Enable Gemini prompt caching by default (RooCodeInc#3225)

Author: cte

.changeset/three-rings-approve.md

@@ -0,0 +1,5 @@
+---
+"roo-cline": patch
+---
+
+Enable Gemini prompt caching by default

evals/packages/types/src/roo-code.ts

@@ -418,7 +418,7 @@ export const providerSettingsSchema = z.object({
 	// Generic
 	includeMaxTokens: z.boolean().optional(),
 	reasoningEffort: reasoningEffortsSchema.optional(),
-	promptCachingEnabled: z.boolean().optional(),
+	promptCachingDisabled: z.boolean().optional(),
 	diffEnabled: z.boolean().optional(),
 	fuzzyMatchThreshold: z.number().optional(),
 	modelTemperature: z.number().nullish(),
@@ -507,7 +507,7 @@ const providerSettingsRecord: ProviderSettingsRecord = {
 	// Generic
 	includeMaxTokens: undefined,
 	reasoningEffort: undefined,
-	promptCachingEnabled: undefined,
+	promptCachingDisabled: undefined,
 	diffEnabled: undefined,
 	fuzzyMatchThreshold: undefined,
 	modelTemperature: undefined,

src/api/providers/__tests__/gemini.test.ts

@@ -291,7 +291,7 @@ describe("Caching Logic", () => {
 			apiKey: "test-key",
 			apiModelId: "gemini-1.5-flash-latest", // Use a model that supports caching
 			geminiApiKey: "test-key",
-			promptCachingEnabled: true, // Enable caching for these tests
+			promptCachingDisabled: false,
 		})
 
 		handlerWithCache["client"] = {
@@ -309,8 +309,8 @@ describe("Caching Logic", () => {
 		} as any
 	})
 
-	it("should not use cache if promptCachingEnabled is false", async () => {
-		handlerWithCache["options"].promptCachingEnabled = false
+	it("should not use cache if promptCachingDisabled is true", async () => {
+		handlerWithCache["options"].promptCachingDisabled = true
 		const stream = handlerWithCache.createMessage(systemPrompt, mockMessagesLong, cacheKey)
 
 		for await (const _ of stream) {

src/api/providers/gemini.ts

@@ -93,7 +93,7 @@ export class GeminiHandler extends BaseProvider implements SingleCompletionHandl
 		// https://ai.google.dev/gemini-api/docs/tokens?lang=node
 		const isCacheAvailable =
 			info.supportsPromptCache &&
-			this.options.promptCachingEnabled &&
+			!this.options.promptCachingDisabled &&
 			cacheKey &&
 			contentsLength > 4 * CONTEXT_CACHE_TOKEN_MINIMUM
 

src/api/providers/openrouter.ts

@@ -94,7 +94,7 @@ export class OpenRouterHandler extends BaseProvider implements SingleCompletionH
 			openAiMessages = convertToR1Format([{ role: "user", content: systemPrompt }, ...messages])
 		}
 
-		const isCacheAvailable = promptCache.supported && (!promptCache.optional || this.options.promptCachingEnabled)
+		const isCacheAvailable = promptCache.supported && (!promptCache.optional || !this.options.promptCachingDisabled)
 
 		// https://openrouter.ai/docs/features/prompt-caching
 		if (isCacheAvailable) {

src/exports/roo-code.d.ts

@@ -127,7 +127,7 @@ type ProviderSettings = {
 	modelMaxThinkingTokens?: number | undefined
 	includeMaxTokens?: boolean | undefined
 	reasoningEffort?: ("low" | "medium" | "high") | undefined
-	promptCachingEnabled?: boolean | undefined
+	promptCachingDisabled?: boolean | undefined
 	diffEnabled?: boolean | undefined
 	fuzzyMatchThreshold?: number | undefined
 	modelTemperature?: (number | null) | undefined

src/exports/types.ts

@@ -128,7 +128,7 @@ type ProviderSettings = {
 	modelMaxThinkingTokens?: number | undefined
 	includeMaxTokens?: boolean | undefined
 	reasoningEffort?: ("low" | "medium" | "high") | undefined
-	promptCachingEnabled?: boolean | undefined
+	promptCachingDisabled?: boolean | undefined
 	diffEnabled?: boolean | undefined
 	fuzzyMatchThreshold?: number | undefined
 	modelTemperature?: (number | null) | undefined

src/schemas/index.ts

@@ -434,7 +434,7 @@ export const providerSettingsSchema = z.object({
 	// Generic
 	includeMaxTokens: z.boolean().optional(),
 	reasoningEffort: reasoningEffortsSchema.optional(),
-	promptCachingEnabled: z.boolean().optional(),
+	promptCachingDisabled: z.boolean().optional(),
 	diffEnabled: z.boolean().optional(),
 	fuzzyMatchThreshold: z.number().optional(),
 	modelTemperature: z.number().nullish(),
@@ -524,7 +524,7 @@ const providerSettingsRecord: ProviderSettingsRecord = {
 	// Generic
 	includeMaxTokens: undefined,
 	reasoningEffort: undefined,
-	promptCachingEnabled: undefined,
+	promptCachingDisabled: undefined,
 	diffEnabled: undefined,
 	fuzzyMatchThreshold: undefined,
 	modelTemperature: undefined,

webview-ui/src/components/settings/PromptCachingControl.tsx

@@ -16,8 +16,8 @@ export const PromptCachingControl = ({ apiConfiguration, setApiConfigurationFiel
 		<>
 			<div>
 				<VSCodeCheckbox
-					checked={apiConfiguration.promptCachingEnabled}
-					onChange={(e: any) => setApiConfigurationField("promptCachingEnabled", e.target.checked)}>
+					checked={apiConfiguration.promptCachingDisabled}
+					onChange={(e: any) => setApiConfigurationField("promptCachingDisabled", e.target.checked)}>
 					<label className="block font-medium mb-1">{t("settings:promptCaching.label")}</label>
 				</VSCodeCheckbox>
 				<div className="text-sm text-vscode-descriptionForeground mt-1">

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

@@ -416,8 +416,8 @@
 		}
 	},
 	"promptCaching": {
-		"label": "Habilitar emmagatzematge en caché de prompts",
-		"description": "Quan està habilitat, Roo utilitzarà aquest model amb la memòria cau de prompts activada per reduir costos."
+		"label": "Desactivar la memòria cau de prompts",
+		"description": "Quan està marcat, Roo no utilitzarà la memòria cau de prompts per a aquest model."
 	},
 	"temperature": {
 		"useCustom": "Utilitzar temperatura personalitzada",