@qwen-code/qwen-code 0.23.5-dev.codemode-responses.20260915.af4da34719 → 0.23.5-preview.0

package/chunks/{chunk-GSS4SZNG.js → chunk-YP2K3WOS.js}

@@ -4,31 +4,27 @@ import {
   DASHSCOPE_PROXY_BASE_URL,
   DEFAULT_DASHSCOPE_BASE_URL,
   DEFAULT_MAX_RETRIES,
+  buildSessionAwareFetch,
+  getEffectiveReasoning,
+  isOpenRouterHostname,
+  resolveReasoningForModel,
   resolveRequestTimeout
-} from "./chunk-JL6APGBX.js";
-import {
-  buildSessionAwareFetch
-} from "./chunk-C75BXMPM.js";
-import {
-  isQwenFamilyWireModel,
-  isTieredEffortWireModel
-} from "./chunk-7MIBD3LG.js";
+} from "./chunk-KEM7MWS7.js";
 import {
   REASONING_EFFORT_TIERS,
   clampReasoningEffort,
-  getGptReasoningCapabilities,
-  isReasoningEffortPlaceholder,
-  parseModelReasoningCapabilities
-} from "./chunk-ARDHE6MM.js";
-import {
   defaultOutputCeiling,
+  getGptReasoningCapabilities,
   hasExplicitOutputLimit,
+  isQwenFamilyWireModel,
+  isReasoningEffortPlaceholder,
+  isTieredEffortWireModel,
   parsePositiveIntegerEnvValue,
   tokenLimit
-} from "./chunk-FPGTNKCP.js";
+} from "./chunk-GJKP6XOB.js";
 import {
   buildRuntimeFetchOptions
-} from "./chunk-EWJTAP3Y.js";
+} from "./chunk-7O5PYQ3M.js";
 import {
   createDebugLogger
 } from "./chunk-ZAFYHRVY.js";
@@ -44,22 +40,6 @@ import {
 
