converse-mcp-server 2.29.2 → 3.0.1

package/src/providers/copilot.js

@@ -26,7 +26,6 @@ const SUPPORTED_MODELS = {
     maxOutputTokens: 16384,
     supportsStreaming: true,
     supportsImages: false,
-    supportsTemperature: false,
     supportsWebSearch: false,
     timeout: 600000,
     description:
@@ -35,223 +34,72 @@ const SUPPORTED_MODELS = {
   },
 
   // OpenAI models
-  'gpt-4.1': {
-    modelName: 'gpt-4.1',
-    friendlyName: 'GPT-4.1 (via Copilot)',
+  // Bare `gpt-5.6` (and the legacy `gpt-5` shortcut) route to Sol, matching
+  // Copilot's own bare-alias behavior. Terra is the recommended balanced tier.
+  'gpt-5.6-sol': {
+    modelName: 'gpt-5.6-sol',
+    friendlyName: 'GPT-5.6 Sol (via Copilot)',
     contextWindow: 1047576,
     maxOutputTokens: 32768,
     supportsStreaming: true,
     supportsImages: false,
-    supportsTemperature: false,
-    supportsWebSearch: false,
-    timeout: 600000,
-    description: 'OpenAI GPT-4.1 via Copilot subscription',
-    aliases: [],
-  },
-  'gpt-5-mini': {
-    modelName: 'gpt-5-mini',
-    friendlyName: 'GPT-5 Mini (via Copilot)',
-    contextWindow: 1047576,
-    maxOutputTokens: 16384,
-    supportsStreaming: true,
-    supportsImages: false,
-    supportsTemperature: false,
     supportsWebSearch: false,
     supportsReasoningEffort: true,
     timeout: 600000,
-    description: 'OpenAI GPT-5 Mini via Copilot subscription',
-    aliases: [],
+    description: 'OpenAI GPT-5.6 Sol via Copilot subscription',
+    aliases: ['gpt-5.6', 'gpt-5'],
   },
-  'gpt-5.1': {
-    modelName: 'gpt-5.1',
-    friendlyName: 'GPT-5.1 (via Copilot)',
+  'gpt-5.6-terra': {
+    modelName: 'gpt-5.6-terra',
+    friendlyName: 'GPT-5.6 Terra (via Copilot)',
     contextWindow: 1047576,
     maxOutputTokens: 32768,
     supportsStreaming: true,
     supportsImages: false,
-    supportsTemperature: false,
     supportsWebSearch: false,
     supportsReasoningEffort: true,
     timeout: 600000,
-    description: 'OpenAI GPT-5.1 via Copilot subscription',
+    description: 'OpenAI GPT-5.6 Terra via Copilot subscription',
     aliases: [],
   },
-  'gpt-5.1-codex': {
-    modelName: 'gpt-5.1-codex',
-    friendlyName: 'GPT-5.1 Codex (via Copilot)',
-    contextWindow: 192000,
-    maxOutputTokens: 16384,
-    supportsStreaming: true,
-    supportsImages: false,
-    supportsTemperature: false,
-    supportsWebSearch: false,
-    supportsReasoningEffort: true,
-    timeout: 600000,
-    description: 'OpenAI GPT-5.1 Codex via Copilot subscription',
-    aliases: [],
-  },
-  'gpt-5.1-codex-mini': {
-    modelName: 'gpt-5.1-codex-mini',
-    friendlyName: 'GPT-5.1 Codex Mini (via Copilot)',
-    contextWindow: 192000,
-    maxOutputTokens: 16384,
-    supportsStreaming: true,
-    supportsImages: false,
-    supportsTemperature: false,
-    supportsWebSearch: false,
-    supportsReasoningEffort: true,
-    timeout: 600000,
-    description: 'OpenAI GPT-5.1 Codex Mini via Copilot subscription',
-    aliases: [],
-  },
-  'gpt-5.1-codex-max': {
-    modelName: 'gpt-5.1-codex-max',
-    friendlyName: 'GPT-5.1 Codex Max (via Copilot)',
-    contextWindow: 192000,
-    maxOutputTokens: 16384,
-    supportsStreaming: true,
-    supportsImages: false,
-    supportsTemperature: false,
-    supportsWebSearch: false,
-    supportsReasoningEffort: true,
-    timeout: 600000,
-    description: 'OpenAI GPT-5.1 Codex Max via Copilot subscription',
-    aliases: [],
-  },
-  'gpt-5.4': {
-    modelName: 'gpt-5.4',
-    friendlyName: 'GPT-5.4 (via Copilot)',
+  'gpt-5.6-luna': {
+    modelName: 'gpt-5.6-luna',
+    friendlyName: 'GPT-5.6 Luna (via Copilot)',
     contextWindow: 1047576,
     maxOutputTokens: 32768,
     supportsStreaming: true,
     supportsImages: false,
-    supportsTemperature: false,
-    supportsWebSearch: false,
-    supportsReasoningEffort: true,
-    timeout: 600000,
-    description: 'OpenAI GPT-5.4 via Copilot subscription',
-    aliases: ['gpt-5', 'codex'],
-  },
-  'gpt-5.2-codex': {
-    modelName: 'gpt-5.2-codex',
-    friendlyName: 'GPT-5.2 Codex (via Copilot)',
-    contextWindow: 192000,
-    maxOutputTokens: 16384,
-    supportsStreaming: true,
-    supportsImages: false,
-    supportsTemperature: false,
     supportsWebSearch: false,
     supportsReasoningEffort: true,
     timeout: 600000,
-    description: 'OpenAI GPT-5.2 Codex via Copilot subscription',
-    aliases: [],
-  },
-  'gpt-5.3-codex': {
-    modelName: 'gpt-5.3-codex',
-    friendlyName: 'GPT-5.3 Codex (via Copilot)',
-    contextWindow: 192000,
-    maxOutputTokens: 16384,
-    supportsStreaming: true,
-    supportsImages: false,
-    supportsTemperature: false,
-    supportsWebSearch: false,
-    supportsReasoningEffort: true,
-    timeout: 600000,
-    description: 'OpenAI GPT-5.3 Codex via Copilot subscription',
+    description: 'OpenAI GPT-5.6 Luna via Copilot subscription',
     aliases: [],
   },
 
