feat: add prompt caching support for LiteLLM

- Add litellmUsePromptCache field to provider settings schema
- Update LiteLLM handler to add cache_control metadata when enabled
- Add UI checkbox for enabling/disabling prompt caching
- Support cache-related token usage tracking (cache reads/writes)
- Add translation keys for new UI elements

Fixes #5791

packages/types/src/provider-settings.ts

@@ -218,6 +218,7 @@ const litellmSchema = baseProviderSettingsSchema.extend({
 	litellmBaseUrl: z.string().optional(),
 	litellmApiKey: z.string().optional(),
 	litellmModelId: z.string().optional(),
+	litellmUsePromptCache: z.boolean().optional(),
 })
 
 const defaultSchema = z.object({

src/api/providers/lite-llm.ts

@@ -39,11 +39,16 @@ export class LiteLLMHandler extends RouterProvider implements SingleCompletionHa
 	): ApiStream {
 		const { id: modelId, info } = await this.fetchModel()
 
-		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+		const baseOpenAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
 			{ role: "system", content: systemPrompt },
 			...convertToOpenAiMessages(messages),
 		]
 
+		// Apply prompt caching if enabled
+		const openAiMessages = this.options.litellmUsePromptCache
+			? this.addCacheControlToMessages(baseOpenAiMessages)
+			: baseOpenAiMessages
+
 		// Required by some providers; others default to max tokens allowed
 		let maxTokens: number | undefined = info.maxTokens ?? undefined
 
@@ -80,12 +85,21 @@ export class LiteLLMHandler extends RouterProvider implements SingleCompletionHa
 			}
 
 			if (lastUsage) {
+				// Extract cache-related information if available
+				const cacheWriteTokens =
+					lastUsage.cache_creation_input_tokens || lastUsage.prompt_cache_miss_tokens || 0
+				const cacheReadTokens =
+					lastUsage.cache_read_input_tokens ||
+					lastUsage.prompt_cache_hit_tokens ||
+					lastUsage.prompt_tokens_details?.cached_tokens ||
+					0
+
 				const usageData: ApiStreamUsageChunk = {
 					type: "usage",
 					inputTokens: lastUsage.prompt_tokens || 0,
 					outputTokens: lastUsage.completion_tokens || 0,
-					cacheWriteTokens: lastUsage.cache_creation_input_tokens || 0,
-					cacheReadTokens: lastUsage.prompt_tokens_details?.cached_tokens || 0,
+					cacheWriteTokens: cacheWriteTokens > 0 ? cacheWriteTokens : undefined,
+					cacheReadTokens: cacheReadTokens > 0 ? cacheReadTokens : undefined,
 				}
 
 				usageData.totalCost = calculateApiCostOpenAI(
@@ -130,9 +144,45 @@ export class LiteLLMHandler extends RouterProvider implements SingleCompletionHa
 			throw error
 		}
 	}
+
+	/**
+	 * Add cache control metadata to messages for prompt caching
+	 * Based on Cline's implementation: adds cache_control to system message and last two user messages
+	 */
+	private addCacheControlToMessages(
+		messages: OpenAI.Chat.ChatCompletionMessageParam[],
+	): OpenAI.Chat.ChatCompletionMessageParam[] {
+		const cacheControl = { cache_control: { type: "ephemeral" } }
+
+		// Find user message indices
+		const userMsgIndices = messages.reduce(
+			(acc, msg, index) => (msg.role === "user" ? [...acc, index] : acc),
+			[] as number[],
+		)
+
+		const lastUserMsgIndex = userMsgIndices[userMsgIndices.length - 1] ?? -1
+		const secondLastUserMsgIndex = userMsgIndices[userMsgIndices.length - 2] ?? -1
+
+		return messages.map((message, index) => {
+			// Add cache control to system message (first message)
+			if (index === 0 && message.role === "system") {
+				return { ...message, ...cacheControl }
+			}
+
+			// Add cache control to last two user messages
+			if (index === lastUserMsgIndex || index === secondLastUserMsgIndex) {
+				return { ...message, ...cacheControl }
+			}
+
+			return message
+		})
+	}
 }
 
 // LiteLLM usage may include an extra field for Anthropic use cases.
 interface LiteLLMUsage extends OpenAI.CompletionUsage {
 	cache_creation_input_tokens?: number
+	prompt_cache_miss_tokens?: number
+	cache_read_input_tokens?: number
+	prompt_cache_hit_tokens?: number
 }

webview-ui/src/components/settings/providers/LiteLLM.tsx

@@ -1,5 +1,5 @@
 import { useCallback, useState, useEffect, useRef } from "react"
-import { VSCodeTextField } from "@vscode/webview-ui-toolkit/react"
+import { VSCodeTextField, VSCodeCheckbox } from "@vscode/webview-ui-toolkit/react"
 
 import { type ProviderSettings, type OrganizationAllowList, litellmDefaultModelId } from "@roo-code/types"
 
@@ -111,6 +111,20 @@ export const LiteLLM = ({
 				{t("settings:providers.apiKeyStorageNotice")}
 			</div>
 
+			<div className="flex flex-col gap-2 mt-4">
+				<VSCodeCheckbox
+					checked={apiConfiguration?.litellmUsePromptCache || false}
+					onChange={(e: any) => {
+						const isChecked = e.target.checked === true
+						setApiConfigurationField("litellmUsePromptCache", isChecked)
+					}}>
+					{t("settings:providers.litellmUsePromptCache")}
+				</VSCodeCheckbox>
+				<div className="text-sm text-vscode-descriptionForeground ml-6">
+					{t("settings:providers.litellmUsePromptCacheDescription")}
+				</div>
+			</div>
+
 			<Button
 				variant="outline"
 				onClick={handleRefreshModels}

webview-ui/src/i18n/locales/en/settings.json

@@ -262,6 +262,8 @@
 		"getXaiApiKey": "Get xAI API Key",
 		"litellmApiKey": "LiteLLM API Key",
 		"litellmBaseUrl": "LiteLLM Base URL",
+		"litellmUsePromptCache": "Use prompt caching",
+		"litellmUsePromptCacheDescription": "Enable prompt caching to improve performance and reduce costs for supported models. Requires a model that supports prompt caching (e.g., Claude 3.7 Sonnet).",
 		"awsCredentials": "AWS Credentials",
 		"awsProfile": "AWS Profile",
 		"awsProfileName": "AWS Profile Name",