 // packages/core/src/core/openaiContentGenerator/provider/default.ts
 init_esbuild_shims();
-
-// packages/core/src/core/openaiContentGenerator/provider/openrouter.ts
-init_esbuild_shims();
-function isOpenRouterHostname(contentGeneratorConfig) {
-  const baseUrl = contentGeneratorConfig.baseUrl ?? "";
-  if (!baseUrl) return false;
-  try {
-    const hostname = new URL(baseUrl).hostname.toLowerCase();
-    return hostname === "openrouter.ai" || hostname.endsWith(".openrouter.ai");
-  } catch {
-    return false;
-  }
-}
-__name(isOpenRouterHostname, "isOpenRouterHostname");
-
-// packages/core/src/core/openaiContentGenerator/provider/default.ts
 var debugLogger = createDebugLogger("DefaultOpenAICompatibleProvider");
 var OPENAI_COMPATIBLE_EFFORTS = [
   "low",
@@ -146,13 +126,18 @@ var DefaultOpenAICompatibleProvider = class {
    * `super.buildRequest` path.
    */
   supportedReasoningEffortsFor(model) {
-    const gpt = getGptReasoningCapabilities(model);
-    if (!gpt) return OPENAI_COMPATIBLE_EFFORTS;
-    const { authType, baseUrl } = this.contentGeneratorConfig;
-    const configured = authType && model ? parseModelReasoningCapabilities(
-      this.cliConfig.getResolvedModelConfig?.(authType, model, baseUrl)?.capabilities.reasoning
-    ) : void 0;
-    return configured && !configured.toggleOnly ? configured.efforts : gpt.efforts;
+    const configured = this.getReasoningCapabilities(model);
+    if (!configured?.profile && !getGptReasoningCapabilities(model))
+      return OPENAI_COMPATIBLE_EFFORTS;
+    if (configured && !configured.toggleOnly) return configured.efforts;
+    return getGptReasoningCapabilities(model)?.efforts ?? OPENAI_COMPATIBLE_EFFORTS;
+  }
+  getReasoningCapabilities(model) {
+    return resolveReasoningForModel(
+      this.cliConfig,
+      this.contentGeneratorConfig,
+      model
+    );
   }
   /**
    * Cap the pipeline-injected `reasoning.effort` at what this endpoint accepts.
@@ -166,7 +151,7 @@ var DefaultOpenAICompatibleProvider = class {
    * merges after this, so both explicit overrides survive unchanged.
    */
   clampConfiguredReasoningEffort(request) {
-    if (this.contentGeneratorConfig.samplingParams?.["reasoning"] !== void 0) {
+    if (this.contentGeneratorConfig.samplingParams?.["reasoning"] !== void 0 || this.getReasoningCapabilities(request.model)?.profile) {
       return request;
     }
     const loose = request;
@@ -200,7 +185,8 @@ var DefaultOpenAICompatibleProvider = class {
     const requestWithTokenLimits = this.clampConfiguredReasoningEffort(
       this.applyOutputTokenLimit(request)
     );
-    const messages = isQwen3Model(request.model) ? requestWithTokenLimits.messages.map(mirrorReasoningContentToReasoning) : requestWithTokenLimits.messages;
+    const profile = this.getReasoningCapabilities(request.model)?.profile;
+    const messages = !profile && isQwen3Model(request.model) || profile === "qwen-chat-template" ? requestWithTokenLimits.messages.map(mirrorReasoningContentToReasoning) : requestWithTokenLimits.messages;
     const result = {
       ...requestWithTokenLimits,
       messages,
@@ -210,7 +196,7 @@ var DefaultOpenAICompatibleProvider = class {
     return result;
   }
   flattenGptReasoningEffort(body) {
-    if (!getGptReasoningCapabilities(body["model"]) || isOpenRouterHostname(this.contentGeneratorConfig) || this.contentGeneratorConfig.samplingParams?.["reasoning"] !== void 0 || this.contentGeneratorConfig.extra_body?.["reasoning"] !== void 0)
+    if (this.getReasoningCapabilities(body["model"])?.profile || !getGptReasoningCapabilities(body["model"]) || isOpenRouterHostname(this.contentGeneratorConfig) || this.contentGeneratorConfig.samplingParams?.["reasoning"] !== void 0 || this.contentGeneratorConfig.extra_body?.["reasoning"] !== void 0)
       return;
     const reasoning = body["reasoning"];
     if (typeof reasoning?.effort !== "string" || !reasoning.effort) return;
@@ -607,15 +593,14 @@ var DashScopeOpenAICompatibleProvider = class _DashScopeOpenAICompatibleProvider
     return merged;
   }
   getConfiguredReasoning(model) {
-    const { authType, baseUrl } = this.contentGeneratorConfig;
-    const wireModel = model ?? this.contentGeneratorConfig.model;
-    const reasoning = authType ? this.cliConfig.getResolvedModelConfig?.(authType, wireModel, baseUrl)?.capabilities.reasoning : void 0;
-    return parseModelReasoningCapabilities(reasoning);
+    return this.getReasoningCapabilities(model);
   }
   isTieredEffortModel(model) {
+    const configured = this.getConfiguredReasoning(model);
+    if (configured?.profile) return configured.profile === "dashscope-effort";
     return isTieredEffortWireModel(
       model ?? this.contentGeneratorConfig.model,
-      this.getConfiguredReasoning(model)
+      configured
     );
   }
   resolveWireModel(model) {
@@ -631,15 +616,20 @@ var DashScopeOpenAICompatibleProvider = class _DashScopeOpenAICompatibleProvider
    * model sharing the DashScope endpoint.
    */
   buildQwenEffortConfig(model) {
-    const reasoning = this.contentGeneratorConfig.reasoning;
+    const configured = this.getConfiguredReasoning(model);
+    if (configured?.profile) return {};
+    const reasoning = getEffectiveReasoning(
+      this.contentGeneratorConfig,
+      configured
+    );
     if (!reasoning || reasoning.effort === void 0) {
       return {};
     }
     const wireModel = this.resolveWireModel(model);
     if (this.isTieredEffortModel(model)) {
-      const configured = this.getConfiguredReasoning(model);
-      if (configured && !configured.toggleOnly) {
-        return configured.efforts.includes(reasoning.effort) ? { reasoning_effort: reasoning.effort } : {};
+      const configured2 = this.getConfiguredReasoning(model);
+      if (configured2 && !configured2.toggleOnly) {
+        return configured2.efforts.includes(reasoning.effort) ? { reasoning_effort: reasoning.effort } : {};
       }
       return { reasoning_effort: this.clampTieredEffort(reasoning.effort) };
     }
@@ -1007,14 +997,8 @@ var DashScopeOpenAICompatibleProvider = class _DashScopeOpenAICompatibleProvider
 };
 
