@prestyj/core 5.28.0 → 5.29.0

package/dist/index.cjs

@@ -69,7 +69,7 @@ __export(index_exports, {
   getContextWindow: () => getContextWindow,
   getDefaultModel: () => getDefaultModel,
   getDefaultThinkingLevel: () => getDefaultThinkingLevel,
-  getFastModel: () => getFastModel,
+  getLowestThinkingLevel: () => getLowestThinkingLevel,
   getMaxThinkingLevel: () => getMaxThinkingLevel,
   getModel: () => getModel,
   getModelsForProvider: () => getModelsForProvider,
@@ -152,6 +152,12 @@ function getAppPaths() {
   };
 }
 
+// src/oauth/request-timeout.ts
+var OAUTH_REQUEST_TIMEOUT_MS = 3e4;
+function oauthRequestSignal() {
+  return AbortSignal.timeout(OAUTH_REQUEST_TIMEOUT_MS);
+}
+
 // src/oauth/kimi.ts
 var CLIENT_ID = "17e5f671-d194-4dfb-9706-5516cb48c098";
 var DEFAULT_OAUTH_HOST = "https://auth.kimi.com";
@@ -239,7 +245,8 @@ async function postForm(endpoint, params) {
       "Content-Type": "application/x-www-form-urlencoded",
       Accept: "application/json"
     },
-    body: new URLSearchParams(params).toString()
+    body: new URLSearchParams(params).toString(),
+    signal: oauthRequestSignal()
   });
   let data = {};
   try {
@@ -713,7 +720,12 @@ async function postTokenRequest(body, label) {
   for (const url of TOKEN_URLS) {
     let response;
     try {
-      response = await fetch(url, { method: "POST", headers, body: encoded });
+      response = await fetch(url, {
+        method: "POST",
+        headers,
+        body: encoded,
+        signal: oauthRequestSignal()
+      });
     } catch (err) {
       lastError = err instanceof Error ? err : new Error(String(err));
       continue;
@@ -962,7 +974,8 @@ async function exchangeOpenAICode(code, verifier) {
       code,
       redirect_uri: REDIRECT_URI2,
       code_verifier: verifier
-    })
+    }),
+    signal: oauthRequestSignal()
   });
   if (!response.ok) {
     const text = await response.text();
@@ -983,7 +996,8 @@ async function refreshOpenAIToken(refreshToken) {
       grant_type: "refresh_token",
       refresh_token: refreshToken,
       client_id: CLIENT_ID3
-    })
+    }),
+    signal: oauthRequestSignal()
   });
   if (!response.ok) {
     const text = await response.text();
@@ -1193,7 +1207,8 @@ async function postTokenRequest2(body) {
   const response = await fetch(TOKEN_URL2, {
     method: "POST",
     headers: { "Content-Type": "application/x-www-form-urlencoded" },
-    body: new URLSearchParams(body)
+    body: new URLSearchParams(body),
+    signal: oauthRequestSignal()
   });
   if (!response.ok) {
     const text = await response.text();
@@ -1322,7 +1337,8 @@ async function codeAssistGet(accessToken, operationName) {
 async function codeAssistRequest(url, accessToken, label, init) {
   const response = await fetch(url, {
     ...init,
-    headers: codeAssistHeaders(accessToken)
+    headers: codeAssistHeaders(accessToken),
+    signal: oauthRequestSignal()
   });
   if (!response.ok) {
     const text = await response.text();
@@ -1408,7 +1424,8 @@ async function postForm2(endpoint, params) {
       Accept: "application/json",
       "User-Agent": `grok-cli/${grokCliVersion()}`
     },
-    body: new URLSearchParams(params).toString()
+    body: new URLSearchParams(params).toString(),
+    signal: oauthRequestSignal()
   });
   let data = {};
   try {
@@ -2244,7 +2261,7 @@ var MODELS = [
   // ── OpenAI (Codex) ─────────────────────────────────────
   {
     // GPT-6 Astra — "Our most capable model for complex, demanding work."
-    // (Codex catalog priority 1, listed for every ChatGPT plan, requires a
+    // (Codex catalog priority 2, listed for every ChatGPT plan, requires a
     // Codex client >= 0.153.0 — see CODEX_CLIENT_VERSION). Same split as 5.6:
     // 1.05M on the public Responses API, 272K on the ChatGPT OAuth route
     // (openai/codex models.json, `gpt-6-astra`). Reasoning ladder low → medium
@@ -2268,32 +2285,36 @@ var MODELS = [
   },
   // GPT-6 Sol + Luna — released 2026-09-22 below Astra, replacing the whole
   // GPT-5.6 family (Sol/Terra/Luna; there is no GPT-6 Terra — OpenAI's Codex
-  // catalog upgrades 5.6 Terra to 6 Sol). Both need a Codex client >= 0.155.0
-  // on the ChatGPT OAuth route. Same window split as Astra: 1.05M on the public
-  // Responses API, 272K on the Codex route; 128K output, text+image input,
-  // freeform apply_patch, responses-lite transport. The 5.6 ids are retired —
-  // a saved session on one falls back to the provider default on next start.
