fix: handle context window exceeded errors during context condensing

src/core/condense/__tests__/index.spec.ts

@@ -1484,6 +1484,122 @@ describe("summarizeConversation", () => {
 
 		expect(result.error).toBeUndefined()
 	})
+
+	it("should return context window exceeded error when API throws context length error", async () => {
+		const messages: ApiMessage[] = [
+			{ role: "user", content: "Hello", ts: 1 },
+			{ role: "assistant", content: "Hi there", ts: 2 },
+			{ role: "user", content: "How are you?", ts: 3 },
+			{ role: "assistant", content: "I'm good", ts: 4 },
+			{ role: "user", content: "What's new?", ts: 5 },
+			{ role: "assistant", content: "Not much", ts: 6 },
+			{ role: "user", content: "Tell me more", ts: 7 },
+		]
+
+		// Create a stream that throws a context length exceeded error
+		// eslint-disable-next-line require-yield
+		const errorStream = (async function* () {
+			throw new Error(
+				"400 This endpoint's maximum context length is 131072 tokens. However, you requested about 145406 tokens.",
+			)
+		})()
+
+		mockApiHandler.createMessage = vi.fn().mockReturnValue(errorStream) as any
+
+		const result = await summarizeConversation(
+			messages,
+			mockApiHandler,
+			defaultSystemPrompt,
+			taskId,
+			DEFAULT_PREV_CONTEXT_TOKENS,
+		)
+
+		// Should return original messages with error
+		expect(result.messages).toEqual(messages)
+		expect(result.summary).toBe("")
+		expect(result.error).toBeTruthy()
+		// The error message comes from the translation key
+		expect(result.error).toContain("condense_context_window_exceeded")
+		expect(result.newContextTokens).toBeUndefined()
+	})
+
+	it("should return context window exceeded error for various context length error messages", async () => {
+		const messages: ApiMessage[] = [
+			{ role: "user", content: "Hello", ts: 1 },
+			{ role: "assistant", content: "Hi there", ts: 2 },
+			{ role: "user", content: "How are you?", ts: 3 },
+			{ role: "assistant", content: "I'm good", ts: 4 },
+			{ role: "user", content: "What's new?", ts: 5 },
+			{ role: "assistant", content: "Not much", ts: 6 },
+			{ role: "user", content: "Tell me more", ts: 7 },
+		]
+
+		// Test various error message patterns
+		const errorPatterns = [
+			"context_length_exceeded",
+			"maximum context length exceeded",
+			"prompt is too long",
+			"too many tokens in your request",
+		]
+
+		for (const errorPattern of errorPatterns) {
+			vi.clearAllMocks()
+
+			// eslint-disable-next-line require-yield
+			const errorStream = (async function* () {
+				throw new Error(errorPattern)
+			})()
+
+			mockApiHandler.createMessage = vi.fn().mockReturnValue(errorStream) as any
+
+			const result = await summarizeConversation(
+				messages,
+				mockApiHandler,
+				defaultSystemPrompt,
+				taskId,
+				DEFAULT_PREV_CONTEXT_TOKENS,
+			)
+
+			expect(result.error).toBeTruthy()
+			// The error message comes from the translation key
+			expect(result.error).toContain("condense_context_window_exceeded")
+		}
+	})
+
+	it("should return generic error for non-context-window API errors", async () => {
+		const messages: ApiMessage[] = [
+			{ role: "user", content: "Hello", ts: 1 },
+			{ role: "assistant", content: "Hi there", ts: 2 },
+			{ role: "user", content: "How are you?", ts: 3 },
+			{ role: "assistant", content: "I'm good", ts: 4 },
+			{ role: "user", content: "What's new?", ts: 5 },
+			{ role: "assistant", content: "Not much", ts: 6 },
+			{ role: "user", content: "Tell me more", ts: 7 },
+		]
+
+		// Create a stream that throws a generic API error
+		// eslint-disable-next-line require-yield
+		const errorStream = (async function* () {
+			throw new Error("Internal server error")
+		})()
+
+		mockApiHandler.createMessage = vi.fn().mockReturnValue(errorStream) as any
+
+		const result = await summarizeConversation(
+			messages,
+			mockApiHandler,
+			defaultSystemPrompt,
+			taskId,
+			DEFAULT_PREV_CONTEXT_TOKENS,
+		)
+
+		// Should return original messages with error
+		expect(result.messages).toEqual(messages)
+		expect(result.summary).toBe("")
+		expect(result.error).toBeTruthy()
+		expect(result.error).toContain("Internal server error")
+		expect(result.newContextTokens).toBeUndefined()
+	})
 })
 
 describe("summarizeConversation with custom settings", () => {

src/core/condense/index.ts

@@ -319,19 +319,45 @@ export async function summarizeConversation(
 		}
 	}
 
-	const stream = handlerToUse.createMessage(promptToUse, requestMessages)
-
 	let summary = ""
 	let cost = 0
 	let outputTokens = 0
 
-	for await (const chunk of stream) {
-		if (chunk.type === "text") {
-			summary += chunk.text
-		} else if (chunk.type === "usage") {
-			// Record final usage chunk only
-			cost = chunk.totalCost ?? 0
-			outputTokens = chunk.outputTokens ?? 0
+	try {
+		const stream = handlerToUse.createMessage(promptToUse, requestMessages)
+
+		for await (const chunk of stream) {
+			if (chunk.type === "text") {
+				summary += chunk.text
+			} else if (chunk.type === "usage") {
+				// Record final usage chunk only
+				cost = chunk.totalCost ?? 0
+				outputTokens = chunk.outputTokens ?? 0
+			}
+		}
+	} catch (error) {
+		const errorMessage = error instanceof Error ? error.message : String(error)
+
+		// Check for context window exceeded error
+		if (
+			errorMessage.includes("context_length_exceeded") ||
+			errorMessage.includes("context length") ||
+			errorMessage.includes("maximum context") ||
+			errorMessage.includes("too long") ||
+			errorMessage.includes("too many tokens")
+		) {
+			return {
+				...response,
+				cost,
+				error: t("common:errors.condense_context_window_exceeded"),
+			}
+		}
+
+		// Return generic error for other failures
+		return {
+			...response,
+			cost,
+			error: `${t("common:errors.condense_failed")}: ${errorMessage}`,
 		}
 	}
 

src/i18n/locales/en/common.json

@@ -63,6 +63,7 @@
 		"condensed_recently": "Context was condensed recently; skipping this attempt",
 		"condense_handler_invalid": "API handler for condensing context is invalid",
 		"condense_context_grew": "Context size increased during condensing; skipping this attempt",
+		"condense_context_window_exceeded": "The selected condensing model's context window is too small for the current conversation. Please select a model with a larger context window or manually reduce the conversation size first.",
 		"url_timeout": "The website took too long to load (timeout). This could be due to a slow connection, heavy website, or the site being temporarily unavailable. You can try again later or check if the URL is correct.",
 		"url_not_found": "The website address could not be found. Please check if the URL is correct and try again.",
 		"no_internet": "No internet connection. Please check your network connection and try again.",