feature flag

proto/cline/models.proto

@@ -98,6 +98,7 @@ message OpenRouterModelInfo {
   repeated ModelTier tiers = 12;
   optional string name = 13;
   optional double temperature = 14;
+  optional ApiFormat api_format = 15;
 }
 
 // Shared response message for model information
@@ -372,6 +373,8 @@ message OcaModelInfo {
   optional string banner = 16;
   // Canonical model identifier as reported by OCA
   string model_name = 17;
+  // The API format used by this model
+  optional ApiFormat api_format = 18;
 }
 
 // Aggregated OCA model catalog keyed by model identifier
@@ -426,6 +429,14 @@ enum ApiProvider {
   NOUSRESEARCH = 39;
 }
 
+enum ApiFormat {
+  ANTHROPIC_CHAT = 0;
+  GEMINI_CHAT = 1;
+  OPENAI_CHAT = 2;
+  R1_CHAT = 3;
+  OPENAI_RESPONSES = 4;
+}
+
 // Model info for OpenAI-compatible models
 message OpenAiCompatibleModelInfo {
   optional int64 max_tokens = 1;
@@ -442,6 +453,7 @@ message OpenAiCompatibleModelInfo {
   repeated ModelTier tiers = 12;
   optional double temperature = 13;
   optional bool is_r1_format_required = 14;
+  optional ApiFormat api_format = 15;
 }
 
 // Model info for LiteLLM models
@@ -459,6 +471,7 @@ message LiteLLMModelInfo {
   optional string description = 11;
   repeated ModelTier tiers = 12;
   optional double temperature = 13;
+  optional ApiFormat api_format = 14;
 }
 
 // Main ApiConfiguration message

src/core/api/providers/openai-native.ts

@@ -2,9 +2,11 @@ import { ModelInfo, OpenAiNativeModelId, openAiNativeDefaultModelId, openAiNativ
 import { calculateApiCostOpenAI } from "@utils/cost"
 import OpenAI from "openai"
 import type { ChatCompletionReasoningEffort, ChatCompletionTool } from "openai/resources/chat/completions"
+import { featureFlagsService } from "@/services/feature-flags"
 import { Logger } from "@/services/logging/Logger"
 import { ClineStorageMessage } from "@/shared/messages/content"
 import { fetch } from "@/shared/net"
+import { ApiFormat } from "@/shared/proto/cline/models"
 import { ApiHandler, CommonApiHandlerOptions } from "../"
 import { withRetry } from "../retry"
 import { convertToOpenAiMessages } from "../transform/openai-format"
@@ -61,13 +63,8 @@ export class OpenAiNativeHandler implements ApiHandler {
 	}
 
 	@withRetry()
-	async *createMessage(
-		systemPrompt: string,
-		messages: ClineStorageMessage[],
-		tools?: ChatCompletionTool[],
-		useResponseFormat = false,
-	): ApiStream {
-		if (useResponseFormat) {
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: ChatCompletionTool[]): ApiStream {
+		if (tools?.length && this.getModel()?.info?.apiFormat === ApiFormat.OPENAI_RESPONSES) {
 			yield* this.createResponseStream(systemPrompt, messages, tools)
 		} else {
 			yield* this.createCompletionStream(systemPrompt, messages, tools)
@@ -406,7 +403,12 @@ export class OpenAiNativeHandler implements ApiHandler {
 		const modelId = this.options.apiModelId
 		if (modelId && modelId in openAiNativeModels) {
 			const id = modelId as OpenAiNativeModelId
-			return { id, info: openAiNativeModels[id] }
+			const info: ModelInfo = { ...openAiNativeModels[id] }
+			// Ensure model is compatible with feature flags
+			if (info.apiFormat === ApiFormat.OPENAI_RESPONSES && !featureFlagsService.isResponseApiEnabled()) {
+				info.apiFormat = undefined
+			}
+			return { id, info }
 		}
 		return {
 			id: openAiNativeDefaultModelId,

