fix(ai/ollama): discover native context length from /api/show

Previously, Ollama model discovery hardcoded contextWindow=128000 and
maxTokens=8192 regardless of the model's actual capacity. That caused
OMP's planner and overflow detector to disagree with Ollama — e.g. a
4k-context model would be planned against 128k, and a 32k-context model
would be silently clamped by OMP.

For each entry returned by /api/tags, POST /api/show and read the
first '<arch>.context_length' field from 'model_info' to pick up the
model's native context window. Fall back to Ollama's default 4096
when the endpoint or field is missing so discovery still works against
older builds. maxTokens is left at 8192 to match the existing default.

Fixes #760

packages/ai/src/provider-models/openai-compat.ts

@@ -246,28 +246,66 @@ async function fetchOllamaNativeModels(baseUrl: string): Promise<Model<"openai-r
 	}
 	const payload = (await response.json()) as { models?: Array<{ name?: string; model?: string }> };
 	const entries = payload.models ?? [];
-	const models: Model<"openai-responses">[] = [];
-	for (const entry of entries) {
-		const id = entry.model ?? entry.name;
-		if (!id) {
-			continue;
-		}
-		models.push({
-			id,
-			name: entry.name ?? id,
-			api: "openai-responses",
-			provider: "ollama",
-			baseUrl,
-			reasoning: false,
-			input: ["text"],
-			cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
-			contextWindow: 128000,
-			maxTokens: 8192,
-		});
-	}
+	const resolved = await Promise.all(
+		entries.map(async (entry): Promise<Model<"openai-responses"> | null> => {
+			const id = entry.model ?? entry.name;
+			if (!id) return null;
+			const { contextWindow, maxTokens } = await fetchOllamaModelLimits(nativeBaseUrl, id);
+			return {
+				id,
+				name: entry.name ?? id,
+				api: "openai-responses",
+				provider: "ollama",
+				baseUrl,
+				reasoning: false,
+				input: ["text"],
+				cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+				contextWindow,
+				maxTokens,
+			};
+		}),
+	);
+	const models: Model<"openai-responses">[] = resolved.filter((m): m is Model<"openai-responses"> => m !== null);
 	return models.sort((left, right) => left.id.localeCompare(right.id));
 }
 
+/** Ollama's default `num_ctx` when the runtime request does not override it. */
+const OLLAMA_DEFAULT_CONTEXT_WINDOW = 4096;
+/** Cap max output tokens at a value that matches OMP's other openai-responses defaults. */
+const OLLAMA_DEFAULT_MAX_TOKENS = 8192;
+
+/**
+	* Query Ollama's `/api/show` endpoint for a single model and pull its native
+	* context length out of `model_info.<arch>.context_length`. Falls back to
+	* Ollama's default context window when the endpoint or field is unavailable
+	* so discovery still succeeds against older Ollama builds.
+	*/
+async function fetchOllamaModelLimits(
+	nativeBaseUrl: string,
+	modelId: string,
+): Promise<{ contextWindow: number; maxTokens: number }> {
+	try {
+		const response = await fetch(`${nativeBaseUrl}/api/show`, {
+			method: "POST",
+			headers: { "Content-Type": "application/json", Accept: "application/json" },
+			body: JSON.stringify({ model: modelId }),
+		});
+		if (!response.ok) {
+			return { contextWindow: OLLAMA_DEFAULT_CONTEXT_WINDOW, maxTokens: OLLAMA_DEFAULT_MAX_TOKENS };
+		}
+		const payload = (await response.json()) as { model_info?: Record<string, unknown> };
+		const info = payload.model_info ?? {};
+		for (const [key, value] of Object.entries(info)) {
+			if (key.endsWith(".context_length") && typeof value === "number" && value > 0) {
+				return { contextWindow: value, maxTokens: OLLAMA_DEFAULT_MAX_TOKENS };
+			}
+		}
+	} catch {
+		// fall through to default
+	}
+	return { contextWindow: OLLAMA_DEFAULT_CONTEXT_WINDOW, maxTokens: OLLAMA_DEFAULT_MAX_TOKENS };
+}
+
 const OPENAI_NON_RESPONSES_PREFIXES = [
 	"text-embedding",
 	"whisper-",