 export {
-  isOpenRouterHostname,
   DefaultOpenAICompatibleProvider,
   selectDashScopeThinkingKnob,
   DASHSCOPE_REGIONAL_HOSTS,
   DashScopeOpenAICompatibleProvider
 };
-/**
- * @license
- * Copyright 2025 Qwen
- * SPDX-License-Identifier: Apache-2.0
- */

package/chunks/{chunk-ZZUXUBEB.js → chunk-YQQ2U7AS.js}

@@ -3,10 +3,11 @@
 import {
   SERVE_CONTROL_EXT_METHODS,
   STATUS_SCHEMA_VERSION
-} from "./chunk-54B2LTKP.js";
+} from "./chunk-M23GFJ5B.js";
 import {
+  AcpChildCapacityExceededError,
   WorkspaceDrainingError
-} from "./chunk-FAWTMDX7.js";
+} from "./chunk-YXHY6BKS.js";
 import {
   init_esbuild_shims
 } from "./chunk-5O2XNYP6.js";
@@ -154,7 +155,8 @@ var WorkspaceRuntimeCoordinator = class {
         );
       } catch (error) {
         this.assertAcceptingWork(error);
-        if (error instanceof WorkspaceRuntimeStillStartingError) throw error;
+        if (error instanceof WorkspaceRuntimeStillStartingError || error instanceof AcpChildCapacityExceededError)
+          throw error;
         throw new WorkspaceRuntimeInitializationError(error);
       }
     }
@@ -339,7 +341,7 @@ var WorkspaceRuntimeCoordinator = class {
           try {
             await this.bridge.preheat({ keepAliveMs: ENSURE_KEEP_ALIVE_MS });
           } catch (error) {
-            if (error instanceof WorkspaceRuntimeStillStartingError)
+            if (error instanceof WorkspaceRuntimeStillStartingError || error instanceof AcpChildCapacityExceededError)
               throw error;
             throw new WorkspaceRuntimeInitializationError(error);
           }
@@ -469,7 +471,8 @@ var WorkspaceRuntimeCoordinator = class {
       try {
         await this.bridge.preheat({ keepAliveMs: ENSURE_KEEP_ALIVE_MS });
       } catch (error) {
-        if (error instanceof WorkspaceRuntimeStillStartingError) throw error;
+        if (error instanceof WorkspaceRuntimeStillStartingError || error instanceof AcpChildCapacityExceededError)
+          throw error;
         throw new WorkspaceRuntimeInitializationError(error);
       }
       snapshot = this.bridge.getWorkspaceRuntimeLifecycleSnapshot();