src/core/api/transform/openai-response-format.ts

@@ -92,8 +92,9 @@ export function convertToOpenAIResponsesInput(messages: ClineStorageMessage[]):
 					case "thinking":
 						// Include reasoning item if it has a call_id, even if thinking is empty
 						// This is required because the API expects reasoning items to be paired with
-						// their corresponding function_calls, and will error if a function_call
-						// references a reasoning item that wasn't sent
+						// their corresponding function_calls or text, and will error if a function_call
+						// references a reasoning item that wasn't sent, or if a reasoning item is not
+						// included with a tool call: Item 'fc_' of type 'function_call' was provided without its required 'reasoning' item: 'rs_'
 						if (part.call_id && part.call_id.length > 0) {
 							assistantItems.push({
 								id: part.call_id,

src/core/prompts/system-prompt/tools/plan_mode_respond.ts

@@ -65,7 +65,7 @@ However, if while writing your response you realize you actually need to do more
 		{
 			name: "response",
 			required: true,
-			instruction: `The response to provide to the user. Do not try to use tools in this parameter, this is simply a chat response. (You MUST use the response parameter, do not simply place the response text directly within <plan_mode_respond> tags.)`,
+			instruction: `The response to provide to the user.)`,
 			usage: "Your response here",
 		},
 		{

src/core/prompts/system-prompt/variants/native-gpt-5-1/config.ts

@@ -26,8 +26,14 @@ export const config = createVariant(ModelFamily.NATIVE_GPT_5_1)
 		const providerInfo = context.providerInfo
 		const modelId = providerInfo.model.id
 
+		// Codex variants will use GPT-5 variant instead for less strict rules.
+		// Chat variants do not support native tool use.
+		if (modelId.includes("codex") && !modelId.includes("chat")) {
+			return false
+		}
+
 		// gpt-5-1-chat models do not support native tool use
-		return isGPT51Model(modelId) && !modelId.includes("chat") && isNextGenModelProvider(providerInfo)
+		return isGPT51Model(modelId) && isNextGenModelProvider(providerInfo)
 	})
 	.template(GPT_5_1_TEMPLATE_OVERRIDES.BASE)
 	.components(

src/core/prompts/system-prompt/variants/native-gpt-5/config.ts

@@ -8,7 +8,7 @@ import { GPT_5_TEMPLATE_OVERRIDES } from "./template"
 
 // Type-safe variant configuration using the builder pattern
 export const config = createVariant(ModelFamily.NATIVE_GPT_5)
-	.description("Prompt tailored to GPT-5 with native tool use support")
+	.description("Prompt tailored to GPT-5 with native tool use support with less strict rules than GPT-5.1 variant")
 	.version(1)
 	.tags("gpt", "gpt-5", "advanced", "production", "native_tools")
 	.labels({
@@ -25,10 +25,11 @@ export const config = createVariant(ModelFamily.NATIVE_GPT_5)
 		const providerInfo = context.providerInfo
 		const modelId = providerInfo.model.id
 
-		// gpt-5-chat models do not support native tool use
+		// gpt-5-chat models do not support native tool use.
 		return (
 			isGPT5ModelFamily(modelId) &&
-			!isGPT51Model(modelId) &&
+			// Exclude gpt-5.1 models except for codex variants
+			(modelId.includes("codex") || !isGPT51Model(modelId)) &&
 			!modelId.includes("chat") &&
 			isNextGenModelProvider(providerInfo)
 		)

src/core/task/StreamResponseHandler.ts

@@ -316,6 +316,8 @@ class ReasoningHandler {
 			thinking: this.pendingReasoning.content,
 			signature: this.pendingReasoning.signature,
 			summary: this.pendingReasoning.summary,
+			// Include the reasoning ID so it can be used when converting back to Responses API format
+			call_id: this.pendingReasoning.id,
 		}
 	}
 

src/core/task/index.ts

@@ -2123,15 +2123,7 @@ export class Task {
 			// saves task history item which we use to keep track of conversation history deleted range
 		}
 
