Requesty provider improvements (#1829)

* Requesty: Add cline headers to requests

* Requesty: Add support for cache read/write tokens

* Requesty: Remove deepseek reasoner conversions that are handled in-flight by Requesty

* Requesty: Read o3-mini reasoning effort from Cline settings

* Requesty: Add total cost per interaction

* Add changeset

.changeset/fair-snails-confess.md

@@ -0,0 +1,10 @@
+---
+"claude-dev": patch
+---
+
+Improve Requesty provider integration
+
+- Adding Cline headers to API requests, to enable targeted optimizations
+- Read o3 reasoning effort from Cline config, not model name
+- Show token information in task header
+- Get total cost from response when available

src/api/providers/requesty.ts

@@ -5,7 +5,6 @@ import { ApiHandlerOptions, ModelInfo, openAiModelInfoSaneDefaults } from "../..
 import { ApiHandler } from "../index"
 import { convertToOpenAiMessages } from "../transform/openai-format"
 import { ApiStream } from "../transform/stream"
-import { convertToR1Format } from "../transform/r1-format"
 
 export class RequestyHandler implements ApiHandler {
 	private options: ApiHandlerOptions
@@ -16,30 +15,32 @@ export class RequestyHandler implements ApiHandler {
 		this.client = new OpenAI({
 			baseURL: "https://router.requesty.ai/v1",
 			apiKey: this.options.requestyApiKey,
+			defaultHeaders: {
+				"HTTP-Referer": "https://cline.bot",
+				"X-Title": "Cline",
+			},
 		})
 	}
 
 	@withRetry()
 	async *createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream {
 		const modelId = this.options.requestyModelId ?? ""
-		const isDeepseekReasoner = modelId.includes("deepseek-reasoner")
 
 		let openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
 			{ role: "system", content: systemPrompt },
 			...convertToOpenAiMessages(messages),
 		]
 
-		if (isDeepseekReasoner) {
-			openAiMessages = convertToR1Format([{ role: "user", content: systemPrompt }, ...messages])
-		}
-
+		// @ts-ignore-next-line
 		const stream = await this.client.chat.completions.create({
 			model: modelId,
 			messages: openAiMessages,
 			temperature: 0,
 			stream: true,
 			stream_options: { include_usage: true },
+			...(modelId === "openai/o3-mini" ? { reasoning_effort: this.options.o3MiniReasoningEffort || "medium" } : {}),
 		})
+
 		for await (const chunk of stream) {
 			const delta = chunk.choices[0]?.delta
 			if (delta?.content) {
@@ -56,11 +57,25 @@ export class RequestyHandler implements ApiHandler {
 				}
 			}
 
+			// Requesty usage includes an extra field for Anthropic use cases.
+			// Safely cast the prompt token details section to the appropriate structure.
+			interface RequestyUsage extends OpenAI.CompletionUsage {
+				prompt_tokens_details?: {
+					caching_tokens?: number
+					cached_tokens?: number
+				}
+				total_cost?: number
+			}
+
 			if (chunk.usage) {
+				const usage = chunk.usage as RequestyUsage
 				yield {
 					type: "usage",
-					inputTokens: chunk.usage.prompt_tokens || 0,
-					outputTokens: chunk.usage.completion_tokens || 0,
+					inputTokens: usage.prompt_tokens || 0,
+					outputTokens: usage.completion_tokens || 0,
+					cacheWriteTokens: usage.prompt_tokens_details?.caching_tokens || undefined,
+					cacheReadTokens: usage.prompt_tokens_details?.cached_tokens || undefined,
+					totalCost: usage.total_cost || undefined,
 				}
 			}
 		}