package/chunks/{chunk-FAWTMDX7.js → chunk-YXHY6BKS.js}

@@ -13,6 +13,20 @@ import {
 // packages/acp-bridge/src/bridgeErrors.ts
 init_esbuild_shims();
 var NOT_CURRENTLY_GENERATING_CANCEL_MESSAGE = "Not currently generating";
+var AcpChildCapacityExceededError = class extends Error {
+  constructor(maxConcurrentChildren, committedAcpChildren) {
+    super(
+      "The service has reached its concurrent process limit. Try again later or cancel this operation."
+    );
+    this.maxConcurrentChildren = maxConcurrentChildren;
+    this.committedAcpChildren = committedAcpChildren;
+  }
+  static {
+    __name(this, "AcpChildCapacityExceededError");
+  }
+  name = "AcpChildCapacityExceededError";
+  code = "acp_child_capacity_exhausted";
+};
 var StandaloneSessionSpawnError = class extends Error {
   constructor(dispatched, cause) {
     super("Daemon-owned standalone session creation failed", { cause });
@@ -498,6 +512,7 @@ var McpAuthenticationInProgressError = class extends Error {
 
 export {
   NOT_CURRENTLY_GENERATING_CANCEL_MESSAGE,
+  AcpChildCapacityExceededError,
   StandaloneSessionSpawnError,
   isNotCurrentlyGeneratingCancelError,
   SessionNotFoundError,

