Class LmStudioHandler

src/api/providers/lm-studio.ts:20–218

import { handleOpenAIError } from "./utils/openai-error-handler"

export class LmStudioHandler extends BaseProvider implements SingleCompletionHandler {
	protected options: ApiHandlerOptions
	private client: OpenAI
	private readonly providerName = "LM Studio"

	constructor(options: ApiHandlerOptions) {
		super()
		this.options = options

		// LM Studio uses "noop" as a placeholder API key
		const apiKey = "noop"

		this.client = new OpenAI({
			baseURL: (this.options.lmStudioBaseUrl || "http://localhost:1234") + "/v1",
			apiKey: apiKey,
			timeout: this.timeoutMs,
		})
	}

	override async *createMessage(
		systemPrompt: string,
		messages: Anthropic.Messages.MessageParam[],
		metadata?: ApiHandlerCreateMessageMetadata,
	): ApiStream {
		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
			{ role: "system", content: systemPrompt },
			...convertToOpenAiMessages(messages),
		]

		// -------------------------
		// Track token usage
		// -------------------------
		const toContentBlocks = (
			blocks: Anthropic.Messages.MessageParam[] | string,
		): Anthropic.Messages.ContentBlockParam[] => {
			if (typeof blocks === "string") {
				return [{ type: "text", text: blocks }]
			}

			const result: Anthropic.Messages.ContentBlockParam[] = []
			for (const msg of blocks) {
				if (typeof msg.content === "string") {
					result.push({ type: "text", text: msg.content })
				} else if (Array.isArray(msg.content)) {
					for (const part of msg.content) {
						if (part.type === "text") {
							result.push({ type: "text", text: part.text })
						}
					}
				}
			}
			return result
		}

		let inputTokens = 0
		try {
			inputTokens = await this.countTokens([{ type: "text", text: systemPrompt }, ...toContentBlocks(messages)])
		} catch (err) {