openfox 2.0.0-beta.11 → 2.0.0-beta.13

package/dist/{chunk-EZUR7OEP.js → chunk-HZODNSSV.js}

@@ -1,12 +1,12 @@
+import {
+  getGlobalConfigPath
+} from "./chunk-CQGTEGKL.js";
 import {
   detectModel
-} from "./chunk-UKTPL5ZG.js";
+} from "./chunk-SW7MQGY3.js";
 import {
   detectBackend
-} from "./chunk-VIIRNJDT.js";
-import {
-  getGlobalConfigPath
-} from "./chunk-CQGTEGKL.js";
+} from "./chunk-LXX2CPM5.js";
 
 // src/cli/config.ts
 import { z } from "zod";
@@ -124,8 +124,15 @@ var oldLlmSchema = z.object({
   model: z.string().default("auto"),
   backend: z.enum(["auto", "vllm", "sglang", "ollama", "llamacpp"]).default("auto"),
   maxContext: z.number().default(2e5),
-  disableThinking: z.boolean().default(false),
+  // Deprecated: use reasoningEffort instead
+  disableThinking: z.boolean().optional(),
+  reasoningEffort: z.string().optional(),
   apiKey: z.string().optional()
+}).transform((data) => {
+  if (data.disableThinking && !data.reasoningEffort) {
+    return { ...data, reasoningEffort: "none", disableThinking: void 0 };
+  }
+  return { ...data, disableThinking: void 0 };
 });
 var oldConfigSchema = z.object({
   llm: oldLlmSchema,
@@ -416,8 +423,7 @@ function mergeConfigs(...configs) {
         url: "http://localhost:8000/v1",
         model: "auto",
         backend: "auto",
-        maxContext: 2e5,
-        disableThinking: false
+        maxContext: 2e5
       },
       server: { port: 10369, host: "127.0.0.1", openBrowser: true },
       logging: { level: "error" }
@@ -443,4 +449,4 @@ export {
   activateProvider,
   mergeConfigs
 };
-//# sourceMappingURL=chunk-EZUR7OEP.js.map
+//# sourceMappingURL=chunk-HZODNSSV.js.map

package/dist/{chunk-LCRJWUW7.js → chunk-LDHRZRCP.js}

@@ -33,7 +33,7 @@ Options:
 }
 async function runNetworkSetup(mode) {
   const { loadAuthConfig, saveAuthConfig, encryptPassword } = await import("./auth-56SIRACI.js");
-  const { saveGlobalConfig } = await import("./config-TDVA7MQN.js");
+  const { saveGlobalConfig } = await import("./config-FAT5XRFB.js");
   const { getAuthKeyPath } = await import("./paths-X46PPOI2.js");
   const existingAuth = await loadAuthConfig(mode);
   if (existingAuth) {
@@ -94,7 +94,7 @@ async function runNetworkSetup(mode) {
   console.log("\u2713 Configuration saved!\n");
 }
 async function runConfig(mode) {
-  const { loadGlobalConfig, getActiveProvider, getDefaultModel } = await import("./config-TDVA7MQN.js");
+  const { loadGlobalConfig, getActiveProvider, getDefaultModel } = await import("./config-FAT5XRFB.js");
   const { getGlobalConfigPath } = await import("./paths-X46PPOI2.js");
   const config = await loadGlobalConfig(mode);
   const configPath = getGlobalConfigPath(mode);
@@ -151,7 +151,7 @@ async function runCli(options) {
       break;
     }
     case "provider": {
-      const { runProviderCommand } = await import("./provider-BGH4MBLH.js");
+      const { runProviderCommand } = await import("./provider-2JYU6KGH.js");
       const [, subcommand] = positionals;
       await runProviderCommand(mode, subcommand);
       break;
@@ -188,12 +188,12 @@ async function runCli(options) {
       break;
     }
     default: {
-      const { configFileExists } = await import("./config-TDVA7MQN.js");
+      const { configFileExists } = await import("./config-FAT5XRFB.js");
       const configExists = await configFileExists(mode);
       if (!configExists) {
         await runNetworkSetup(mode);
       }
-      const { runServe } = await import("./serve-I6T7V62Z.js");
+      const { runServe } = await import("./serve-NSTNUQ2K.js");
       const serveOptions = { mode };
       if (values.port) serveOptions.port = parseInt(values.port);
       if (values["no-browser"] === true) serveOptions.openBrowser = false;
@@ -205,4 +205,4 @@ async function runCli(options) {
 export {
   runCli
 };
-//# sourceMappingURL=chunk-LCRJWUW7.js.map
+//# sourceMappingURL=chunk-LDHRZRCP.js.map