package/chunks/{chunk-VIV2I3R7.js → chunk-YZXSDPOQ.js}

@@ -4,7 +4,7 @@ import {
   OpenAIContentConverter,
   TaggedThinkingParser,
   openaiRequestCaptureContext
-} from "./chunk-SROLVJFY.js";
+} from "./chunk-3WWPEFMX.js";
 import {
   applyOfficialOpenAIPromptCaching,
   isOfficialOpenAIEndpoint,
@@ -12,9 +12,8 @@ import {
 } from "./chunk-WFNEY5G7.js";
 import {
   DashScopeOpenAICompatibleProvider,
-  DefaultOpenAICompatibleProvider,
-  isOpenRouterHostname
-} from "./chunk-GSS4SZNG.js";
+  DefaultOpenAICompatibleProvider
+} from "./chunk-YP2K3WOS.js";
 import {
   getCurrentAgentId
 } from "./chunk-BW2SMO6L.js";
@@ -31,7 +30,7 @@ import {
   resolveStreamIdleTimeoutMs,
   resolveStreamMaxLifetimeMs,
   withStreamGuards
-} from "./chunk-HG46TTAG.js";
+} from "./chunk-HXUI5D4H.js";
 import {
   runtimeDiagnostics
 } from "./chunk-VNOVK4I7.js";
@@ -49,32 +48,32 @@ import {
 import {
   getRateLimitErrorDetails,
   getTransportCode
-} from "./chunk-ZKIAGIGH.js";
+} from "./chunk-3XBUCOWS.js";
 import {
   reportOpenAiChunk,
   reportOpenAiRequest,
   reportOpenAiResponse
 } from "./chunk-AXPOOQAW.js";
+import {
+  getEffectiveReasoning,
+  isOpenRouterHostname,
+  resolveReasoningForModel
+} from "./chunk-KEM7MWS7.js";
 import {
   ProtocolTagSanitizedEvent,
   logProtocolTagSanitized
-} from "./chunk-UB2MATP5.js";
-import {
-  isQwenFamilyWireModel,
-  isTieredEffortWireModel
-} from "./chunk-7MIBD3LG.js";
+} from "./chunk-CDDP4ONL.js";
 import {
   REASONING_EFFORT_TIERS,
   getGptReasoningCapabilities,
+  isQwenFamilyWireModel,
   isReasoningEffortPlaceholder,
-  parseModelReasoningCapabilities
-} from "./chunk-ARDHE6MM.js";
-import {
+  isTieredEffortWireModel,
   reconcileMaxTokens
-} from "./chunk-FPGTNKCP.js";
+} from "./chunk-GJKP6XOB.js";
 import {
   redactProxyError
-} from "./chunk-EWJTAP3Y.js";
+} from "./chunk-7O5PYQ3M.js";
 import {
   createDebugLogger,
   getErrorMessage,
@@ -203,7 +202,7 @@ var DeepSeekOpenAICompatibleProvider = class extends DefaultOpenAICompatibleProv
    * `isDeepSeekHostname`. Elsewhere the generic ceiling applies.
    */
   supportedReasoningEffortsFor(model) {
-    return isDeepSeekHostname(this.contentGeneratorConfig) ? REASONING_EFFORT_TIERS : super.supportedReasoningEffortsFor(model);
+    return !this.getReasoningCapabilities(model)?.profile && isDeepSeekHostname(this.contentGeneratorConfig) ? REASONING_EFFORT_TIERS : super.supportedReasoningEffortsFor(model);
   }
   /**
    * Text-only DeepSeek APIs require message content to be a plain string, not
@@ -220,15 +219,15 @@ var DeepSeekOpenAICompatibleProvider = class extends DefaultOpenAICompatibleProv
    */
   buildRequest(request, userPromptId) {
     const baseRequest = super.buildRequest(request, userPromptId);
-    const reshaped = isDeepSeekHostname(this.contentGeneratorConfig) ? translateReasoningEffort(baseRequest) : baseRequest;
+    const profile = this.getReasoningCapabilities(request.model)?.profile;
+    const reshaped = (!profile || this.contentGeneratorConfig.samplingParams?.["reasoning"] !== void 0 || this.contentGeneratorConfig.extra_body?.["reasoning"] !== void 0) && isDeepSeekHostname(this.contentGeneratorConfig) ? translateReasoningEffort(baseRequest) : baseRequest;
     if (!reshaped.messages?.length) {
       return reshaped;
     }
-    const messages = reshaped.messages.map(
-      (message) => ensureReasoningContentOnAssistantMessage(
-        this.contentGeneratorConfig.modalities?.image ? message : flattenContentParts(message)
-      )
-    );
+    const messages = reshaped.messages.map((message) => {
+      const content = this.contentGeneratorConfig.modalities?.image ? message : flattenContentParts(message);
+      return isDeepSeekHostname(this.contentGeneratorConfig) || !profile || profile === "deepseek-openai" ? ensureReasoningContentOnAssistantMessage(content) : content;
+    });
     return {
       ...reshaped,
       messages
@@ -402,6 +401,33 @@ var MiniMaxOpenAICompatibleProvider = class extends DefaultOpenAICompatibleProvi
       return false;
     }
   }
+  /**
+   * MiniMax rejects a function tool that carries no `parameters` at all
+   * (#11834: `400 invalid params, function parameters is empty (2013)`), so
+   * zero-argument tools get an empty object schema injected here.
+   *
