fix: remove unnecessary vercel stream transform

src/core/api/providers/vercel-ai-gateway.ts

@@ -5,14 +5,16 @@ import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/com
 import { fetch } from "@/shared/net"
 import { ApiHandler, CommonApiHandlerOptions } from "../index"
 import { withRetry } from "../retry"
+import { createOpenRouterStream } from "../transform/openrouter-stream"
 import { ApiStream } from "../transform/stream"
 import { ToolCallProcessor } from "../transform/tool-call-processor"
-import { createVercelAIGatewayStream } from "../transform/vercel-ai-gateway-stream"
 
 interface VercelAIGatewayHandlerOptions extends CommonApiHandlerOptions {
 	vercelAiGatewayApiKey?: string
 	openRouterModelId?: string
 	openRouterModelInfo?: ModelInfo
+	openRouterProviderSorting?: string
+	reasoningEffort?: string
 	thinkingBudgetTokens?: number
 }
 
@@ -53,12 +55,14 @@ export class VercelAIGatewayHandler implements ApiHandler {
 		const modelInfo = this.getModel().info
 
 		try {
-			const stream = await createVercelAIGatewayStream(
+			const stream = await createOpenRouterStream(
 				client,
 				systemPrompt,
 				messages,
 				{ id: modelId, info: modelInfo },
+				this.options.reasoningEffort,
 				this.options.thinkingBudgetTokens,
+				this.options.openRouterProviderSorting,
 				tools,
 			)
 			let didOutputUsage: boolean = false

src/core/api/transform/openrouter-stream.ts

@@ -34,95 +34,48 @@ export async function createOpenRouterStream(
 		model.id = model.id.slice(0, -CLAUDE_SONNET_1M_SUFFIX.length)
 	}
 
-	// prompt caching: https://openrouter.ai/docs/prompt-caching
-	// this was initially specifically for claude models (some models may 'support prompt caching' automatically without this)
-	// handles direct model.id match logic
-	switch (model.id) {
-		case "anthropic/claude-haiku-4.5":
-		case "anthropic/claude-4.5-haiku":
-		case "anthropic/claude-sonnet-4.5":
-		case "anthropic/claude-4.5-sonnet": // OpenRouter accidentally included this in model list for a brief moment, and users may be using this model id. And to support prompt caching, we need to add it here.
-		case "anthropic/claude-sonnet-4":
-		case "anthropic/claude-opus-4.1":
-		case "anthropic/claude-opus-4":
-		case "anthropic/claude-3.7-sonnet":
-		case "anthropic/claude-3.7-sonnet:beta":
-		case "anthropic/claude-3.7-sonnet:thinking":
-		case "anthropic/claude-3-7-sonnet":
-		case "anthropic/claude-3-7-sonnet:beta":
-		case "anthropic/claude-3.5-sonnet":
-		case "anthropic/claude-3.5-sonnet:beta":
-		case "anthropic/claude-3.5-sonnet-20240620":
-		case "anthropic/claude-3.5-sonnet-20240620:beta":
-		case "anthropic/claude-3-5-haiku":
-		case "anthropic/claude-3-5-haiku:beta":
-		case "anthropic/claude-3-5-haiku-20241022":
-		case "anthropic/claude-3-5-haiku-20241022:beta":
-		case "anthropic/claude-3-haiku":
-		case "anthropic/claude-3-haiku:beta":
-		case "anthropic/claude-3-opus":
-		case "anthropic/claude-3-opus:beta":
-			openAiMessages[0] = {
-				role: "system",
-				content: [
-					{
-						type: "text",
-						text: systemPrompt,
-						// @ts-ignore-next-line
-						cache_control: { type: "ephemeral" },
-					},
-				],
-			}
-			// Add cache_control to the last two user messages
-			// (note: this works because we only ever add one user message at a time, but if we added multiple we'd need to mark the user message before the last assistant message)
-			const lastTwoUserMessages = openAiMessages.filter((msg) => msg.role === "user").slice(-2)
-			lastTwoUserMessages.forEach((msg) => {
-				if (typeof msg.content === "string") {
-					msg.content = [{ type: "text", text: msg.content }]
-				}
-				if (Array.isArray(msg.content)) {
-					// NOTE: this is fine since env details will always be added at the end. but if it weren't there, and the user added a image_url type message, it would pop a text part before it and then move it after to the end.
-					let lastTextPart = msg.content.filter((part) => part.type === "text").pop()
+	const isAnthropicModel = model.id.startsWith("anthropic/")
 
