openfox 1.6.20 → 1.6.21

package/dist/{auto-compaction-NFLWKJAO.js → auto-compaction-U75EINMB.js}

@@ -2,10 +2,10 @@ import {
   maybeAutoCompactContext,
   performManualContextCompaction,
   resolveCompactionStatsIdentity
-} from "./chunk-KOBGCNUE.js";
+} from "./chunk-HZP2V5Z6.js";
 import "./chunk-NBU6KIOD.js";
 import "./chunk-574HZVLE.js";
-import "./chunk-55N6FAAZ.js";
+import "./chunk-XYD5JXRM.js";
 import "./chunk-PMDJEJYY.js";
 import "./chunk-EEPU4INU.js";
 import "./chunk-DZHZ3UUR.js";
@@ -22,4 +22,4 @@ export {
   performManualContextCompaction,
   resolveCompactionStatsIdentity
 };
-//# sourceMappingURL=auto-compaction-NFLWKJAO.js.map
+//# sourceMappingURL=auto-compaction-U75EINMB.js.map

package/dist/{chat-handler-2RVTVNPF.js → chat-handler-JEMPI5UF.js}

@@ -10,11 +10,11 @@ import {
 } from "./chunk-AV45GQ7B.js";
 import {
   runChatTurn
-} from "./chunk-U3KMU3UE.js";
-import "./chunk-KOBGCNUE.js";
+} from "./chunk-AU33PQAB.js";
+import "./chunk-HZP2V5Z6.js";
 import "./chunk-NBU6KIOD.js";
 import "./chunk-574HZVLE.js";
-import "./chunk-55N6FAAZ.js";
+import "./chunk-XYD5JXRM.js";
 import {
   getEventStore
 } from "./chunk-PMDJEJYY.js";
@@ -61,7 +61,7 @@ async function startChatSession(sessionId, content, deps, options) {
   sessionManager.setRunning(sessionId, true);
   broadcastForSession(sessionId, createSessionRunningMessage(true));
   try {
-    const { maybeAutoCompactContext } = await import("./auto-compaction-NFLWKJAO.js");
+    const { maybeAutoCompactContext } = await import("./auto-compaction-U75EINMB.js");
     await maybeAutoCompactContext({
       sessionManager,
       sessionId,
@@ -170,4 +170,4 @@ export {
   startChatSession,
   stopSessionExecution
 };
-//# sourceMappingURL=chat-handler-2RVTVNPF.js.map
+//# sourceMappingURL=chat-handler-JEMPI5UF.js.map

package/dist/{chunk-IN5EP4ZB.js → chunk-6NPMZFMP.js}

@@ -29,7 +29,7 @@ function isVisionFallbackEnabled() {
 async function ensureVisionFallbackConfigLoaded() {
   if (configLoaded) return;
   try {
-    const { loadGlobalConfig, getVisionFallback } = await import("./config-67AX6CNS.js");
+    const { loadGlobalConfig, getVisionFallback } = await import("./config-SWDAJMQ3.js");
     const { getRuntimeConfig } = await import("./runtime-config-2XJJIMSC.js");
     const runtimeConfig = getRuntimeConfig();
     const mode = runtimeConfig.mode ?? "production";
@@ -135,4 +135,4 @@ export {
   describeImage,
   describeImageFromDataUrl
 };
-//# sourceMappingURL=chunk-IN5EP4ZB.js.map
+//# sourceMappingURL=chunk-6NPMZFMP.js.map

package/dist/{chunk-U3KMU3UE.js → chunk-AU33PQAB.js}

@@ -13,7 +13,7 @@ import {
   getToolRegistryForAgent,
   loadAllAgentsDefault,
   runTopLevelAgentLoop
-} from "./chunk-KOBGCNUE.js";
+} from "./chunk-HZP2V5Z6.js";
 import {
   getCurrentContextWindowId,
   getEventStore
@@ -293,4 +293,4 @@ export {
   runBuilderTurn,
   runVerifierTurn
 };
-//# sourceMappingURL=chunk-U3KMU3UE.js.map
+//# sourceMappingURL=chunk-AU33PQAB.js.map