package/dist/{chunk-VIIRNJDT.js → chunk-LXX2CPM5.js}

@@ -2,6 +2,176 @@ import {
   logger
 } from "./chunk-K44MW7JJ.js";
 
+// src/server/llm/profiles.ts
+var DEFAULT_PROFILE = {
+  name: "default",
+  temperature: 0.7,
+  topP: 0.9,
+  defaultMaxTokens: 4096,
+  supportsVision: true
+};
+var MOCK_PROFILE = {
+  name: "mock",
+  temperature: 0.7,
+  topP: 0.9,
+  defaultMaxTokens: 1024,
+  supportsVision: false
+};
+var MODEL_PROFILES = [
+  {
+    pattern: "mistral",
+    profile: {
+      name: "Mistral",
+      temperature: 0.7,
+      topP: 0.9,
+      defaultMaxTokens: 16384,
+      supportsVision: false
+    }
+  },
+  {
+    pattern: "qwen3-coder-next",
+    profile: {
+      name: "Qwen3-Coder-Next",
+      // Per Qwen docs: "temperature=1.0, top_p=0.95, top_k=40"
+      temperature: 1,
+      topP: 0.95,
+      topK: 40,
+      // "This model supports only non-thinking mode and does not generate <think></think> blocks"
+      defaultMaxTokens: 16384,
+      supportsVision: false
+    }
+  },
+  {
+    pattern: "qwen3",
+    profile: {
+      name: "Qwen3",
+      temperature: 0.7,
+      topP: 0.9,
+      defaultMaxTokens: 16384,
+      supportsVision: false
+    }
+  },
+  {
+    pattern: "qwen3-vl",
+    profile: {
+      name: "Qwen3-VL",
+      temperature: 0.7,
+      topP: 0.9,
+      defaultMaxTokens: 16384,
+      supportsVision: true
+    }
+  },
+  {
+    pattern: "deepseek",
+    profile: {
+      name: "DeepSeek",
+      temperature: 0.6,
+      topP: 0.95,
+      defaultMaxTokens: 16384,
+      supportsVision: false
+    }
+  },
+  {
+    pattern: "minimax-m2.5",
+    profile: {
+      name: "MiniMax-M2.5",
+      temperature: 1,
+      topP: 0.95,
+      topK: 40,
+      defaultMaxTokens: 16384,
+      supportsVision: false
+    }
+  },
+  {
+    pattern: "minimax-m2.7",
+    profile: {
+      name: "MiniMax-M2.7",
+      temperature: 1,
+      topP: 0.95,
+      topK: 40,
+      defaultMaxTokens: 16384,
+      supportsVision: false
+    }
+  },
+  {
+    pattern: "minimax",
+    profile: {
+      name: "MiniMax",
+      temperature: 1,
+      topP: 0.95,
+      topK: 40,
+      defaultMaxTokens: 16384,
+      supportsVision: false
+    }
+  },
+  {
+    pattern: "minimax-m3",
+    profile: {
+      name: "MiniMax-M3",
+      temperature: 1,
+      topP: 0.95,
+      topK: 40,
+      defaultMaxTokens: 16384,
+      supportsVision: true
+    }
+  },
+  {
+    pattern: "llava",
+    profile: {
+      name: "LLaVA",
+      temperature: 0.7,
+      topP: 0.9,
+      defaultMaxTokens: 16384,
+      supportsVision: true
+    }
+  },
+  {
+    pattern: "llama",
+    profile: {
+      name: "Llama",
+      temperature: 0.7,
+      topP: 0.9,
+      defaultMaxTokens: 16384,
+      supportsVision: false
+    }
+  },
+  {
+    pattern: "claude",
+    profile: {
+      name: "Claude",
+      temperature: 0.7,
+      topP: 0.9,
+      defaultMaxTokens: 16384,
+      supportsVision: false
+    }
+  },
+  {
+    pattern: "gemma-4",
+    profile: {
+      name: "Gemma 4",
+      temperature: 0.7,
+      topP: 0.9,
+      defaultMaxTokens: 16384,
+      supportsVision: true
+    }
+  }
+];
+function getModelProfile(modelName) {
+  const lowerName = modelName.toLowerCase();
+  if (lowerName.includes("mock")) {
+    return MOCK_PROFILE;
+  }
+  for (const { pattern, profile } of MODEL_PROFILES) {
+    if (lowerName.includes(pattern.toLowerCase())) {
+      return profile;
+    }
+  }
+  return DEFAULT_PROFILE;
+}
+function modelSupportsVision(modelName) {
+  return getModelProfile(modelName).supportsVision;
+}
+
 // src/server/llm/url-utils.ts
 var VERSION_PREFIX_REGEX = /\/v\d+(\/|$)/;
 function hasVersionPrefix(url) {
@@ -21,33 +191,27 @@ function buildModelsUrl(baseUrl) {
 // src/server/llm/backend.ts
 var BACKEND_CAPABILITIES = {
   vllm: {
-    supportsReasoningField: true,
     supportsChatTemplateKwargs: true,
     supportsTopK: true
   },
   sglang: {
-    supportsReasoningField: true,
     supportsChatTemplateKwargs: true,
     supportsTopK: true
   },
   ollama: {
-    supportsReasoningField: false,
     supportsChatTemplateKwargs: false,
     supportsTopK: false
   },
   llamacpp: {
-    supportsReasoningField: false,
     supportsChatTemplateKwargs: false,
     supportsTopK: true
   },
   "opencode-go": {
-    supportsReasoningField: false,
     supportsChatTemplateKwargs: false,
     supportsTopK: true
   },
   unknown: {
     // Assume vLLM-like for unknown backends
-    supportsReasoningField: true,
     supportsChatTemplateKwargs: true,
     supportsTopK: true
   }
@@ -179,206 +343,6 @@ function getBackendDisplayName(backend) {
   }
 }
 