   // Anthropic models
-  'claude-haiku-4.5': {
-    modelName: 'claude-haiku-4.5',
-    friendlyName: 'Claude Haiku 4.5 (via Copilot)',
-    contextWindow: 200000,
-    maxOutputTokens: 8192,
-    supportsStreaming: true,
-    supportsImages: false,
-    supportsTemperature: false,
-    supportsWebSearch: false,
-    timeout: 600000,
-    description: 'Anthropic Claude Haiku 4.5 via Copilot subscription',
-    aliases: ['haiku'],
-  },
-  'claude-sonnet-4': {
-    modelName: 'claude-sonnet-4',
-    friendlyName: 'Claude Sonnet 4 (via Copilot)',
-    contextWindow: 200000,
-    maxOutputTokens: 16384,
-    supportsStreaming: true,
-    supportsImages: false,
-    supportsTemperature: false,
-    supportsWebSearch: false,
-    timeout: 600000,
-    description: 'Anthropic Claude Sonnet 4 via Copilot subscription',
-    aliases: [],
-  },
-  'claude-sonnet-4.5': {
-    modelName: 'claude-sonnet-4.5',
-    friendlyName: 'Claude Sonnet 4.5 (via Copilot)',
-    contextWindow: 200000,
-    maxOutputTokens: 16384,
-    supportsStreaming: true,
-    supportsImages: false,
-    supportsTemperature: false,
-    supportsWebSearch: false,
-    timeout: 600000,
-    description: 'Anthropic Claude Sonnet 4.5 via Copilot subscription',
-    aliases: [],
-  },
-  'claude-sonnet-4.6': {
-    modelName: 'claude-sonnet-4.6',
-    friendlyName: 'Claude Sonnet 4.6 (via Copilot)',
-    contextWindow: 200000,
-    maxOutputTokens: 16384,
-    supportsStreaming: true,
-    supportsImages: false,
-    supportsTemperature: false,
-    supportsWebSearch: false,
-    timeout: 600000,
-    description: 'Anthropic Claude Sonnet 4.6 via Copilot subscription',
-    aliases: ['sonnet'],
-  },
-  'claude-opus-4.5': {
-    modelName: 'claude-opus-4.5',
-    friendlyName: 'Claude Opus 4.5 (via Copilot)',
-    contextWindow: 200000,
-    maxOutputTokens: 32768,
-    supportsStreaming: true,
-    supportsImages: false,
-    supportsTemperature: false,
-    supportsWebSearch: false,
-    timeout: 600000,
-    description: 'Anthropic Claude Opus 4.5 via Copilot subscription',
-    aliases: [],
-  },
-  'claude-opus-4.6': {
-    modelName: 'claude-opus-4.6',
-    friendlyName: 'Claude Opus 4.6 (via Copilot)',
+  'claude-fable-5': {
+    modelName: 'claude-fable-5',
+    friendlyName: 'Claude Fable 5 (via Copilot)',
     contextWindow: 200000,
     maxOutputTokens: 32768,
     supportsStreaming: true,
     supportsImages: false,
-    supportsTemperature: false,
     supportsWebSearch: false,
     timeout: 600000,
-    description: 'Anthropic Claude Opus 4.6 via Copilot subscription',
-    aliases: [],
+    description: 'Anthropic Claude Fable 5 via Copilot subscription',
+    aliases: ['fable'],
   },
-  'claude-opus-4.7': {
-    modelName: 'claude-opus-4.7',
-    friendlyName: 'Claude Opus 4.7 (via Copilot)',
+  'claude-sonnet-5': {
+    modelName: 'claude-sonnet-5',
+    friendlyName: 'Claude Sonnet 5 (via Copilot)',
     contextWindow: 200000,
     maxOutputTokens: 32768,
     supportsStreaming: true,
     supportsImages: false,
-    supportsTemperature: false,
     supportsWebSearch: false,
     timeout: 600000,
-    description: 'Anthropic Claude Opus 4.7 via Copilot subscription',
-    aliases: [],
+    description: 'Anthropic Claude Sonnet 5 via Copilot subscription',
+    aliases: ['sonnet'],
   },
   'claude-opus-4.8': {
     modelName: 'claude-opus-4.8',
@@ -260,7 +108,6 @@ const SUPPORTED_MODELS = {
     maxOutputTokens: 32768,
     supportsStreaming: true,
     supportsImages: false,
-    supportsTemperature: false,
     supportsWebSearch: false,
     timeout: 600000,
     description: 'Anthropic Claude Opus 4.8 via Copilot subscription',
@@ -268,31 +115,30 @@ const SUPPORTED_MODELS = {
   },
 
