Method attemptApiRequest

src/core/task/Task.ts:3972–4383

(
		retryAttempt: number = 0,
		options: { skipProviderRateLimit?: boolean } = {},
	)

	}

	public async *attemptApiRequest(
		retryAttempt: number = 0,
		options: { skipProviderRateLimit?: boolean } = {},
	): ApiStream {
		const state = await this.providerRef.deref()?.getState()

		const {
			apiConfiguration,
			autoApprovalEnabled,
			requestDelaySeconds,
			mode,
			autoCondenseContext = true,
			autoCondenseContextPercent = 100,
			profileThresholds = {},
		} = state ?? {}

		// Get condensing configuration for automatic triggers.
		const customCondensingPrompt = state?.customSupportPrompts?.CONDENSE

		if (!options.skipProviderRateLimit) {
			await this.maybeWaitForProviderRateLimit(retryAttempt)
		}

		// Update last request time right before making the request so that subsequent
		// requests — even from new subtasks — will honour the provider's rate-limit.
		//
		// NOTE: When recursivelyMakeClineRequests handles rate limiting, it sets the
		// timestamp earlier to include the environment details build. We still set it
		// here for direct callers (tests) and for the case where we didn't rate-limit
		// in the caller.
		this.rateLimitClock.recordRequest()

		const systemPrompt = await this.getSystemPrompt()
		const { contextTokens } = this.getTokenUsage()

		if (contextTokens) {
			const modelInfo = this.api.getModel().info

			const maxTokens = getModelMaxOutputTokens({
				modelId: this.api.getModel().id,
				model: modelInfo,
				settings: this.apiConfiguration,
			})

			// vscode-lm condenses against its static-table maxInputTokens (not the inflated live window);
			// only it implements getCondenseContextWindow, so others fall back to the full contextWindow.
			const contextWindow = this.api.getCondenseContextWindow?.() ?? modelInfo.contextWindow
			const useAvailableInputForContextPercent = typeof this.api.getCondenseContextWindow === "function"

			// Get the current profile ID using the helper method
			const currentProfileId = this.getCurrentProfileId(state)
			// Check if context management will likely run (threshold check)
			// This allows us to show an in-progress indicator to the user
			// We use the centralized willManageContext helper to avoid duplicating threshold logic
			const lastMessage = this.apiConversationHistory[this.apiConversationHistory.length - 1]
			const lastMessageContent = lastMessage?.content
			let lastMessageTokens = 0
			if (lastMessageContent) {