package/dist/{chunk-QY7BMXWT.js → chunk-B7E3BICY.js}

@@ -69,7 +69,11 @@ function buildSessionStats(messagesWithStats) {
         completionTime: call.completionTime,
         prefillSpeed: call.prefillSpeed,
         generationSpeed: call.generationSpeed,
-        totalTime: call.totalTime
+        totalTime: call.totalTime,
+        ...call.temperature !== void 0 && { temperature: call.temperature },
+        ...call.topP !== void 0 && { topP: call.topP },
+        ...call.topK !== void 0 && { topK: call.topK },
+        ...call.maxTokens !== void 0 && { maxTokens: call.maxTokens }
       });
     }
   }
@@ -121,4 +125,4 @@ function computeSessionStats(messages) {
 export {
   computeSessionStats
 };
-//# sourceMappingURL=chunk-QY7BMXWT.js.map
+//# sourceMappingURL=chunk-B7E3BICY.js.map

package/dist/{chunk-OVLFEBRR.js → chunk-FX4SUY2S.js}

@@ -1,172 +1,15 @@
+import {
+  getBackendCapabilities,
+  getModelProfile
+} from "./chunk-XYD5JXRM.js";
 import {
   describeImageFromDataUrl,
   ensureVisionFallbackConfigLoaded
-} from "./chunk-IN5EP4ZB.js";
+} from "./chunk-6NPMZFMP.js";
 import {
   logger
 } from "./chunk-PNBH3RAX.js";
 
-// 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
-  }
-};
-function getBackendCapabilities(backend) {
-  return BACKEND_CAPABILITIES[backend];
-}
-async function detectBackend(baseUrl, explicitBackend, silent = false) {
-  if (explicitBackend && explicitBackend !== "unknown") {
-    if (silent) {
-      logger.debug("Using explicit backend", { backend: explicitBackend });
-    } else {
-      logger.info("Using explicit backend", { backend: explicitBackend });
-    }
-    return explicitBackend;
-  }
-  const probeUrl = baseUrl.replace(/\/v1\/?$/, "");
-  try {
-    if (await probeOllama(probeUrl)) {
-      if (silent) {
-        logger.debug("Detected Ollama backend", { url: probeUrl });
-      } else {
-        logger.info("Detected Ollama backend", { url: probeUrl });
-      }
-      return "ollama";
-    }
-    if (await probeLlamaCpp(probeUrl)) {
-      if (silent) {
-        logger.debug("Detected llama.cpp backend", { url: probeUrl });
-      } else {
-        logger.info("Detected llama.cpp backend", { url: probeUrl });
-      }
-      return "llamacpp";
-    }
-    if (await probeSGLang(probeUrl)) {
-      if (silent) {
-        logger.debug("Detected SGLang backend", { url: probeUrl });
-      } else {
-        logger.info("Detected SGLang backend", { url: probeUrl });
-      }
-      return "sglang";
-    }
-    if (await probeOpenAI(baseUrl)) {
-      if (silent) {
-        logger.debug("Detected vLLM backend (OpenAI-compatible)", { url: baseUrl });
-      } else {
-        logger.info("Detected vLLM backend (OpenAI-compatible)", { url: baseUrl });
-      }
-      return "vllm";
-    }
-    if (silent) {
-      logger.debug("Could not detect backend, using unknown", { url: baseUrl });
-    } else {
-      logger.warn("Could not detect backend, using unknown", { url: baseUrl });
-    }
-    return "unknown";
-  } catch (error) {
-    if (silent) {
-      logger.debug("Backend detection failed", {
-        url: baseUrl,
-        error: error instanceof Error ? error.message : String(error)
-      });
-    } else {
-      logger.warn("Backend detection failed", {
-        url: baseUrl,
-        error: error instanceof Error ? error.message : String(error)
-      });
-    }
-    return "unknown";
-  }
-}
-async function probeOllama(baseUrl) {
-  try {
-    const response = await fetch(`${baseUrl}/api/tags`, {
-      signal: AbortSignal.timeout(5e3)
-    });
-    return response.ok;
-  } catch {
-    return false;
-  }
-}
-async function probeLlamaCpp(baseUrl) {
-  try {
-    const response = await fetch(`${baseUrl}/health`, {
-      signal: AbortSignal.timeout(5e3)
-    });
-    if (!response.ok) return false;
-    const data = await response.json();
-    return "slots_idle" in data || "slots_processing" in data;
-  } catch {
-    return false;
-  }
-}
-async function probeSGLang(baseUrl) {
-  try {
-    const response = await fetch(`${baseUrl}/get_model_info`, {
-      signal: AbortSignal.timeout(5e3)
-    });
-    return response.ok;
-  } catch {
-    return false;
-  }
-}
-async function probeOpenAI(baseUrl) {
-  try {
-    const url = baseUrl.includes("/v1") ? baseUrl : `${baseUrl}/v1`;
-    const response = await fetch(`${url}/models`, {
-      signal: AbortSignal.timeout(5e3)
-    });
-    return response.ok;
-  } catch {
-    return false;
-  }
-}
-function getBackendDisplayName(backend) {
-  switch (backend) {
-    case "vllm":
-      return "vLLM";
-    case "sglang":
-      return "SGLang";
-    case "ollama":
-      return "Ollama";
-    case "llamacpp":
-      return "llama.cpp";
-    case "opencode-go":
-      return "OpenCode Go";
-    case "unknown":
-      return "Unknown";
-  }
-}
-
 // src/server/llm/models.ts
 var modelCache = /* @__PURE__ */ new Map();
 var llmStatus = "unknown";
@@ -295,203 +138,6 @@ var LLMError = class extends OpenFoxError {
   }
 };
 