-// src/server/llm/profiles.ts
-var DEFAULT_PROFILE = {
-  name: "default",
-  temperature: 0.7,
-  topP: 0.9,
-  supportsReasoning: false,
-  reasoningAsContent: false,
-  defaultMaxTokens: 4096,
-  supportsVision: true
-};
-var MOCK_PROFILE = {
-  name: "mock",
-  temperature: 0.7,
-  topP: 0.9,
-  supportsReasoning: false,
-  reasoningAsContent: false,
-  defaultMaxTokens: 1024,
-  supportsVision: false
-};
-var MODEL_PROFILES = [
-  {
-    pattern: "mistral",
-    profile: {
-      name: "Mistral",
-      temperature: 0.7,
-      topP: 0.9,
-      supportsReasoning: false,
-      reasoningAsContent: false,
-      defaultMaxTokens: 16384,
-      supportsVision: false
-    }
-  },
-  {
-    pattern: "qwen3-coder-next",
-    profile: {
-      name: "Qwen3-Coder-Next",
-      // Per Qwen docs: "temperature=1.0, top_p=0.95, top_k=40"
-      temperature: 1,
-      topP: 0.95,
-      topK: 40,
-      // "This model supports only non-thinking mode and does not generate <think></think> blocks"
-      supportsReasoning: false,
-      reasoningAsContent: false,
-      defaultMaxTokens: 16384,
-      supportsVision: false
-    }
-  },
-  {
-    pattern: "qwen3",
-    profile: {
-      name: "Qwen3",
-      temperature: 0.7,
-      topP: 0.9,
-      supportsReasoning: true,
-      reasoningAsContent: false,
-      defaultMaxTokens: 16384,
-      supportsVision: false
-    }
-  },
-  {
-    pattern: "qwen3-vl",
-    profile: {
-      name: "Qwen3-VL",
-      temperature: 0.7,
-      topP: 0.9,
-      supportsReasoning: true,
-      reasoningAsContent: false,
-      defaultMaxTokens: 16384,
-      supportsVision: true
-    }
-  },
-  {
-    pattern: "deepseek",
-    profile: {
-      name: "DeepSeek",
-      temperature: 0.6,
-      topP: 0.95,
-      supportsReasoning: true,
-      reasoningAsContent: false,
-      defaultMaxTokens: 16384,
-      supportsVision: false
-    }
-  },
-  {
-    pattern: "minimax-m2.5",
-    profile: {
-      name: "MiniMax-M2.5",
-      temperature: 1,
-      topP: 0.95,
-      topK: 40,
-      supportsReasoning: true,
-      reasoningAsContent: false,
-      defaultMaxTokens: 16384,
-      supportsVision: false
-    }
-  },
-  {
-    pattern: "minimax-m2.7",
-    profile: {
-      name: "MiniMax-M2.7",
-      temperature: 1,
-      topP: 0.95,
-      topK: 40,
-      supportsReasoning: true,
-      reasoningAsContent: false,
-      defaultMaxTokens: 16384,
-      supportsVision: false
-    }
-  },
-  {
-    pattern: "minimax",
-    profile: {
-      name: "MiniMax",
-      temperature: 1,
-      topP: 0.95,
-      topK: 40,
-      supportsReasoning: true,
-      reasoningAsContent: false,
-      defaultMaxTokens: 16384,
-      supportsVision: false
-    }
-  },
-  {
-    pattern: "minimax-m3",
-    profile: {
-      name: "MiniMax-M3",
-      temperature: 1,
-      topP: 0.95,
-      topK: 40,
-      supportsReasoning: true,
-      reasoningAsContent: false,
-      defaultMaxTokens: 16384,
-      supportsVision: true
-    }
-  },
-  {
-    pattern: "llava",
-    profile: {
-      name: "LLaVA",