   // Google models
-  'gemini-3-pro-preview': {
-    modelName: 'gemini-3-pro-preview',
-    friendlyName: 'Gemini 3 Pro Preview (via Copilot)',
+  // GA `gemini-3.1-pro` does not exist upstream — keep the -preview suffix.
+  'gemini-3.1-pro-preview': {
+    modelName: 'gemini-3.1-pro-preview',
+    friendlyName: 'Gemini 3.1 Pro Preview (via Copilot)',
     contextWindow: 1048576,
     maxOutputTokens: 65536,
     supportsStreaming: true,
     supportsImages: false,
-    supportsTemperature: false,
     supportsWebSearch: false,
     timeout: 600000,
-    description: 'Google Gemini 3 Pro Preview via Copilot subscription',
-    aliases: ['gemini-3-pro'],
+    description: 'Google Gemini 3.1 Pro Preview via Copilot subscription',
+    aliases: ['gemini', 'gemini-3.1-pro'],
   },
-  'gemini-3.1-pro-preview': {
-    modelName: 'gemini-3.1-pro-preview',
-    friendlyName: 'Gemini 3.1 Pro Preview (via Copilot)',
+  'gemini-3.5-flash': {
+    modelName: 'gemini-3.5-flash',
+    friendlyName: 'Gemini 3.5 Flash (via Copilot)',
     contextWindow: 1048576,
     maxOutputTokens: 65536,
     supportsStreaming: true,
     supportsImages: false,
-    supportsTemperature: false,
     supportsWebSearch: false,
     timeout: 600000,
-    description: 'Google Gemini 3.1 Pro Preview via Copilot subscription',
-    aliases: ['gemini', 'gemini-3.1-pro'],
+    description: 'Google Gemini 3.5 Flash via Copilot subscription',
+    aliases: ['gemini-flash'],
   },
 };
 