-					if (!lastTextPart) {
-						lastTextPart = { type: "text", text: "..." }
-						msg.content.push(lastTextPart)
-					}
+	// prompt caching: https://openrouter.ai/docs/prompt-caching
+
+	if (isAnthropicModel) {
+		openAiMessages[0] = {
+			role: "system",
+			content: [
+				{
+					type: "text",
+					text: systemPrompt,
 					// @ts-ignore-next-line
-					lastTextPart["cache_control"] = { type: "ephemeral" }
+					cache_control: { type: "ephemeral" },
+				},
+			],
+		}
+		// Add cache_control to the last two user messages
+		// (note: this works because we only ever add one user message at a time, but if we added multiple we'd need to mark the user message before the last assistant message)
+		const lastTwoUserMessages = openAiMessages.filter((msg) => msg.role === "user").slice(-2)
+		lastTwoUserMessages.forEach((msg) => {
+			if (typeof msg.content === "string") {
+				msg.content = [{ type: "text", text: msg.content }]
+			}
+			if (Array.isArray(msg.content)) {
+				// NOTE: this is fine since env details will always be added at the end. but if it weren't there, and the user added a image_url type message, it would pop a text part before it and then move it after to the end.
+				let lastTextPart = msg.content.filter((part) => part.type === "text").pop()
+
+				if (!lastTextPart) {
+					lastTextPart = { type: "text", text: "..." }
+					msg.content.push(lastTextPart)
 				}
-			})
-			break
-		default:
-			break
+				// @ts-ignore-next-line
+				lastTextPart["cache_control"] = { type: "ephemeral" }
+			}
+		})
 	}
 
-	// Not sure how openrouter defaults max tokens when no value is provided, but the anthropic api requires this value and since they offer both 4096 and 8192 variants, we should ensure 8192.
+	// The anthropic api requires this value and since they offer both 4096 and 8192 variants, we should ensure 8192.
 	// (models usually default to max tokens allowed)
 	let maxTokens: number | undefined
-	switch (model.id) {
-		case "anthropic/claude-haiku-4.5":
-		case "anthropic/claude-4.5-haiku":
-		case "anthropic/claude-sonnet-4.5":
-		case "anthropic/claude-4.5-sonnet":
-		case "anthropic/claude-sonnet-4":
-		case "anthropic/claude-opus-4.1":
-		case "anthropic/claude-opus-4":
-		case "anthropic/claude-3.7-sonnet":
-		case "anthropic/claude-3.7-sonnet:beta":
-		case "anthropic/claude-3.7-sonnet:thinking":
-		case "anthropic/claude-3-7-sonnet":
-		case "anthropic/claude-3-7-sonnet:beta":
-		case "anthropic/claude-3.5-sonnet":
-		case "anthropic/claude-3.5-sonnet:beta":
-		case "anthropic/claude-3.5-sonnet-20240620":
-		case "anthropic/claude-3.5-sonnet-20240620:beta":
-		case "anthropic/claude-3-5-haiku":
-		case "anthropic/claude-3-5-haiku:beta":
-		case "anthropic/claude-3-5-haiku-20241022":
-		case "anthropic/claude-3-5-haiku-20241022:beta":
-			maxTokens = 8_192
-			break
+	if (isAnthropicModel) {
+		maxTokens = 8_192
 	}
 
 	let temperature: number | undefined = 0
@@ -140,32 +93,18 @@ export async function createOpenRouterStream(
 	}
 
 	let reasoning: { max_tokens: number } | undefined
-	switch (model.id) {
-		case "anthropic/claude-haiku-4.5":
-		case "anthropic/claude-4.5-haiku":
-		case "anthropic/claude-sonnet-4.5":
-		case "anthropic/claude-4.5-sonnet":
-		case "anthropic/claude-sonnet-4":
-		case "anthropic/claude-opus-4.1":
-		case "anthropic/claude-opus-4":
-		case "anthropic/claude-3.7-sonnet":
-		case "anthropic/claude-3.7-sonnet:beta":
-		case "anthropic/claude-3.7-sonnet:thinking":
-		case "anthropic/claude-3-7-sonnet":
-		case "anthropic/claude-3-7-sonnet:beta":
-			const budget_tokens = thinkingBudgetTokens || 0
-			const reasoningOn = budget_tokens !== 0
-			if (reasoningOn) {
-				temperature = undefined // extended thinking does not support non-1 temperature
-				reasoning = { max_tokens: budget_tokens }
-			}
-			break
-		default:
-			if (thinkingBudgetTokens && model.info?.thinkingConfig && thinkingBudgetTokens > 0) {
-				temperature = undefined // extended thinking does not support non-1 temperature
-				reasoning = { max_tokens: thinkingBudgetTokens }
-				break
-			}
+	if (isAnthropicModel) {
+		const budget_tokens = thinkingBudgetTokens || 0
+		const reasoningOn = budget_tokens !== 0
+		if (reasoningOn) {
+			temperature = undefined // extended thinking does not support non-1 temperature
+			reasoning = { max_tokens: budget_tokens }
+		}
+	} else {
+		if (thinkingBudgetTokens && model.info?.thinkingConfig && thinkingBudgetTokens > 0) {
+			temperature = undefined // extended thinking does not support non-1 temperature
+			reasoning = { max_tokens: thinkingBudgetTokens }
+		}
 	}
 