+   * This deliberately reverses the converter's invariant one layer down:
+   * converter.ts sets `parameters = undefined` for parameterless tools
+   * (#11431), because the default-provider endpoints #10080 was written for
+   * (llama.cpp, LM Studio, vLLM) reject the empty-object shape. Keep this
+   * MiniMax-scoped: do not hoist it into DefaultOpenAICompatibleProvider,
+   * and do not move it into the converter ahead of
+   * `relaxSchemaForFunctionCalling`, which strips empty `properties` and
+   * would emit the bare `{"type":"object"}` that #11410 reports as a 400.
+   */
+  buildRequest(request, userPromptId) {
+    const baseRequest = super.buildRequest(request, userPromptId);
+    baseRequest.tools = baseRequest.tools?.map(
+      (tool) => tool.function.parameters === void 0 ? {
+        ...tool,
+        function: {
+          ...tool.function,
+          parameters: { type: "object", properties: {} }
+        }
+      } : tool
+    );
+    return baseRequest;
+  }
   getResponseParsingOptions() {
     return { taggedThinkingTags: true };
   }
@@ -443,6 +469,8 @@ var MistralOpenAICompatibleProvider = class extends DefaultOpenAICompatibleProvi
   static isMistralProvider = isMistralProvider;
   buildRequest(request, userPromptId) {
     const baseRequest = super.buildRequest(request, userPromptId);
+    if (!isMistralHostname(this.contentGeneratorConfig) && this.getReasoningCapabilities(request.model)?.profile === "deepseek-openai")
+      return baseRequest;
     return {
       ...baseRequest,
       messages: baseRequest.messages.map(stripReasoningContent)
@@ -548,7 +576,7 @@ var MiMoOpenAICompatibleProvider = class extends DefaultOpenAICompatibleProvider
   static isMiMoProvider = isMiMoProvider;
   buildRequest(request, userPromptId) {
     const baseRequest = super.buildRequest(request, userPromptId);
-    if (!baseRequest.messages?.length) {
+    if (this.getReasoningCapabilities(request.model)?.profile && !isMiMoProvider({ ...this.contentGeneratorConfig, model: "" }) || !baseRequest.messages?.length) {
       return baseRequest;
     }
     return {
@@ -1054,7 +1082,44 @@ function asObject(value) {
   return value;
 }
 __name(asObject, "asObject");
+function profileReasoning(profile, reasoning) {
+  if (reasoning === void 0) return {};
+  const enabled = reasoning !== false;
+  const effort = reasoning && reasoning.effort;
+  switch (profile) {
+    case "openai-reasoning":
+      return { reasoning: enabled ? reasoning : { enabled: false } };
+    case "openai-effort":
+    case "dashscope-effort":
+      return effort || !enabled ? { reasoning_effort: enabled ? effort : "none" } : {};
+    case "deepseek-openai":
+      return {
+        thinking: { type: enabled ? "enabled" : "disabled" },
+        ...effort ? { reasoning_effort: effort } : {}
+      };
+    case "dashscope-thinking":
+      return { enable_thinking: enabled };
+    case "qwen-chat-template":
+      return { chat_template_kwargs: { enable_thinking: enabled } };
+    default:
+      return {};
+  }
+}
+__name(profileReasoning, "profileReasoning");
 function applyConfiguredReasoningEffort(request, capabilities) {
+  if (capabilities?.profile) {
+    const loose2 = request;
+    const { reasoning: reasoning2, ...rest2 } = loose2;
+    const { effort: _effort, ...siblings } = asObject(reasoning2) ?? {};
+    return {
+      ...Object.keys(siblings).length ? { reasoning: siblings } : {},
+      ...profileReasoning(
+        capabilities.profile,
+        reasoning2
+      ),
+      ...rest2
+    };
+  }
   if (!capabilities || capabilities.toggleOnly || !Array.isArray(capabilities.efforts)) {
     return request;
   }
@@ -1555,10 +1620,16 @@ var ContentGenerationPipeline = class {
     return true;
   }
   async buildRequest(request, userPromptId, context, isStreaming) {
-    const messages = OpenAIContentConverter.convertLlmRequestToOpenAI(
+    const reasoningCapabilities = resolveReasoningForModel(
+      this.config.cliConfig,
+      this.contentGeneratorConfig,
+      context.model
+    );
+    const convertedMessages = OpenAIContentConverter.convertLlmRequestToOpenAI(
       request,
       context
     );
+    const messages = reasoningCapabilities?.profile === "deepseek-openai" ? convertedMessages.map(ensureReasoningContentOnAssistantMessage) : convertedMessages;
     let baseRequest = {