@@ -892,8 +738,6 @@ export const copilotProvider = {
       stream = false,
       signal,
       reasoning_effort,
-      temperature,
-      use_websearch,
     } = options;
 
     if (!config) {
@@ -903,16 +747,6 @@ export const copilotProvider = {
       );
     }
 
-    if (temperature !== undefined) {
-      debugLog(
-        '[Copilot SDK] Parameter "temperature" not supported by Copilot SDK (ignored)',
-      );
-    }
-    if (use_websearch) {
-      debugLog(
-        '[Copilot SDK] Parameter "use_websearch" not supported by Copilot SDK (ignored)',
-      );
-    }
     try {
       const cwd = config.server?.client_cwd || process.cwd();
       const client = await getCopilotClient(cwd, config);

package/src/providers/deepseek.js

@@ -8,61 +8,48 @@
 import { createOpenAICompatibleProvider } from './openai-compatible.js';
 import { debugLog } from '../utils/console.js';
 
-// Define supported DeepSeek models with their capabilities
+// Define supported DeepSeek models with their capabilities.
+// V4 unified the catalog: both tiers share a 1M context window and a 384K output
+// ceiling, are text-only (no vision), and expose thinking mode via the
+// top-level `thinking` + `reasoning_effort` request fields (see transformRequest).
+// The legacy `deepseek-chat` / `deepseek-reasoner` IDs hard-retire 2026-07-24
+// 15:59 UTC upstream; they are intentionally absent here but keep working as
+// explicit pass-through IDs (resolveModelName returns unknown IDs unchanged)
+// until that date.
 const SUPPORTED_MODELS = {
-  'deepseek-chat': {
-    modelName: 'deepseek-chat',
-    friendlyName: 'DeepSeek Chat (V3-0324)',
-    contextWindow: 128000, // API supports 128K context
-    maxOutputTokens: 8000, // Maximum 8K output tokens
-    defaultMaxTokens: 4000, // Default 4K tokens
+  'deepseek-v4-pro': {
+    modelName: 'deepseek-v4-pro',
+    friendlyName: 'DeepSeek V4 Pro',
+    contextWindow: 1000000, // 1M context (V4 default across all official services)
+    maxOutputTokens: 384000, // 384K max output ceiling
+    defaultMaxTokens: 8000, // Conservative default despite the high ceiling
     supportsStreaming: true,
-    supportsImages: false,
-    supportsTemperature: true,
+    supportsImages: false, // Text-only; V4 has no vision input
     supportsWebSearch: false,
+    supportsReasoning: true,
     supportsJsonOutput: true,
     supportsFunctionCalling: true,
-    supportsChatPrefixCompletion: true, // Beta
-    supportsFIMCompletion: true, // Beta
-    timeout: 300000,
+    timeout: 600000, // Longer timeout for reasoning
     description:
-      'DeepSeek-V3-0324 - Strong MoE model with 671B total/37B active parameters',
-    aliases: [
-      'deepseek',
-      'chat',
-      'deepseek chat',
-      'deepseek-v3',
-      'deepseek-chat-v3',
-      'deepseek v3',
-    ],
+      'DeepSeek V4 Pro - flagship MoE model with 1M context and thinking mode',
+    aliases: ['deepseek', 'deepseek-pro'],
   },
-  'deepseek-reasoner': {
-    modelName: 'deepseek-reasoner',
-    friendlyName: 'DeepSeek Reasoner (R1-0528)',
-    contextWindow: 128000, // API supports 128K context
-    maxOutputTokens: 64000, // Maximum 64K output tokens (including CoT)
-    defaultMaxTokens: 32000, // Default 32K tokens
+  'deepseek-v4-flash': {
+    modelName: 'deepseek-v4-flash',
+    friendlyName: 'DeepSeek V4 Flash',
+    contextWindow: 1000000, // 1M context
+    maxOutputTokens: 384000, // 384K max output ceiling
+    defaultMaxTokens: 8000, // Conservative default despite the high ceiling
     supportsStreaming: true,
-    supportsImages: false,
-    supportsTemperature: true,
+    supportsImages: false, // Text-only; V4 has no vision input
     supportsWebSearch: false,
+    supportsReasoning: true,
     supportsJsonOutput: true,
     supportsFunctionCalling: true,
-    supportsChatPrefixCompletion: true, // Beta
-    supportsFIMCompletion: false, // Not supported
-    supportsReasoning: true,
     timeout: 600000, // Longer timeout for reasoning
     description:
-      'DeepSeek-R1-0528 - Advanced reasoning model with CoT capabilities',
-    aliases: [
-      'deepseek reasoner',
-      'reasoner',
-      'r1',
-      'deepseek r1',
-      'deepseek-r1',
-      'deepseek-reasoner-r1',
-      'deepseek-r1-0528',
-    ],
+      'DeepSeek V4 Flash - faster, lower-cost V4 tier with 1M context and thinking mode',
+    aliases: ['deepseek-flash'],
   },
 };
 