-      temperature: 0.7,
-      topP: 0.9,
-      supportsReasoning: false,
-      reasoningAsContent: false,
-      defaultMaxTokens: 16384,
-      supportsVision: true
-    }
-  },
-  {
-    pattern: "llama",
-    profile: {
-      name: "Llama",
-      temperature: 0.7,
-      topP: 0.9,
-      supportsReasoning: false,
-      reasoningAsContent: false,
-      defaultMaxTokens: 16384,
-      supportsVision: false
-    }
-  },
-  {
-    pattern: "claude",
-    profile: {
-      name: "Claude",
-      temperature: 0.7,
-      topP: 0.9,
-      supportsReasoning: true,
-      reasoningAsContent: false,
-      defaultMaxTokens: 16384,
-      supportsVision: false
-    }
-  },
-  {
-    pattern: "gemma-4",
-    profile: {
-      name: "Gemma 4",
-      temperature: 0.7,
-      topP: 0.9,
-      supportsReasoning: true,
-      reasoningAsContent: false,
-      defaultMaxTokens: 16384,
-      supportsVision: true
-    }
-  }
-];
-function getModelProfile(modelName) {
-  const lowerName = modelName.toLowerCase();
-  if (lowerName.includes("mock")) {
-    return MOCK_PROFILE;
-  }
-  for (const { pattern, profile } of MODEL_PROFILES) {
-    if (lowerName.includes(pattern.toLowerCase())) {
-      return profile;
-    }
-  }
-  return DEFAULT_PROFILE;
-}
-function modelSupportsVision(modelName) {
-  return getModelProfile(modelName).supportsVision;
-}
-
 // src/server/llm/client-pure.ts
 function buildModelParams(params) {
   return {
@@ -408,7 +372,14 @@ function convertToolCalls(toolCalls) {
     }
   }));
 }
-function buildAssistantMessage(msg) {
+function getThinking(msg, override) {
+  if (override) {
+    const val = msg[override];
+    if (val) return val;
+  }
+  return msg["reasoning"] ?? msg["reasoning_content"] ?? msg["thinking"];
+}
+function buildAssistantMessage(msg, thinkingField) {
   const result = {
     role: "assistant",
     content: msg.content || null
@@ -417,7 +388,7 @@ function buildAssistantMessage(msg) {
     result["tool_calls"] = convertToolCalls(msg.toolCalls);
   }
   if (msg.thinkingContent) {
-    result["reasoning_content"] = msg.thinkingContent;
+    result[thinkingField ?? "reasoning"] = msg.thinkingContent;
   }
   return result;
 }
@@ -433,7 +404,7 @@ function convertAttachmentSync(attachment, modelSupportsVision2) {
     text: `[Image: ${attachment.filename || "image"}] (vision not supported, cannot describe)`
   };
 }
-function convertMessages(messages, modelSupportsVision2) {
+function convertMessages(messages, modelSupportsVision2, thinkingField) {
   const filtered = messages.filter((msg) => {
     return !(msg.role === "assistant" && !msg.content?.trim() && (!msg.toolCalls || msg.toolCalls.length === 0));
   });
@@ -454,7 +425,7 @@ function convertMessages(messages, modelSupportsVision2) {
       };
     }
     if (msg.role === "assistant") {
-      return buildAssistantMessage(msg);
+      return buildAssistantMessage(msg, thinkingField);
     }
     if (msg.role === "user" && msg.attachments && msg.attachments.length > 0) {