+  // catalog upgrades 5.6 Terra to 6 Sol). GPT-6.1 Sol (2026-09-29) then replaced
+  // GPT-6 Sol. Same window split as Astra: 1.05M on the public Responses API,
+  // 272K on the Codex route; 128K output, text+image input, freeform
+  // apply_patch, responses-lite transport. The GPT-5.6 ids and `gpt-6-sol` are
+  // retired — a saved session on one falls back to the provider default on
+  // next start.
   {
-    // Sol — "Workhorse model for coding and everyday work." (Codex priority 2,
-    // default medium). $2/$10 MTok. Ladder low → medium → high → xhigh → max →
-    // ultra; ultra is the Codex orchestration preset (max effort on the wire +
-    // proactive local subagent delegation).
-    id: "gpt-6-sol",
-    name: "GPT-6 Sol",
+    // GPT-6.1 Sol — "Latest workhorse model for coding and everyday work."
+    // (Codex priority 1, default low). The catalog says Codex client >= 0.153.0,
+    // but the ChatGPT backend only serves it from 0.159.0 (see
+    // CODEX_CLIENT_VERSION). $2/$10 MTok, cached input $0.10. Ladder low →
+    // medium → high → xhigh → max → ultra; ultra is the Codex orchestration
+    // preset (max effort on the wire + proactive local subagent delegation).
+    id: "gpt-6.1-sol",
+    name: "GPT-6.1 Sol",
     provider: "openai",
     contextWindow: 105e4,
     codexContextWindow: 272e3,
     maxOutputTokens: 128e3,
     supportsThinking: true,
-    defaultThinkingLevel: "medium",
+    defaultThinkingLevel: "low",
     supportsImages: true,
     supportsVideo: false,
     costTier: "medium",
     maxThinkingLevel: "ultra"
   },
   {
-    // Luna — "Fast and affordable model for easier tasks." (Codex priority 3,
-    // default medium). $0.10/$0.50 MTok. Reasoning tops out at `max`.
+    // Luna — "Fast and affordable model for easier tasks." (Codex priority 4,
+    // default medium, needs a Codex client >= 0.155.0). $0.10/$0.50 MTok.
+    // Reasoning tops out at `max`.
     id: "gpt-6-luna",
     name: "GPT-6 Luna",
     provider: "openai",
@@ -2426,9 +2447,7 @@ var MODELS = [
     // Sent over our Code Assist (OAuth) transport ahead of gemini-cli — upstream
     // hasn't listed 3.7 yet (google-gemini/gemini-cli#28802, still open) — so
     // free/personal accounts 404 (entitlement-gated) while Code Assist
-    // Standard/Enterprise accounts get it. Kept after the working flash-lite:
-    // getFastModel picks the first low-tier entry, and flash-lite is the one
-    // that works on every account.
+    // Standard/Enterprise accounts get it.
     id: "gemini-3.7-flash",
     name: "Gemini 3.7 Flash",
     provider: "gemini",
@@ -2789,7 +2808,7 @@ function getVideoByteLimit(modelId) {
 }
 function getDefaultModel(provider) {
   if (provider === "xiaomi") return MODELS.find((m) => m.id === "mimo-v2.6-pro");
-  if (provider === "openai") return MODELS.find((m) => m.id === "gpt-6-sol");
+  if (provider === "openai") return MODELS.find((m) => m.id === "gpt-6.1-sol");
   if (provider === "gemini") return MODELS.find((m) => m.id === "gemini-3.1-flash-lite");
   if (provider === "glm") return MODELS.find((m) => m.id === "glm-5.3");
   if (provider === "moonshot") return MODELS.find((m) => m.id === "kimi-k3");
@@ -2839,19 +2858,17 @@ function getDefaultThinkingLevel(modelId, options) {
   if (model?.id === "kimi-k3" && isKimiCodingEndpoint(options?.baseUrl)) return "high";
   return model?.defaultThinkingLevel ?? model?.maxThinkingLevel ?? "high";
 }
+var SUMMARY_COST_TIER = {
+  anthropic: "medium",
+  openai: "low",
+  glm: "low",
+  deepseek: "low",
+  huggingface: "low"
+};
 function getSummaryModel(provider, currentModelId) {
-  if (provider === "anthropic") {
-    return MODELS.find((m) => m.id === "claude-sonnet-5-5");
-  }
-  if (provider === "openai" || provider === "glm" || provider === "deepseek" || provider === "huggingface") {
-    const low = getModelsForProvider(provider).find((m) => m.costTier === "low");
-    if (low) return low;
-  }
-  return getModel(currentModelId) ?? getDefaultModel(provider);
-}
-function getFastModel(provider, currentModelId) {
-  const low = getModelsForProvider(provider).find((m) => m.costTier === "low");
-  return low ?? getModel(currentModelId) ?? getDefaultModel(provider);
+  const tier = SUMMARY_COST_TIER[provider];
+  const tagged = tier ? getModelsForProvider(provider).find((m) => m.costTier === tier) : void 0;
+  return tagged ?? getModel(currentModelId) ?? getDefaultModel(provider);
 }
 