 	const providerPreferences = OPENROUTER_PROVIDER_PREFERENCES[model.id]
@@ -187,7 +126,6 @@ export async function createOpenRouterStream(
 		...(reasoning ? { reasoning } : {}),
 		...(openRouterProviderSorting && !providerPreferences ? { provider: { sort: openRouterProviderSorting } } : {}),
 		...(providerPreferences ? { provider: providerPreferences } : {}),
-		...(isClaudeSonnet1m ? { provider: { order: ["anthropic", "google-vertex/global"], allow_fallbacks: false } } : {}),
 		...getOpenAIToolParams(tools),
 	})
 

src/core/api/transform/vercel-ai-gateway-stream.ts

@@ -1,91 +0,0 @@
-import { Anthropic } from "@anthropic-ai/sdk"
-import {
-	CLAUDE_SONNET_1M_SUFFIX,
-	ModelInfo,
-	openRouterClaudeSonnet41mModelId,
-	openRouterClaudeSonnet451mModelId,
-} from "@shared/api"
-import OpenAI from "openai"
-import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
-import { convertToOpenAiMessages } from "../transform/openai-format"
-import { getOpenAIToolParams } from "./tool-call-processor"
-
-export async function createVercelAIGatewayStream(
-	client: OpenAI,
-	systemPrompt: string,
-	messages: Anthropic.Messages.MessageParam[],
-	model: { id: string; info: ModelInfo },
-	thinkingBudgetTokens?: number,
-	tools?: OpenAITool[],
-) {
-	// Convert Anthropic messages to OpenAI format
-	const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
-		{ role: "system", content: systemPrompt },
-		...convertToOpenAiMessages(messages),
-	]
-
-	const isClaudeSonnet1m = model.id === openRouterClaudeSonnet41mModelId || model.id === openRouterClaudeSonnet451mModelId
-	if (isClaudeSonnet1m) {
-		// remove the custom :1m suffix, to create the model id openrouter API expects
-		model.id = model.id.slice(0, -CLAUDE_SONNET_1M_SUFFIX.length)
-	}
-
-	const isAnthropicModel = model.id.startsWith("anthropic/")
-
-	if (isAnthropicModel) {
-		openAiMessages[0] = {
-			role: "system",
-			content: systemPrompt,
-			// @ts-ignore-next-line
-			cache_control: { type: "ephemeral" },
-		}
-
-		// Add cache_control to the last two user messages for conversation context caching
-		const lastTwoUserMessages = openAiMessages.filter((msg) => msg.role === "user").slice(-2)
-		lastTwoUserMessages.forEach((msg) => {
-			if (typeof msg.content === "string" && msg.content.length > 0) {
-				msg.content = [{ type: "text", text: msg.content }]
-			}
-			if (Array.isArray(msg.content)) {
-				// Find the last text part in the message content
-				const lastTextPart = msg.content.filter((part) => part.type === "text").pop()
-
-				if (lastTextPart && lastTextPart.text && lastTextPart.text.length > 0) {
-					// @ts-ignore-next-line
-					lastTextPart["cache_control"] = { type: "ephemeral" }
-				}
-			}
-		})
-	}
-
-	// Configure reasoning parameters similar to OpenRouter
-	let temperature: number | undefined = 0
-	let reasoning: { max_tokens: number } | undefined
-
-	if (isAnthropicModel) {
-		const budget_tokens = thinkingBudgetTokens || 0
-		const reasoningOn = budget_tokens !== 0
-		if (reasoningOn) {
-			temperature = undefined // extended thinking does not support non-1 temperature
-			reasoning = { max_tokens: budget_tokens }
-		}
-	} else if (thinkingBudgetTokens && model.info?.thinkingConfig && thinkingBudgetTokens > 0) {
-		temperature = undefined // extended thinking does not support non-1 temperature
-		reasoning = { max_tokens: thinkingBudgetTokens }
-	}
-
-	// @ts-ignore-next-line
-	const stream = await client.chat.completions.create({
-		model: model.id,
-		max_tokens: model.info.maxTokens,
-		temperature: temperature,
-		messages: openAiMessages,
-		stream: true,
-		stream_options: { include_usage: true },
-		include_reasoning: true,
-		...(reasoning ? { reasoning } : {}),
-		...getOpenAIToolParams(tools),
-	})
-
-	return stream
-}