       const content = buildAttachmentContent(msg.content, msg.attachments, modelSupportsVision2);
@@ -479,10 +450,10 @@ function convertTools(tools) {
     }
   }));
 }
-async function buildChatCompletionCreateParams(model, request, profile, capabilities, disableThinking, isStreaming) {
+async function buildChatCompletionCreateParams(model, request, profile, capabilities, reasoningEffort, isStreaming, thinkingField) {
   const userVisionOverride = request.modelSettings?.supportsVision;
   const modelSupportsVision2 = userVisionOverride ?? profile.supportsVision ?? false;
-  const convertedMessages = convertMessages(request.messages, modelSupportsVision2);
+  const convertedMessages = convertMessages(request.messages, modelSupportsVision2, thinkingField);
   const temperature = request.modelSettings?.temperature ?? request.temperature ?? profile.temperature;
   const maxTokens = request.modelSettings?.maxTokens ?? request.maxTokens ?? profile.defaultMaxTokens;
   const topP = request.modelSettings?.topP ?? profile.topP;
@@ -502,23 +473,34 @@ async function buildChatCompletionCreateParams(model, request, profile, capabili
     ;
     params["top_k"] = topK;
   }
-  const shouldDisableThinking = disableThinking || request.disableThinking;
-  if (capabilities.supportsChatTemplateKwargs && profile.supportsReasoning && shouldDisableThinking) {
+  const resolvedEffort = reasoningEffort ?? request.reasoningEffort;
+  if (resolvedEffort) {
     ;
-    params["chat_template_kwargs"] = { enable_thinking: false };
+    params["reasoning_effort"] = resolvedEffort;
+    const chatTemplateKwargs = request.modelSettings?.chatTemplateKwargs;
+    if (chatTemplateKwargs) {
+      ;
+      params["chat_template_kwargs"] = chatTemplateKwargs;
+    } else if (capabilities.supportsChatTemplateKwargs) {
+      ;
+      params["chat_template_kwargs"] = {
+        enable_thinking: resolvedEffort !== "none"
+      };
+    }
   }
   const modelParams = buildModelParams({ temperature, topP, topK, maxTokens });
   return { params, modelParams };
 }
 async function buildCreateParamsFromInput(input, isStreaming) {
-  const { model, request, profile, capabilities, disableThinking } = input;
+  const { model, request, profile, capabilities, reasoningEffort, thinkingField } = input;
   return buildChatCompletionCreateParams(
     model,
     request,
     profile,
     capabilities,
-    !!disableThinking,
-    isStreaming
+    reasoningEffort,
+    isStreaming,
+    thinkingField
   );
 }
 var buildNonStreamingCreateParams = (input) => buildCreateParamsFromInput(input, false);
@@ -537,20 +519,6 @@ function mapFinishReason(reason) {
       return "stop";
   }
 }
-function extractThinking(content) {
-  const thinkRegex = /<think>([\s\S]*?)<\/think>/g;
-  let thinkingContent = "";
-  let cleanContent = content;
-  let match;
-  while ((match = thinkRegex.exec(content)) !== null) {
-    thinkingContent += match[1];
-    cleanContent = cleanContent.replace(match[0], "");
-  }
-  return {
-    content: cleanContent.trim(),
-    thinkingContent: thinkingContent.trim() || null
-  };
-}
 
