openfox 1.6.20 → 1.6.21

package/dist/chunk-XYD5JXRM.js

@@ -0,0 +1,483 @@
+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/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/streaming.ts
+var XML_TOOL_PATTERNS = ["<tool_call>", "<function=", "</tool_call>", "<parameter="];
+function hasXmlToolPattern(text) {
+  return XML_TOOL_PATTERNS.some((p) => text.includes(p));
+}
+async function* streamWithSegments(client, request) {
+  const xmlAbortController = new AbortController();
+  const combinedSignal = request.signal ? AbortSignal.any([request.signal, xmlAbortController.signal]) : xmlAbortController.signal;
+  let content = "";
+  let thinkingContent = "";
+  let response = null;
+  const segments = [];
+  let currentTextSegment = "";
+  let currentThinkingSegment = "";
+  const startTime = performance.now();
+  let firstTokenTime = null;
+  const flushText = () => {
+    if (currentTextSegment.trim()) {
+      segments.push({ type: "text", content: currentTextSegment });
+    }
+    currentTextSegment = "";
+  };
+  const flushThinking = () => {
+    if (currentThinkingSegment.trim()) {
+      segments.push({ type: "thinking", content: currentThinkingSegment });
+    }
+    currentThinkingSegment = "";
+  };
+  try {
+    for await (const event of client.stream({ ...request, signal: combinedSignal })) {
+      switch (event.type) {
+        case "text_delta":
+          if (firstTokenTime === null) {
+            firstTokenTime = performance.now();
+          }
+          flushThinking();
+          content += event.content;
+          currentTextSegment += event.content;
+          if (hasXmlToolPattern(content)) {
+            xmlAbortController.abort();
+            yield { type: "xml_tool_abort" };
+            return null;
+          }
+          yield { type: "text_delta", content: event.content };
+          break;
+        case "thinking_delta":
+          if (firstTokenTime === null) {
+            firstTokenTime = performance.now();
+          }
+          flushText();
+          thinkingContent += event.content;
+          currentThinkingSegment += event.content;
+          if (hasXmlToolPattern(thinkingContent)) {
+            xmlAbortController.abort();
+            yield { type: "xml_tool_abort" };
+            return null;
+          }
+          yield { type: "thinking_delta", content: event.content };
+          break;
+        case "tool_call_delta":
+          yield {
+            type: "tool_call_delta",
+            index: event.index,
+            ...event.id !== void 0 ? { id: event.id } : {},
+            ...event.name !== void 0 ? { name: event.name } : {},
+            ...event.arguments !== void 0 ? { arguments: event.arguments } : {}
+          };
+          break;
+        case "done":
+          flushThinking();
+          flushText();
+          response = event.response;
+          yield { type: "done", response: event.response };
+          break;
+        case "error":
+          yield { type: "error", error: event.error };
+          return null;
+      }
+    }
+  } catch (error) {
+    if (error instanceof Error && error.name === "AbortError") {
+      return null;
+    }
+    yield { type: "error", error: error instanceof Error ? error.message : "Unknown error" };
+    return null;
+  }
+  if (!response) {
+    return null;
+  }
+  const toolCalls = response.toolCalls ?? [];
+  for (const tc of toolCalls) {
+    segments.push({ type: "tool_call", toolCallId: tc.id });
+  }
+  const endTime = performance.now();
+  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,
+    toolCalls,
+    response,
+    segments,
+    timing: {
+      ttft,
+      completionTime,
+      tps: completionTime > 0 ? completionTokens / completionTime : 0,
+      prefillTps: ttft > 0 ? promptTokens / ttft : 0
+    }
+  };
+}
+
+export {
+  getModelProfile,
+  getBackendCapabilities,
+  detectBackend,
+  getBackendDisplayName,
+  streamWithSegments
+};
+//# sourceMappingURL=chunk-XYD5JXRM.js.map

package/dist/cli/dev.js

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import {
   runCli
-} from "../chunk-LTPZ4GTW.js";
+} from "../chunk-HWOYZ245.js";
 import {
   logger
 } from "../chunk-PNBH3RAX.js";