       model: context.model,
       messages,
@@ -1571,21 +1642,17 @@ var ContentGenerationPipeline = class {
     } else {
       baseRequest.stream = false;
     }
-    const authType = this.contentGeneratorConfig.authType;
-    const reasoningCapabilities = authType ? parseModelReasoningCapabilities(
-      this.config.cliConfig.getResolvedModelConfig?.(
-        authType,
-        context.model,
-        this.contentGeneratorConfig.baseUrl
-      )?.capabilities.reasoning
-    ) : void 0;
-    if (reasoningCapabilities && !("reasoning" in baseRequest) && this.contentGeneratorConfig.reasoning) {
+    const effectiveReasoning = getEffectiveReasoning(
+      this.contentGeneratorConfig,
+      reasoningCapabilities
+    );
+    if (reasoningCapabilities && !("reasoning" in baseRequest) && effectiveReasoning && request.config?.thinkingConfig?.includeThoughts !== false) {
       baseRequest = {
         ...baseRequest,
-        reasoning: this.contentGeneratorConfig.reasoning
+        reasoning: effectiveReasoning
       };
     }
-    if (this.contentGeneratorConfig.samplingParams?.["reasoning"] === void 0 && !isOpenRouterHostname(this.contentGeneratorConfig)) {
+    if (this.contentGeneratorConfig.samplingParams?.["reasoning"] === void 0 && (!reasoningCapabilities?.profile || this.contentGeneratorConfig.extra_body?.["reasoning"] === void 0) && (reasoningCapabilities?.profile || !isOpenRouterHostname(this.contentGeneratorConfig))) {
       baseRequest = applyConfiguredReasoningEffort(
         baseRequest,
         reasoningCapabilities
@@ -1621,9 +1688,36 @@ var ContentGenerationPipeline = class {
       this.contentGeneratorConfig
     );
     const explicitThinkingMandatory = reasoningCapabilities?.canDisable === false || this.requiresThinking(model);
-    const thinkingMandatory = explicitThinkingMandatory || getGptReasoningCapabilities(model)?.thinkingMandatory === true;
+    const profile = reasoningCapabilities?.profile;
+    const thinkingMandatory = explicitThinkingMandatory || !profile && getGptReasoningCapabilities(model)?.thinkingMandatory === true;
     const reasoningDisabled = request.config?.thinkingConfig?.includeThoughts === false || this.contentGeneratorConfig.reasoning === false;
-    if (reasoningDisabled) {
+    if ((profile === "openai-effort" || profile === "dashscope-effort") && isReasoningEffortPlaceholder(providerRequest.reasoning_effort) && effectiveReasoning && this.contentGeneratorConfig.samplingParams?.["reasoning"] === void 0 && this.contentGeneratorConfig.extra_body?.["reasoning"] === void 0)
+      providerRequest.reasoning_effort = effectiveReasoning.effort;
+    if (reasoningDisabled && profile) {
+      const typed2 = providerRequest;
+      if (!thinkingMandatory || request.config?.thinkingConfig?.includeThoughts === false) {
+        delete typed2["reasoning"];
+        delete typed2["reasoning_effort"];
+      }
+      if (!thinkingMandatory) {
+        if (isDashScope && profile === "dashscope-effort") {
+          delete typed2["thinking_budget"];
+          delete typed2["enable_thinking"];
+        }
+        const template = asObject(typed2["chat_template_kwargs"]);
+        Object.assign(typed2, profileReasoning(profile, false));
+        if (profile === "qwen-chat-template") {
+          delete typed2["enable_thinking"];
+          typed2["chat_template_kwargs"] = {
+            ...template,
+            enable_thinking: false
+          };
+        } else if (reasoningCapabilities?.disableField === "enable_thinking") {
+          delete typed2["reasoning_effort"];
+          typed2["enable_thinking"] = false;
+        }
+      }
+    } else if (reasoningDisabled) {
       const typed2 = providerRequest;
       if (!thinkingMandatory && isQwenFamilyWireModel(model)) {
         if (isDashScope) {
@@ -1788,7 +1882,14 @@ var ContentGenerationPipeline = class {
     if (request.config?.thinkingConfig?.includeThoughts === false) {
       return {};
     }
-    const reasoning = this.contentGeneratorConfig.reasoning;
+    const reasoning = getEffectiveReasoning(
+      this.contentGeneratorConfig,
+      resolveReasoningForModel(
+        this.config.cliConfig,
+        this.contentGeneratorConfig,
+        request.model
+      )
+    );
     if (reasoning === false || reasoning === void 0) {
       return {};
     }

