[Fix] Ollama tasks condense before their configured context limit

src/api/providers/__tests__/native-ollama.spec.ts

@@ -910,6 +910,50 @@ describe("NativeOllamaHandler", () => {
 		})
 	})
 
+	describe("ensureModelFetched", () => {
+		it("makes the detected context window available before a request", async () => {
+			mockGetOllamaModels.mockResolvedValueOnce({
+				llama2: {
+					contextWindow: 1_000_000,
+					maxTokens: 4096,
+					supportsImages: false,
+					supportsPromptCache: true,
+				},
+			})
+
+			expect(handler.getModel().info.contextWindow).toBe(128_000)
+
+			await handler.ensureModelFetched()
+
+			expect(handler.getModel().info.contextWindow).toBe(1_000_000)
+			expect(mockGetOllamaModels).toHaveBeenCalledTimes(1)
+		})
+
+		it("skips subsequent fetches after models are populated", async () => {
+			await handler.ensureModelFetched()
+			mockGetOllamaModels.mockClear()
+
+			await handler.fetchModel()
+
+			expect(mockGetOllamaModels).not.toHaveBeenCalled()
+		})
+
+		it("deduplicates concurrent metadata fetches", async () => {
+			await Promise.all([handler.ensureModelFetched(), handler.ensureModelFetched()])
+
+			expect(mockGetOllamaModels).toHaveBeenCalledTimes(1)
+		})
+
+		it("allows a later fetch after a rejected request", async () => {
+			mockGetOllamaModels.mockRejectedValueOnce(new Error("network down"))
+
+			await expect(handler.ensureModelFetched()).rejects.toThrow("network down")
+			await handler.ensureModelFetched()
+
+			expect(mockGetOllamaModels).toHaveBeenCalledTimes(2)
+		})
+	})
+
 	describe("tool calling", () => {
 		it("should include tools when tools are provided", async () => {
 			// Model metadata should not gate tool inclusion; metadata.tools controls it.

src/api/providers/native-ollama.ts

@@ -227,6 +227,7 @@ export class NativeOllamaHandler extends BaseProvider implements SingleCompletio
 	protected options: ApiHandlerOptions
 	private client: Ollama | undefined
 	protected models: Record<string, ModelInfo> = {}
+	private modelFetchPromise?: Promise<{ id: string; info: ModelInfo }>
 
 	constructor(options: ApiHandlerOptions) {
 		super()
@@ -538,8 +539,26 @@ export class NativeOllamaHandler extends BaseProvider implements SingleCompletio
 	}
 
 	async fetchModel() {
-		this.models = await getOllamaModels(this.options.ollamaBaseUrl, this.options.ollamaApiKey)
-		return this.getModel()
+		if (Object.keys(this.models).length > 0) {
+			return this.getModel()
+		}
+
+		if (!this.modelFetchPromise) {
+			this.modelFetchPromise = getOllamaModels(this.options.ollamaBaseUrl, this.options.ollamaApiKey)
+				.then((models) => {
+					this.models = models
+					return this.getModel()
+				})
+				.finally(() => {
+					this.modelFetchPromise = undefined
+				})
+		}
+
+		return this.modelFetchPromise
+	}
+
+	async ensureModelFetched(): Promise<void> {
+		await this.fetchModel()
 	}
 
 	override getModel(): { id: string; info: ModelInfo } {