package/dist/cli/index.js

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import {
   runCli
-} from "../chunk-LTPZ4GTW.js";
+} from "../chunk-HWOYZ245.js";
 import {
   logger
 } from "../chunk-PNBH3RAX.js";

package/dist/{config-67AX6CNS.js → config-SWDAJMQ3.js}

@@ -13,11 +13,11 @@ import {
   saveGlobalConfig,
   setDefaultModelSelection,
   trySmartDefaults
-} from "./chunk-ZDNXCVW4.js";
-import "./chunk-OVLFEBRR.js";
-import "./chunk-55N6FAAZ.js";
+} from "./chunk-XRGMFOVG.js";
+import "./chunk-FX4SUY2S.js";
+import "./chunk-XYD5JXRM.js";
 import "./chunk-R4HADRYO.js";
-import "./chunk-IN5EP4ZB.js";
+import "./chunk-6NPMZFMP.js";
 import "./chunk-PNBH3RAX.js";
 export {
   activateProvider,
@@ -35,4 +35,4 @@ export {
   setDefaultModelSelection,
   trySmartDefaults
 };
-//# sourceMappingURL=config-67AX6CNS.js.map
+//# sourceMappingURL=config-SWDAJMQ3.js.map

package/dist/{orchestrator-ZICQ5NIZ.js → orchestrator-T2IPRZ7B.js}

@@ -3,7 +3,7 @@ import {
   runBuilderTurn,
   runChatTurn,
   runVerifierTurn
-} from "./chunk-U3KMU3UE.js";
+} from "./chunk-AU33PQAB.js";
 import {
   TurnMetrics,
   createChatDoneEvent,
@@ -11,10 +11,10 @@ import {
   createMessageStartEvent,
   createToolCallEvent,
   createToolResultEvent
-} 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";
@@ -38,4 +38,4 @@ export {
   runChatTurn,
   runVerifierTurn
 };
-//# sourceMappingURL=orchestrator-ZICQ5NIZ.js.map
+//# sourceMappingURL=orchestrator-T2IPRZ7B.js.map

package/dist/package.json

@@ -1,6 +1,6 @@
 {
   "name": "openfox",
-  "version": "1.6.20",
+  "version": "1.6.21",
   "description": "Local-LLM-first agentic coding assistant",
   "type": "module",
   "bin": {

package/dist/{processor-MPSYT534.js → processor-VKFEOK46.js}

@@ -175,7 +175,7 @@ var QueueProcessor = class {
       backend: provider?.backend ?? llmClient.getBackend(),
       model: llmClient.getModel()
     };
-    const { runChatTurn } = await import("./orchestrator-ZICQ5NIZ.js");
+    const { runChatTurn } = await import("./orchestrator-T2IPRZ7B.js");
     const runChatTurnParams = buildRunChatTurnParams({
       sessionManager,
       sessionId,
@@ -208,4 +208,4 @@ var QueueProcessor = class {
 export {
   QueueProcessor
 };
-//# sourceMappingURL=processor-MPSYT534.js.map
+//# sourceMappingURL=processor-VKFEOK46.js.map

package/dist/{protocol-I7rn7Msg.d.ts → protocol-BYM6rZvW.d.ts}

@@ -106,6 +106,10 @@ interface LLMCallStats {
     generationSpeed: number;
     totalTime: number;
     timestamp?: string;
+    temperature?: number;
+    topP?: number;
+    topK?: number;
+    maxTokens?: number;
 }
 interface StatsDataPoint {
     messageId: string;
@@ -142,6 +146,10 @@ interface CallStatsDataPoint {
     prefillSpeed: number;
     generationSpeed: number;
     totalTime: number;
+    temperature?: number;
+    topP?: number;
+    topK?: number;
+    maxTokens?: number;
 }
 interface SessionStats {
     totalTime: number;
@@ -405,6 +413,10 @@ interface ModelConfig {
     id: string;
     contextWindow: number;
     source: 'backend' | 'user' | 'default';
+    temperature?: number;
+    topP?: number;
+    topK?: number;
+    maxTokens?: number;
 }
 /** LLM provider configuration */
 interface Provider {