package/chunks/{chunk-KZ63S3HX.js → chunk-ZFHKUQZ4.js}

@@ -2,15 +2,15 @@
 "use strict";
 import {
   appEvents
-} from "./chunk-6IHJLXNF.js";
+} from "./chunk-CXY6BICQ.js";
 import {
   assembleMcpServers,
   getPendingGatedMcpServers,
   getPromptableMcpServers
-} from "./chunk-RD2HAF25.js";
+} from "./chunk-WL36KHUQ.js";
 import {
   getMCPServerStatus
-} from "./chunk-UB2MATP5.js";
+} from "./chunk-CDDP4ONL.js";
 import {
   require_fast_deep_equal
 } from "./chunk-RSWBT7AZ.js";

package/chunks/{chunk-B6KSTADQ.js → chunk-ZMYKNKET.js}

@@ -5,7 +5,7 @@ import {
 } from "./chunk-6H67XLET.js";
 import {
   DEFAULT_RING_SIZE
-} from "./chunk-7U5G6JXI.js";
+} from "./chunk-KWYTSSTJ.js";
 import {
   DEFAULT_COMPACTED_REPLAY_MAX_BYTES,
   DEFAULT_MAX_JOURNAL_BYTES,
@@ -27,10 +27,10 @@ import {
 } from "./chunk-3VUENPWF.js";
 import {
   loadSettings
-} from "./chunk-65BKWMMH.js";
+} from "./chunk-RRI764F7.js";
 import {
   MCP_BUDGET_WARN_FRACTION
-} from "./chunk-AYMSNIYL.js";
+} from "./chunk-KAYC5FBN.js";
 import {
   parsePositiveIntegerEnv
 } from "./chunk-6PVPNMXU.js";
@@ -272,9 +272,9 @@ var serveCommand = {
     default: "observe",
     description: "Whether the daemon derives a memory-pressure level from its own RSS and V8 heap. `observe` (default) reports the level in daemon status and raises a status issue when it leaves normal. `off` still reports the underlying figures but raises no issue, so the overall status rollup is unchanged \u2014 use it while calibrating, or if you alert on the top-level status. Nothing remediates in either mode."
   }).option("child-heap-mode", {
-    choices: ["off", "observe"],
+    choices: ["off", "observe", "admit"],
     default: "observe",
-    description: "Whether the daemon models a per-child heap partition of the memory budget. `observe` (default) reports the partition it would apply \u2014 `limits.memory.childHeap.perChildCeilingMb` and `maxConcurrentChildren` \u2014 and counts spawns that would have exceeded it. Nothing is applied: no child is sized from the budget and no spawn is refused. `off` models nothing. Note a refusal count of 0 does NOT mean the partition would be safe to apply; children still run on the much larger host-derived ceiling, so a workload needing more old space than the modeled ceiling looks healthy here."
+    description: "Whether the daemon models a per-child heap partition of the memory budget. `observe` (default) reports the partition it would apply \u2014 `limits.memory.childHeap.perChildCeilingMb` and `maxConcurrentChildren` \u2014 and counts spawns that would have exceeded it. Nothing is applied: no child is sized from the budget and no spawn is refused. `off` models nothing. Note a refusal count of 0 does NOT mean the partition would be safe to apply; children still run on the much larger host-derived ceiling, so a workload needing more old space than the modeled ceiling looks healthy here. `admit` rejects starts past the modeled process limit but keeps the existing child heap arguments."
   }).option("mcp-client-budget", {
     type: "number",
     description: "Cap on live MCP clients spawned inside the ACP child for the bound workspace. Positive integer. Combine with --mcp-budget-mode to control behavior at the cap. When unset, mode defaults to off (no accounting-driven enforcement, but GET /workspace/mcp still reports `clientCount`). Distinct from claude-code MCP_SERVER_CONNECTION_BATCH_SIZE which gates startup concurrency, not the total client count."
@@ -447,7 +447,7 @@ var serveCommand = {
     delete process.env[EXTERNAL_TOOL_GUARD_TOKEN_ENV];
     const openWithAuth = argv["open-with-auth"];
     const open = argv.open || openWithAuth;
-    const { runQwenServe } = await import("./run-qwen-serve-UFEOUIDD.js");
+    const { runQwenServe } = await import("./run-qwen-serve-ZVEKGS7Y.js");
     try {
       const serveOptions = {
         port: argv.port,