@@ -79,20 +66,51 @@ function validateApiKey(apiKey) {
 }
 
 /**
- * Transform request to handle DeepSeek-specific requirements
+ * Map a Converse reasoning_effort level to DeepSeek's thinking-mode request
+ * fields. DeepSeek exposes only two independent controls: a `thinking` toggle
+ * ({type:"enabled"|"disabled"}) and a `reasoning_effort` field that accepts
+ * ONLY "high" or "max". Every enabled level below max maps to "high" (there is
+ * no lower documented tier), preserving enabled-reasoning intent so the default
+ * `medium` runs thinking-on rather than silently disabling it.
+ */
+function applyReasoning(requestPayload, reasoningEffort) {
+  if (reasoningEffort === 'none') {
+    // Disable thinking entirely; omit reasoning_effort.
+    requestPayload.thinking = { type: 'disabled' };
+    return;
+  }
+
+  requestPayload.thinking = { type: 'enabled' };
+  requestPayload.reasoning_effort = reasoningEffort === 'max' ? 'max' : 'high';
+}
+
+/**
+ * Transform request to handle DeepSeek-specific requirements.
+ *
+ * Reads the requested reasoning effort from the shared base's widened context
+ * and builds the thinking-mode fields. Reasoning is capability-gated: the fields
+ * are attached only when the resolved model supports reasoning, so an
+ * unknown/retired pass-through ID never receives them.
  */
-async function transformRequest(requestPayload) {
-  // DeepSeek uses standard OpenAI format, no transformation needed
+async function transformRequest(requestPayload, context = {}) {
+  const { modelConfig, reasoningEffort } = context;
+
+  if (modelConfig?.supportsReasoning && reasoningEffort) {
+    applyReasoning(requestPayload, reasoningEffort);
+  }
+
   debugLog('[DeepSeek] Request payload prepared');
   return requestPayload;
 }
 
 /**
- * Transform response to handle DeepSeek-specific fields
+ * Transform response to handle DeepSeek-specific fields.
+ *
+ * `reasoning_content` (the thinking-mode sibling of `content`) is surfaced to
+ * metadata by the shared OpenAI-compatible base when supportsReasoning is set,
+ * so no provider-specific reasoning extraction is needed here.
  */
 async function transformResponse(result, rawResponse) {
-  // DeepSeek returns standard OpenAI-compatible responses
-  // Add any DeepSeek-specific metadata if needed
   if (rawResponse.system_fingerprint) {
     result.metadata.system_fingerprint = rawResponse.system_fingerprint;
   }
@@ -101,19 +119,21 @@ async function transformResponse(result, rawResponse) {
 }
 
