feat(ai): added OpenRouter provider with strict mode detection

- Added support for OpenRouter provider with strict mode detection.

packages/ai/CHANGELOG.md

@@ -1,13 +1,13 @@
 # Changelog
 
 ## [Unreleased]
-
 ### Breaking Changes
 
 - Removed `coerceNullStrings` function and its automatic null-string coercion behavior from JSON parsing
 
 ### Added
 
+- Added support for OpenRouter provider with strict mode detection
 - Added automatic cleaning of literal escape sequences (`\n`, `\t`, `\r`) in JSON parsing to handle LLM encoding confusion
 - Added support for healing JSON with trailing junk after balanced containers (e.g., `]\n</invoke>`)
 - Added `CODEX_STARTUP_EVENT_CHANNEL` constant and `CodexStartupEvent` type for monitoring Codex provider initialization status

packages/ai/src/providers/openai-completions-compat.ts

@@ -13,6 +13,7 @@ export type ResolvedOpenAICompat = Required<
 function detectStrictModeSupport(provider: string, baseUrl: string): boolean {
 	if (
 		provider === "openai" ||
+		provider === "openrouter" ||
 		provider === "cerebras" ||
 		provider === "together" ||
 		provider === "github-copilot" ||
@@ -28,6 +29,7 @@ function detectStrictModeSupport(provider: string, baseUrl: string): boolean {
 		normalizedBaseUrl.includes("models.inference.ai.azure.com") ||
 		normalizedBaseUrl.includes("api.cerebras.ai") ||
 		normalizedBaseUrl.includes("api.together.xyz") ||
+		normalizedBaseUrl.includes("openrouter.ai") ||
 		normalizedBaseUrl.includes("api.deepseek.com") ||
 		normalizedBaseUrl.includes("deepseek.com")
 	);

packages/ai/test/openai-tool-strict-mode.test.ts

@@ -91,6 +91,16 @@ describe("OpenAI tool strict mode", () => {
 		};
 		expect(payload.tools?.[0]?.function?.strict).toBe(true);
 	});
+
+	it("sends strict=true for openai-completions tool schemas on OpenRouter", async () => {
+		const model = getBundledModel("openrouter", "anthropic/claude-sonnet-4") as Model<"openai-completions">;
+
+		const payload = (await captureCompletionsPayload(model)) as {
+			tools?: Array<{ function?: { strict?: boolean } }>;
+		};
+		expect(payload.tools?.[0]?.function?.strict).toBe(true);
+	});
+
 	it("omits stream_options usage requests for Cerebras chat completions", async () => {
 		const model = getBundledModel("cerebras", "gpt-oss-120b") as Model<"openai-completions">;