 // src/server/llm/streaming.ts
 async function* streamWithSegments(client, request) {
@@ -630,10 +598,9 @@ async function* streamWithSegments(client, request) {
   const ttft = ((firstTokenTime ?? endTime) - startTime) / 1e3;
   const completionTime = (endTime - (firstTokenTime ?? startTime)) / 1e3;
   const { promptTokens, completionTokens } = response.usage;
-  const effectiveThinkingContent = thinkingContent || content === "" && response.reasoning_content || "";
   return {
     content,
-    thinkingContent: effectiveThinkingContent,
+    thinkingContent,
     toolCalls,
     response,
     segments,
@@ -656,10 +623,10 @@ export {
   detectBackend,
   getBackendDisplayName,
   buildModelParams,
+  getThinking,
   buildNonStreamingCreateParams,
   buildStreamingCreateParams,
   mapFinishReason,
-  extractThinking,
   streamWithSegments
 };
-//# sourceMappingURL=chunk-VIIRNJDT.js.map
+//# sourceMappingURL=chunk-LXX2CPM5.js.map

package/dist/{chunk-EXYMWI7A.js → chunk-OKLV3ZNN.js}

@@ -1,7 +1,7 @@
 import {
   getEventStore,
   updateSessionMetadata
-} from "./chunk-NSBYORD5.js";
+} from "./chunk-DHAKJSE7.js";
 import {
   buildMessagesFromStoredEvents,
   foldPendingConfirmations
@@ -56,7 +56,7 @@ Example inputs and outputs:
 
 User message: {message}`;
 async function generateSessionName(options) {
-  const { userMessage, llmClient, signal } = options;
+  const { userMessage, llmClient, signal, modelSettings, nonThinkingEnabled } = options;
   try {
     logger.debug("Generating session name", { messagePreview: userMessage.slice(0, 50) });
     const prompt = SESSION_NAME_PROMPT.replace("{message}", userMessage);
@@ -72,9 +72,13 @@ async function generateSessionName(options) {
       messages,
       tools: [],
       signal: composedSignal,
-      disableThinking: true
+      // Default to non-thinking (reasoningEffort: 'none') when not explicitly configured
+      // to prevent thinking output in session names. Only skip when user explicitly
+      // set nonThinkingEnabled to false.
+      ...nonThinkingEnabled !== false ? { reasoningEffort: "none" } : {},
+      ...modelSettings ? { modelSettings } : {}
     });
-    let name = response.content.trim();
+    let name = (response.content || response.thinkingContent || "").trim();
     if (name.length > 50) {
       name = name.substring(0, 47) + "...";
     }
@@ -143,4 +147,4 @@ export {
   needsNameGenerationCheck,
   applyGeneratedSessionName
 };
-//# sourceMappingURL=chunk-EXYMWI7A.js.map
+//# sourceMappingURL=chunk-OKLV3ZNN.js.map