 /**
- * Create DeepSeek provider using OpenAI-compatible base
+ * Create DeepSeek provider using OpenAI-compatible base.
+ *
+ * baseURL is the bare host (no `/v1`) matching the current official docs and all
+ * OpenAI-SDK code samples. `frequency_penalty`/`presence_penalty` are omitted
+ * from defaultParams: the API now marks them deprecated no-ops, and thinking
+ * mode ignores them (along with temperature/top_p) regardless.
  */
 export const deepseekProvider = createOpenAICompatibleProvider({
-  baseURL: 'https://api.deepseek.com/v1',
+  baseURL: 'https://api.deepseek.com',
   providerName: 'DeepSeek',
   supportedModels: SUPPORTED_MODELS,
   validateApiKey,
   transformRequest,
   transformResponse,
   defaultParams: {
-    // DeepSeek default parameters
     top_p: 0.95,
-    frequency_penalty: 0,
-    presence_penalty: 0,
   },
 });

package/src/providers/gemini-cli.js

@@ -62,7 +62,6 @@ const SUPPORTED_MODELS = {
     maxOutputTokens: 65536,
     supportsStreaming: true,
     supportsImages: false,
-    supportsTemperature: false,
     supportsWebSearch: false,
     supportsThinking: true,
     timeout: DEFAULT_TIMEOUT_MS,
@@ -79,7 +78,6 @@ const SUPPORTED_MODELS = {
     maxOutputTokens: 65536,
     supportsStreaming: true,
     supportsImages: false,
-    supportsTemperature: false,
     supportsWebSearch: false,
     supportsThinking: true,
     timeout: DEFAULT_TIMEOUT_MS,
@@ -95,7 +93,6 @@ const SUPPORTED_MODELS = {
     maxOutputTokens: 65536,
     supportsStreaming: true,
     supportsImages: false,
-    supportsTemperature: false,
     supportsWebSearch: false,
     supportsThinking: true,
     timeout: DEFAULT_TIMEOUT_MS,

package/src/providers/google.js

@@ -17,7 +17,6 @@ const SUPPORTED_MODELS = {
     maxOutputTokens: 65536,
     supportsStreaming: true,
     supportsImages: true,
-    supportsTemperature: true,
     supportsThinking: true,
     supportsWebSearch: true,
     maxThinkingTokens: 24576,
@@ -43,7 +42,6 @@ const SUPPORTED_MODELS = {
     maxOutputTokens: 65536,
     supportsStreaming: true,
     supportsImages: true,
-    supportsTemperature: true,
     supportsThinking: true,
     supportsWebSearch: true,
     maxThinkingTokens: 24576,
@@ -68,7 +66,6 @@ const SUPPORTED_MODELS = {
     supportsStreaming: true,
     supportsWebSearch: true,
     supportsImages: true,
-    supportsTemperature: true,
     supportsThinking: true,
     maxThinkingTokens: 32768,
     timeout: 300000,
@@ -83,7 +80,6 @@ const SUPPORTED_MODELS = {
     maxOutputTokens: 64000,
     supportsStreaming: true,
     supportsImages: true,
-    supportsTemperature: true,
     supportsThinking: true,
     supportsWebSearch: true,
     thinkingMode: 'level',
@@ -113,7 +109,6 @@ const SUPPORTED_MODELS = {
     maxOutputTokens: 65536,
     supportsStreaming: true,
     supportsImages: true,
-    supportsTemperature: true,
     supportsThinking: true,
     supportsWebSearch: true,
     thinkingMode: 'level',
@@ -417,11 +412,9 @@ export const googleProvider = {
   async invoke(messages, options = {}) {
     const {
       model = 'gemini-2.5-flash',
-      temperature = 0.7,
       maxTokens = null,
       stream = false,
       reasoning_effort = 'medium',
-      use_websearch = false,
       media_resolution = null,
       signal,
       config,
@@ -497,14 +490,6 @@ export const googleProvider = {
     // Build generation config
     const generationConfig = {};
 
