Method createMessage

src/api/providers/opencode-go.ts:160–237

* Streams a chat completion response, yielding typed chunks for text, * reasoning, partial tool calls, and token usage. * * Anthropic-format models (Qwen/MiniMax) are streamed via * streamAnthropicMessage against `/v1/messages`; all other models * use the OpenAI-compatible chat com

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

	 * model advertises `supportsReasoningEffort`.
	 */
	override async *createMessage(
		systemPrompt: string,
		messages: Anthropic.Messages.MessageParam[],
		metadata?: ApiHandlerCreateMessageMetadata,
	): ApiStream {
		const { id: modelId, info, format, temperature, reasoningEffort, maxTokens } = await this.resolveModel()

		if (format === "anthropic") {
			yield* this.streamAnthropicMessage(modelId, info, temperature, maxTokens, systemPrompt, messages, metadata)
			return
		}

		// preserveReasoning models (GLM/DeepSeek/MiMo/MiniMax/Qwen) require
		// reasoning_content to be carried across tool-call continuations.
		const preserveReasoning = info.preserveReasoning === true
		const convertedMessages = preserveReasoning
			? convertToR1Format(messages, { mergeToolResultText: true })
			: convertToOpenAiMessages(messages)

		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
			{ role: "system", content: systemPrompt },
			...convertedMessages,
		]

		const body: OpenAI.Chat.ChatCompletionCreateParams = {
			model: modelId,
			messages: openAiMessages,
			temperature: this.supportsTemperature(modelId) ? temperature : undefined,
			max_completion_tokens:
				this.options.includeMaxTokens === true ? this.options.modelMaxTokens || maxTokens : maxTokens,
			stream: true,
			stream_options: { include_usage: true },
			tools: this.convertToolsForOpenAI(metadata?.tools),
			tool_choice: metadata?.tool_choice,
			parallel_tool_calls: metadata?.parallelToolCalls ?? true,
			...(reasoningEffort && {
				reasoning_effort: reasoningEffort as OpenAI.Chat.ChatCompletionCreateParams["reasoning_effort"],
			}),
		}

		const completion = await this.client.chat.completions.create(body)

		for await (const chunk of completion) {
			const delta = chunk.choices[0]?.delta

			if (delta?.content) {
				yield { type: "text", text: delta.content }
			}

			// Several Go-plan models (GLM, DeepSeek) stream reasoning via this field.
			const reasoningText = extractReasoningFromDelta(delta)
			if (reasoningText) {
				yield { type: "reasoning", text: reasoningText }
			}

			// Emit raw tool call chunks - NativeToolCallParser handles state management.
			if (delta?.tool_calls) {
				for (const toolCall of delta.tool_calls) {