-		// Response API requires native tool calls to be enabled
-		const useResponseApi = this.useNativeToolCalls && featureFlagsService.isResponseApiEnabled()
-
-		const stream = this.api.createMessage(
-			systemPrompt,
-			contextManagementMetadata.truncatedConversationHistory,
-			tools,
-			useResponseApi,
-		)
+		const stream = this.api.createMessage(systemPrompt, contextManagementMetadata.truncatedConversationHistory, tools)
 
 		const iterator = stream[Symbol.asyncIterator]()
 

src/shared/api.ts

@@ -1,4 +1,5 @@
 import type { LanguageModelChatSelector } from "../core/api/providers/types"
+import { ApiFormat } from "./proto/cline/models"
 
 export type ApiProvider =
 	| "anthropic"
@@ -257,6 +258,7 @@ export interface ModelInfo {
 		cacheReadsPrice?: number
 	}[]
 	temperature?: number
+	apiFormat?: ApiFormat
 }
 
 export interface OpenAiCompatibleModelInfo extends ModelInfo {
@@ -1400,8 +1402,45 @@ export const geminiModels = {
 // OpenAI Native
 // https://openai.com/api/pricing/
 export type OpenAiNativeModelId = keyof typeof openAiNativeModels
-export const openAiNativeDefaultModelId: OpenAiNativeModelId = "gpt-5-2025-08-07"
+export const openAiNativeDefaultModelId: OpenAiNativeModelId = "gpt-5.1-2025-11-13"
 export const openAiNativeModels = {
+	"gpt-5.1-2025-11-13": {
+		maxTokens: 8_192,
+		contextWindow: 272000,
+		supportsImages: true,
+		supportsPromptCache: true,
+		inputPrice: 1.25,
+		outputPrice: 10.0,
+		cacheReadsPrice: 0.125,
+	},
+	"gpt-5.1": {
+		maxTokens: 8_192, // 128000 breaks context window truncation
+		contextWindow: 272000,
+		supportsImages: true,
+		supportsPromptCache: true,
+		inputPrice: 1.25,
+		outputPrice: 10.0,
+		cacheReadsPrice: 0.125,
+	},
+	"gpt-5.1-codex": {
+		maxTokens: 8_192, // 128000 breaks context window truncation
+		contextWindow: 400000,
+		supportsImages: true,
+		supportsPromptCache: true,
+		inputPrice: 1.25,
+		outputPrice: 10.0,
+		cacheReadsPrice: 0.125,
+		apiFormat: ApiFormat.OPENAI_RESPONSES,
+	},
+	"gpt-5.1-chat-latest": {
+		maxTokens: 8_192,
+		contextWindow: 400000,
+		supportsImages: true,
+		supportsPromptCache: true,
+		inputPrice: 1.25,
+		outputPrice: 10,
+		cacheReadsPrice: 0.125,
+	},
 	"gpt-5-2025-08-07": {
 		maxTokens: 8_192, // 128000 breaks context window truncation
 		contextWindow: 272000,
@@ -1411,6 +1450,16 @@ export const openAiNativeModels = {
 		outputPrice: 10.0,
 		cacheReadsPrice: 0.125,
 	},
+	"gpt-5-codex": {
+		maxTokens: 8_192, // 128000 breaks context window truncation
+		contextWindow: 400000,
+		supportsImages: true,
+		supportsPromptCache: true,
+		inputPrice: 1.25,
+		outputPrice: 10.0,
+		cacheReadsPrice: 0.125,
+		apiFormat: ApiFormat.OPENAI_RESPONSES,
+	},
 	"gpt-5-mini-2025-08-07": {
 		maxTokens: 8_192,
 		contextWindow: 272000,
@@ -1438,33 +1487,6 @@ export const openAiNativeModels = {
 		outputPrice: 10,
 		cacheReadsPrice: 0.125,
 	},
-	"gpt-5.1-2025-11-13": {
-		maxTokens: 8_192,
-		contextWindow: 272000,
-		supportsImages: true,
-		supportsPromptCache: true,
-		inputPrice: 1.25,
-		outputPrice: 10.0,
-		cacheReadsPrice: 0.125,
-	},
-	"gpt-5.1": {
-		maxTokens: 8_192,
-		contextWindow: 272000,
-		supportsImages: true,
-		supportsPromptCache: true,
-		inputPrice: 1.25,
-		outputPrice: 10.0,
-		cacheReadsPrice: 0.125,
-	},
-	"gpt-5.1-chat-latest": {
-		maxTokens: 8_192,
-		contextWindow: 400000,
-		supportsImages: true,
-		supportsPromptCache: true,
-		inputPrice: 1.25,
-		outputPrice: 10,
-		cacheReadsPrice: 0.125,
-	},
 	o3: {
 		maxTokens: 100_000,
 		contextWindow: 200_000,