Method createMessage

src/api/providers/lite-llm.ts:113–312

(
		systemPrompt: string,
		messages: Anthropic.Messages.MessageParam[],
		metadata?: ApiHandlerCreateMessageMetadata,
	)

	}

	override async *createMessage(
		systemPrompt: string,
		messages: Anthropic.Messages.MessageParam[],
		metadata?: ApiHandlerCreateMessageMetadata,
	): ApiStream {
		const { id: modelId, info } = await this.fetchModel()

		const openAiMessages = convertToOpenAiMessages(messages, {
			normalizeToolCallId: sanitizeOpenAiCallId,
		})

		// Prepare messages with cache control if enabled and supported
		let systemMessage: OpenAI.Chat.ChatCompletionMessageParam
		let enhancedMessages: OpenAI.Chat.ChatCompletionMessageParam[]

		if (this.options.litellmUsePromptCache && info.supportsPromptCache) {
			// Create system message with cache control in the proper format
			systemMessage = {
				role: "system",
				content: [
					{
						type: "text",
						text: systemPrompt,
						cache_control: { type: "ephemeral" },
					} as any,
				],
			}

			// Find the last two user messages to apply caching
			const userMsgIndices = openAiMessages.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

			// Apply cache_control to the last two user messages
			enhancedMessages = openAiMessages.map((message, index) => {
				if ((index === lastUserMsgIndex || index === secondLastUserMsgIndex) && message.role === "user") {
					// Handle both string and array content types
					if (typeof message.content === "string") {
						return {
							...message,
							content: [
								{
									type: "text",
									text: message.content,
									cache_control: { type: "ephemeral" },
								} as any,
							],
						}
					} else if (Array.isArray(message.content)) {
						// Apply cache control to the last content item in the array
						return {
							...message,
							content: message.content.map((content, contentIndex) =>
								contentIndex === message.content.length - 1
									? ({