 // src/thinking-level.ts
@@ -2936,11 +2953,15 @@ function getSupportedThinkingLevels(provider, model) {
     return GLM_THINKING_LEVELS.slice(0, maxIndex2 + 1);
   }
   if (!isOpenAIGptModel(provider, model)) return [maxLevel];
-  const levels = model.startsWith("gpt-5.6-") || model.startsWith("gpt-6-") ? OPENAI_GPT_56_THINKING_LEVELS : OPENAI_GPT_THINKING_LEVELS;
+  const levels = model.startsWith("gpt-5.6-") || model.startsWith("gpt-6-") || model.startsWith("gpt-6.") ? OPENAI_GPT_56_THINKING_LEVELS : OPENAI_GPT_THINKING_LEVELS;
   const maxIndex = levels.indexOf(maxLevel);
   if (maxIndex === -1) return ["medium"];
   return levels.slice(0, maxIndex + 1);
 }
+function getLowestThinkingLevel(provider, model) {
+  if (getModel(model)?.supportsThinking === false) return void 0;
+  return getSupportedThinkingLevels(provider, model)[0];
+}
 function isThinkingLevelSupported(provider, model, level) {
   return getSupportedThinkingLevels(provider, model).includes(level);
 }
@@ -3488,12 +3509,16 @@ async function fetchSubscriptionUsage(provider, credentials, options = {}) {
 
 // src/telegram.ts
 var TELEGRAM_API = "https://api.telegram.org";
+var LONG_POLL_SECONDS = 30;
+var API_TIMEOUT_MS = 3e4;
 var MAX_MESSAGE_LENGTH = 4096;
 var TelegramBot = class {
   token;
   allowedUserId;
   offset = 0;
   running = false;
+  /** Aborts the in-flight long poll so stop() takes effect immediately. */
+  pollAbort = null;
   onMessage = null;
   onVoiceMessage = null;
   onCallback = null;
@@ -3546,6 +3571,7 @@ var TelegramBot = class {
   /** Stop long polling. */
   stop() {
     this.running = false;
+    this.pollAbort?.abort();
   }
   /** Send a text message to a specific chat. Converts markdown and splits long messages. */
   async send(chatId, text, buttons) {
@@ -3592,12 +3618,26 @@ var TelegramBot = class {
   }
   // ── Private ───────────────────────────────────────────
   async getUpdates() {
-    const result = await this.apiCall("getUpdates", {
-      offset: this.offset,
-      timeout: 30,
-      allowed_updates: ["message", "callback_query", "my_chat_member"]
-    });
-    if (!result.ok || !Array.isArray(result.result)) return [];
+    const poll = new AbortController();
+    this.pollAbort = poll;
+    let result;
+    try {
+      result = await this.apiCall(
+        "getUpdates",
+        {
+          offset: this.offset,
+          timeout: LONG_POLL_SECONDS,
+          allowed_updates: ["message", "callback_query", "my_chat_member"]
+        },
+        // A half-open socket would otherwise hang the poll forever.
+        AbortSignal.any([poll.signal, AbortSignal.timeout((LONG_POLL_SECONDS + 15) * 1e3)])
+      );
+    } finally {
+      if (this.pollAbort === poll) this.pollAbort = null;
+    }
+    if (!result.ok || !Array.isArray(result.result)) {
+      throw new Error(`getUpdates failed: ${JSON.stringify(result)}`);
+    }
     const updates = result.result;
     if (updates.length > 0) {
       this.offset = updates[updates.length - 1].update_id + 1;
@@ -3645,12 +3685,13 @@ var TelegramBot = class {
       }
     }
   }
-  async apiCall(method, body) {
+  async apiCall(method, body, signal = AbortSignal.timeout(API_TIMEOUT_MS)) {
     const url = `${TELEGRAM_API}/bot${this.token}/${method}`;
     const response = await fetch(url, {
       method: "POST",
       headers: { "Content-Type": "application/json" },
-      body: body ? JSON.stringify(body) : void 0
+      body: body ? JSON.stringify(body) : void 0,
+      signal
     });
     if (!response.ok) {
       return { ok: false };
@@ -3783,7 +3824,7 @@ function isModelLoaded() {
   return transcriber !== null;
 }
 async function transcribeVoice(fileUrl) {
-  const response = await fetch(fileUrl);
+  const response = await fetch(fileUrl, { signal: AbortSignal.timeout(6e4) });
   if (!response.ok) throw new Error(`Failed to download voice file: ${response.status}`);
   const buffer = new Uint8Array(await response.arrayBuffer());
   const pcm = await decodeOggOpus(buffer);
@@ -4002,7 +4043,7 @@ function createAutoUpdater(config) {
   getContextWindow,
   getDefaultModel,
   getDefaultThinkingLevel,
-  getFastModel,
+  getLowestThinkingLevel,
   getMaxThinkingLevel,
   getModel,
   getModelsForProvider,