package/dist/{provider-DKGBQHUS.js → provider-CDM4LQAC.js}

@@ -5,17 +5,18 @@ import {
   loadGlobalConfig,
   removeProvider,
   saveGlobalConfig
-} from "./chunk-ZDNXCVW4.js";
+} from "./chunk-XRGMFOVG.js";
 import {
   fetchAvailableModelsFromBackend
-} from "./chunk-KOUMYBYM.js";
+} from "./chunk-WINI5HX7.js";
 import {
-  detectBackend,
   detectModel
-} from "./chunk-OVLFEBRR.js";
-import "./chunk-55N6FAAZ.js";
+} from "./chunk-FX4SUY2S.js";
+import {
+  detectBackend
+} from "./chunk-XYD5JXRM.js";
 import "./chunk-R4HADRYO.js";
-import "./chunk-IN5EP4ZB.js";
+import "./chunk-6NPMZFMP.js";
 import "./chunk-PNBH3RAX.js";
 
 // src/cli/provider.ts
@@ -248,7 +249,7 @@ async function runProviderAdd(mode) {
     isActive: makeActive
   });
   if (makeActive) {
-    const { setDefaultModelSelection } = await import("./config-67AX6CNS.js");
+    const { setDefaultModelSelection } = await import("./config-SWDAJMQ3.js");
     newConfig = setDefaultModelSelection(newConfig, newConfig.providers[newConfig.providers.length - 1].id, selectedModel);
   }
   await saveGlobalConfig(mode, newConfig);
@@ -372,4 +373,4 @@ export {
   runProviderRemove,
   runProviderUse
 };
-//# sourceMappingURL=provider-DKGBQHUS.js.map
+//# sourceMappingURL=provider-CDM4LQAC.js.map

package/dist/{serve-B5A52252.js → serve-X3PKC27B.js}

@@ -2,19 +2,19 @@ import {
   getActiveProvider,
   getDefaultModel,
   loadGlobalConfig
-} from "./chunk-ZDNXCVW4.js";
+} from "./chunk-XRGMFOVG.js";
 import {
   VERSION,
   createServer
-} from "./chunk-R7UAGXQW.js";
+} from "./chunk-SJVEX2NR.js";
 import "./chunk-AV45GQ7B.js";
-import "./chunk-U3KMU3UE.js";
-import "./chunk-KOBGCNUE.js";
+import "./chunk-AU33PQAB.js";
+import "./chunk-HZP2V5Z6.js";
 import "./chunk-NBU6KIOD.js";
 import "./chunk-574HZVLE.js";
-import "./chunk-KOUMYBYM.js";
-import "./chunk-OVLFEBRR.js";
-import "./chunk-55N6FAAZ.js";
+import "./chunk-WINI5HX7.js";
+import "./chunk-FX4SUY2S.js";
+import "./chunk-XYD5JXRM.js";
 import "./chunk-PMDJEJYY.js";
 import "./chunk-EEPU4INU.js";
 import "./chunk-DZHZ3UUR.js";
@@ -22,7 +22,7 @@ import "./chunk-22CTURMH.js";
 import "./chunk-RN2O6JK7.js";
 import "./chunk-XKFPU2FA.js";
 import "./chunk-3EHGGBWE.js";
-import "./chunk-QY7BMXWT.js";
+import "./chunk-B7E3BICY.js";
 import "./chunk-Y6HBEACI.js";
 import {
   ensureDataDirExists,
@@ -32,7 +32,7 @@ import {
 import {
   loadConfig
 } from "./chunk-TVQOONDR.js";
-import "./chunk-IN5EP4ZB.js";
+import "./chunk-6NPMZFMP.js";
 import {
   logger
 } from "./chunk-PNBH3RAX.js";
@@ -188,4 +188,4 @@ async function runServe(options) {
 export {
   runServe
 };
-//# sourceMappingURL=serve-B5A52252.js.map
+//# sourceMappingURL=serve-X3PKC27B.js.map