page-sub{color:var(--muted);max-width:70ch;margin:4px 0 22px;font-size:13.5px}.page-sub b{color:var(--text);font-weight:600}.btn{border-radius:var(--radius-sm);font:inherit;cursor:pointer;white-space:nowrap;border:1px solid #0000;justify-content:center;align-items:center;gap:7px;padding:7px 14px;font-size:13px;font-weight:550;transition:background .12s,border-color .12s,opacity .12s;display:inline-flex}.btn svg{width:15px;height:15px}.btn:disabled{opacity:.55;cursor:default}.btn-primary{background:var(--accent);color:var(--accent-ink)}.btn-primary:hover:not(:disabled){background:var(--accent-hover)}.btn-ghost{background:var(--raised);color:var(--text);border-color:var(--border)}.btn-ghost:hover:not(:disabled){background:var(--raised-hover)}.btn-danger{color:var(--red);background:0 0;border-color:#f871714d}.btn-danger:hover:not(:disabled){background:var(--red-soft)}.btn-sm{border-radius:var(--radius-xs);padding:4px 9px;font-size:12px}.btn-icon{padding:5px}.card{background:var(--surface);border:1px solid var(--border);border-radius:var(--radius)}.panel{background:var(--surface);border:1px solid var(--border);border-radius:var(--radius);padding:18px}.panel-accent{background:linear-gradient(180deg, var(--accent-soft), transparent 120%), var(--surface);border-color:#7c5cff47}.stat-row{grid-template-columns:repeat(4,1fr);gap:12px;margin-bottom:28px;display:grid}.stat{background:var(--surface);border:1px solid var(--border);border-radius:var(--radius);padding:15px 16px}.stat .label{color:var(--muted);align-items:center;gap:6px;margin-bottom:7px;font-size:12px;display:flex}.stat .label svg{width:14px;height:14px}.stat .value{letter-spacing:-.02em;font-size:23px;font-weight:700;line-height:1.1}.stat .value.mono{font-family:var(--mono);font-size:19px}.badge{font-size:11px;font-weight:600;font-family:var(--mono);letter-spacing:.01em;border-radius:99px;align-items:center;gap:5px;padding:2px 8px;display:inline-flex}.badge-accent{background:var(--accent-soft);color:var(--accent-hover)}.badge-green{background:var(--green-soft);color:var(--green)}.badge-amber{background:var(--amber-soft);color:var(--amber)}.badge-muted{background:var(--raised);color:var(--muted);border:1px solid var(--border)}.dot{border-radius:50%;flex-shrink:0;width:7px;height:7px}.dot-green{background:var(--green);box-shadow:0 0 0 3px var(--green-soft)}.dot-red{background:var(--red);box-shadow:0 0 0 3px var(--red-soft)}.tbl{border-collapse:collapse;width:100%;font-size:13px}.tbl thead th{text-align:left;color:var(--muted);text-transform:uppercase;letter-spacing:.04em;border-bottom:1px solid var(--border);padding:9px 12px;font-size:11.5px;font-weight:600}.tbl tbody td{border-bottom:1px solid var(--border-soft);padding:10px 12px}.tbl tbody tr:last-child td{border-bottom:none}.tbl tbody tr:hover td{background:#ffffff05}.tbl .num{text-align:right;font-family:var(--mono)}.tbl-wrap{border:1px solid var(--border);border-radius:var(--radius);overflow:hidden}.input,textarea.input{border-radius:var(--radius-sm);background:var(--raised);border:1px solid var(--border);width:100%;color:var(--text);font:inherit;padding:8px 11px;font-size:13px;transition:border-color .12s}.input::placeholder{color:var(--faint)}.input:focus{border-color:var(--accent);outline:none}textarea.input{resize:vertical;font-family:var(--mono);line-height:1.55}.field-label{color:var(--muted);margin-bottom:5px;font-size:12px;font-weight:500;display:block}select.input{appearance:none}.switch{cursor:pointer;background:var(--border);border:none;border-radius:99px;flex-shrink:0;width:34px;height:19px;padding:0;transition:background .15s;position:relative}.switch.on{background:var(--accent)}.switch:disabled{opacity:.6;cursor:default}.switch .knob{background:#fff;border-radius:50%;width:15px;height:15px;transition:left .15s;position:absolute;top:2px;left:2px}.switch.on .knob{left:17px}.muted{color:var(--muted)}.faint{color:var(--faint)}.row{align-items:center;gap:10px;display:flex}.spread{justify-content:space-between;align-items:center;gap:12px;display:flex}.stack{flex-direction:column;display:flex}.chip{font-family:var(--mono);background:var(--raised);border:1px solid var(--border);border-radius:var(--radius-xs);color:var(--text);padding:1px 7px;font-size:12px}.empty{text-align:center;border:1px dashed var(--border);border-radius:var(--radius);color:var(--muted);padding:56px 20px}.empty svg{width:30px;height:30px;color:var(--faint);margin-bottom:12px}.empty .title{color:var(--text);margin-bottom:6px;font-weight:600}.notice{border-radius:var(--radius-sm);align-items:center;gap:8px;margin-bottom:14px;padding:9px 12px;font-size:13px;display:flex}.notice svg{flex-shrink:0;width:15px;height:15px}.notice-ok{background:var(--green-soft);color:var(--green)}.notice-err{background:var(--red-soft);color:var(--red)}.h-section{color:var(--text);align-items:center;gap:8px;margin:30px 0 12px;font-size:13px;font-weight:600;display:flex}.h-section .count{color:var(--muted);font-weight:500;font-family:var(--mono);font-size:12px}.spin{border:2px solid var(--border);border-top-color:var(--accent);border-radius:50%;width:14px;height:14px;animation:.7s linear infinite spin;display:inline-block}@keyframes spin{to{transform:rotate(360deg)}}@media (prefers-reduced-motion:reduce){*{transition:none!important;animation:none!important}}.modal-overlay{-webkit-backdrop-filter:blur(2px);backdrop-filter:blur(2px);z-index:50;background:#0009;justify-content:center;align-items:flex-start;padding:8vh 16px;display:flex;position:fixed;inset:0}.modal-card{background:var(--surface);border:1px solid var(--border);border-radius:var(--radius);width:100%;max-width:520px;box-shadow:var(--shadow);max-height:84vh;padding:20px;overflow-y:auto}.modal-head{justify-content:space-between;align-items:center;margin-bottom:16px;display:flex}.modal-head h3{font-size:16px}.list-row{text-align:left;border-radius:var(--radius-sm);border:1px solid var(--border);background:var(--raised);cursor:pointer;width:100%;color:var(--text);font:inherit;justify-content:space-between;align-items:center;gap:10px;padding:11px 13px;transition:background .12s,border-color .12s;display:flex}.list-row:hover{background:var(--raised-hover);border-color:#34343f}.list-row .title{font-size:14px;font-weight:600}.list-row .sub{color:var(--muted);margin-top:2px;font-size:12px}.prov-card{justify-content:space-between;align-items:flex-start;gap:12px;padding:15px 16px;display:flex}.link-btn{color:var(--accent-hover);font:inherit;cursor:pointer;background:0 0;border:none;padding:6px 2px;font-size:13px;text-decoration:underline}@media (width<=760px){.app{grid-template-columns:1fr}.sidebar{border-right:none;border-bottom:1px solid var(--border-soft);flex-flow:wrap;align-items:center;height:auto;position:static}.brand{width:100%;padding:6px 8px}.nav-item{width:auto}.sidebar-foot{margin:0;padding:0}.main-inner{padding:22px 18px 48px}.stat-row{grid-template-columns:repeat(2,1fr)}}