-// 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;
-}
-
 // src/server/llm/client-pure.ts
 function buildAttachmentContent(msgContent, attachments, modelSupportsVision) {
   const content = [];
@@ -705,27 +351,37 @@ async function buildChatCompletionCreateParams(model, request, profile, capabili
     onVisionFallbackStart,
     onVisionFallbackDone
   );
+  const temperature = request.temperature ?? profile.temperature;
+  const maxTokens = request.maxTokens ?? profile.defaultMaxTokens;
+  const topP = profile.topP;
+  const topK = capabilities.supportsTopK ? profile.topK : void 0;
   const params = {
     model,
     messages: convertedMessages,
     ...request.tools ? { tools: convertTools(request.tools) } : {},
     ...request.toolChoice ? { tool_choice: request.toolChoice } : {},
-    temperature: request.temperature ?? profile.temperature,
-    max_tokens: request.maxTokens ?? profile.defaultMaxTokens,
-    top_p: profile.topP,
+    temperature,
+    max_tokens: maxTokens,
+    top_p: topP,
     stream: isStreaming,
     ...isStreaming ? { stream_options: { include_usage: true } } : {}
   };
-  if (capabilities.supportsTopK && profile.topK !== void 0) {
+  if (topK !== void 0) {
     ;
-    params["top_k"] = profile.topK;
+    params["top_k"] = topK;
   }
   const shouldDisableThinking = isStreaming ? disableThinking || request.disableThinking : disableThinking;
   if (capabilities.supportsChatTemplateKwargs && profile.supportsReasoning && shouldDisableThinking) {
     ;
     params["chat_template_kwargs"] = { enable_thinking: false };
   }
-  return params;
+  const modelParams = {
+    ...temperature !== void 0 && { temperature },
+    ...topP !== void 0 && { topP },
+    ...topK !== void 0 && { topK },
+    ...maxTokens !== void 0 && { maxTokens }
+  };
+  return { params, modelParams };
 }
 async function buildCreateParamsFromInput(input, isStreaming) {
   const { model, request, profile, capabilities, disableThinking, visionFallbackEnabled = false, onVisionFallbackStart, onVisionFallbackDone } = input;
@@ -812,7 +468,7 @@ function createLLMClient(config, initialBackend = "unknown") {
       try {
         const shouldDisableThinking = disableThinking || request.disableThinking === true;
         await ensureVisionFallbackConfigLoaded();
-        const { isVisionFallbackEnabled } = await import("./vision-fallback-ADYRFFD4.js");
+        const { isVisionFallbackEnabled } = await import("./vision-fallback-VXGCQKLA.js");
         const paramsOptions = {
           model,
           request,
@@ -827,7 +483,7 @@ function createLLMClient(config, initialBackend = "unknown") {
         if (request.onVisionFallbackDone) {
           paramsOptions.onVisionFallbackDone = request.onVisionFallbackDone;
         }
-        const createParams = await buildNonStreamingCreateParams(paramsOptions);
+        const { params: createParams } = await buildNonStreamingCreateParams(paramsOptions);
         const response = await openai.chat.completions.create(createParams, {
           signal: request.signal
         });
@@ -890,7 +546,7 @@ function createLLMClient(config, initialBackend = "unknown") {
       });
       try {
         await ensureVisionFallbackConfigLoaded();
-        const { isVisionFallbackEnabled } = await import("./vision-fallback-ADYRFFD4.js");
+        const { isVisionFallbackEnabled } = await import("./vision-fallback-VXGCQKLA.js");
         const shouldDisableThinking = disableThinking || request.disableThinking === true;
         const createParams = await buildStreamingCreateParams({
           model,
@@ -902,7 +558,8 @@ function createLLMClient(config, initialBackend = "unknown") {
           onVisionFallbackStart: request.onVisionFallbackStart,
           onVisionFallbackDone: request.onVisionFallbackDone
         });
-        const stream = await openai.chat.completions.create(createParams, {
+        const { params: streamingParams } = createParams;
+        const stream = await openai.chat.completions.create(streamingParams, {
           signal: request.signal
         });
         let fullContent = "";
@@ -1097,13 +754,10 @@ function createLLMClient(config, initialBackend = "unknown") {
 