-    // Add temperature if model supports it
-    if (
-      modelConfig.supportsTemperature !== false &&
-      temperature !== undefined
-    ) {
-      generationConfig.temperature = Math.max(0, Math.min(2, temperature));
-    }
-
     // Add max tokens if specified
     if (maxTokens) {
       generationConfig.maxOutputTokens = Math.min(
@@ -579,8 +564,9 @@ export const googleProvider = {
       }
     }
 
-    // Add web search grounding if requested and model supports it
-    if (use_websearch && modelConfig.supportsWebSearch) {
+    // Attach web search grounding where the model supports it; the model
+    // decides per-request whether to actually search.
+    if (modelConfig.supportsWebSearch) {
       generationConfig.tools = [{ googleSearch: {} }];
     }
 
@@ -599,7 +585,6 @@ export const googleProvider = {
           generationConfig,
           modelConfig,
           reasoning_effort,
-          use_websearch,
           signal,
         );
       }
@@ -607,7 +592,7 @@ export const googleProvider = {
 
     try {
       debugLog(
-        `[Google] Calling ${resolvedModel} with ${messages.length} messages${use_websearch && modelConfig.supportsWebSearch ? ' (with grounding)' : ''}`,
+        `[Google] Calling ${resolvedModel} with ${messages.length} messages${modelConfig.supportsWebSearch ? ' (with grounding)' : ''}`,
       );
 
       // Check if already aborted before making request
@@ -666,7 +651,7 @@ export const googleProvider = {
             ? reasoning_effort
             : null,
           provider: 'google',
-          web_search_used: use_websearch && modelConfig.supportsWebSearch,
+          web_search_used: !!modelConfig.supportsWebSearch,
           grounding_metadata: response.groundingMetadata || null,
         },
       };
@@ -735,7 +720,6 @@ export const googleProvider = {
    * @param {Object} generationConfig - Generation configuration
    * @param {Object} modelConfig - Model configuration
    * @param {string} reasoning_effort - Reasoning effort level
-   * @param {boolean} use_websearch - Whether web search is enabled
    * @returns {AsyncGenerator} - Streaming generator yielding chunks
    */
   async *_createStreamingGenerator(
@@ -745,11 +729,10 @@ export const googleProvider = {
     generationConfig,
     modelConfig,
     reasoning_effort,
-    use_websearch,
     signal,
   ) {
     debugLog(
-      `[Google] Starting streaming for ${resolvedModel} with ${geminiContents.length} messages${use_websearch && modelConfig.supportsWebSearch ? ' (with grounding)' : ''}`,
+      `[Google] Starting streaming for ${resolvedModel} with ${geminiContents.length} messages${modelConfig.supportsWebSearch ? ' (with grounding)' : ''}`,
     );
 
     const startTime = Date.now();
@@ -771,7 +754,7 @@ export const googleProvider = {
         model: resolvedModel,
         provider: 'google',
         thinking_mode: modelConfig.supportsThinking && reasoning_effort,
-        web_search: use_websearch && modelConfig.supportsWebSearch,
+        web_search: !!modelConfig.supportsWebSearch,
       };
 
       // Create streaming request with retry logic and abort signal support
@@ -838,8 +821,8 @@ export const googleProvider = {
           total_tokens: finalResponse.usageMetadata?.totalTokenCount || 0,
         };
 
-        // Extract grounding metadata if web search was used
-        if (use_websearch && modelConfig.supportsWebSearch) {
+        // Extract grounding metadata if the model supports web search
+        if (modelConfig.supportsWebSearch) {
           groundingMetadata = finalResponse.groundingMetadata || null;
         }
 
@@ -875,7 +858,7 @@ export const googleProvider = {
             ? reasoning_effort
             : null,
           provider: 'google',
-          web_search_used: use_websearch && modelConfig.supportsWebSearch,
+          web_search_used: !!modelConfig.supportsWebSearch,
           grounding_metadata: groundingMetadata,
           thinking_mode_enabled: !!(
             modelConfig.supportsThinking && reasoning_effort