 export {
   SessionNotFoundError,
-  getModelProfile,
-  detectBackend,
-  getBackendDisplayName,
   createLLMClient,
   detectModel,
   getLlmStatus,
   setLlmStatus,
   clearModelCache
 };
-//# sourceMappingURL=chunk-OVLFEBRR.js.map
+//# sourceMappingURL=chunk-FX4SUY2S.js.map

package/dist/{chunk-LTPZ4GTW.js → chunk-HWOYZ245.js}

@@ -27,7 +27,7 @@ Options:
 }
 async function runNetworkSetup(mode) {
   const { loadAuthConfig, saveAuthConfig, encryptPassword } = await import("./auth-JF2HDJ6P.js");
-  const { saveGlobalConfig } = await import("./config-67AX6CNS.js");
+  const { saveGlobalConfig } = await import("./config-SWDAJMQ3.js");
   const { getAuthKeyPath } = await import("./paths-ZRI56O63.js");
   const existingAuth = await loadAuthConfig(mode);
   if (existingAuth) {
@@ -88,7 +88,7 @@ async function runNetworkSetup(mode) {
   console.log("\u2713 Configuration saved!\n");
 }
 async function runConfig(mode) {
-  const { loadGlobalConfig, getActiveProvider, getDefaultModel } = await import("./config-67AX6CNS.js");
+  const { loadGlobalConfig, getActiveProvider, getDefaultModel } = await import("./config-SWDAJMQ3.js");
   const { getGlobalConfigPath } = await import("./paths-ZRI56O63.js");
   const config = await loadGlobalConfig(mode);
   const configPath = getGlobalConfigPath(mode);
@@ -143,18 +143,18 @@ async function runCli(options) {
       break;
     }
     case "provider": {
-      const { runProviderCommand } = await import("./provider-DKGBQHUS.js");
+      const { runProviderCommand } = await import("./provider-CDM4LQAC.js");
       const [, subcommand] = positionals;
       await runProviderCommand(mode, subcommand);
       break;
     }
     default: {
-      const { configFileExists } = await import("./config-67AX6CNS.js");
+      const { configFileExists } = await import("./config-SWDAJMQ3.js");
       const configExists = await configFileExists(mode);
       if (!configExists) {
         await runNetworkSetup(mode);
       }
-      const { runServe } = await import("./serve-B5A52252.js");
+      const { runServe } = await import("./serve-X3PKC27B.js");
       await runServe({
         mode,
         port: values.port ? parseInt(values.port) : void 0,
@@ -167,4 +167,4 @@ async function runCli(options) {
 export {
   runCli
 };
-//# sourceMappingURL=chunk-LTPZ4GTW.js.map
+//# sourceMappingURL=chunk-HWOYZ245.js.map