fix(ai): map groq qwen3 reasoning effort values closes #1745

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

@@ -428,7 +428,7 @@ function buildParams(model: Model<"openai-completions">, context: Context, optio
 		(params as any).enable_thinking = !!options?.reasoningEffort;
 	} else if (options?.reasoningEffort && model.reasoning && compat.supportsReasoningEffort) {
 		// OpenAI-style reasoning_effort
-		params.reasoning_effort = options.reasoningEffort;
+		(params as any).reasoning_effort = mapReasoningEffort(options.reasoningEffort, compat.reasoningEffortMap);
 	}
 
 	// OpenRouter provider routing preferences
@@ -450,6 +450,13 @@ function buildParams(model: Model<"openai-completions">, context: Context, optio
 	return params;
 }
 
+function mapReasoningEffort(
+	effort: NonNullable<OpenAICompletionsOptions["reasoningEffort"]>,
+	reasoningEffortMap: Partial<Record<NonNullable<OpenAICompletionsOptions["reasoningEffort"]>, string>>,
+): string {
+	return reasoningEffortMap[effort] ?? effort;
+}
+
 function maybeAddOpenRouterAnthropicCacheControl(
 	model: Model<"openai-completions">,
 	messages: ChatCompletionMessageParam[],
@@ -777,13 +784,26 @@ function detectCompat(model: Model<"openai-completions">): Required<OpenAIComple
 	const useMaxTokens = provider === "mistral" || baseUrl.includes("mistral.ai") || baseUrl.includes("chutes.ai");
 
 	const isGrok = provider === "xai" || baseUrl.includes("api.x.ai");
+	const isGroq = provider === "groq" || baseUrl.includes("groq.com");
 
 	const isMistral = provider === "mistral" || baseUrl.includes("mistral.ai");
 
+	const reasoningEffortMap =
+		isGroq && model.id === "qwen/qwen3-32b"
+			? {
+					minimal: "default",
+					low: "default",
+					medium: "default",
+					high: "default",
+					xhigh: "default",
+				}
+			: {};
+
 	return {
 		supportsStore: !isNonStandard,
 		supportsDeveloperRole: !isNonStandard,
 		supportsReasoningEffort: !isGrok && !isZai,
+		reasoningEffortMap,
 		supportsUsageInStreaming: true,
 		maxTokensField: useMaxTokens ? "max_tokens" : "max_completion_tokens",
 		requiresToolResultName: isMistral,
@@ -809,6 +829,7 @@ function getCompat(model: Model<"openai-completions">): Required<OpenAICompletio
 		supportsStore: model.compat.supportsStore ?? detected.supportsStore,
 		supportsDeveloperRole: model.compat.supportsDeveloperRole ?? detected.supportsDeveloperRole,
 		supportsReasoningEffort: model.compat.supportsReasoningEffort ?? detected.supportsReasoningEffort,
+		reasoningEffortMap: model.compat.reasoningEffortMap ?? detected.reasoningEffortMap,
 		supportsUsageInStreaming: model.compat.supportsUsageInStreaming ?? detected.supportsUsageInStreaming,
 		maxTokensField: model.compat.maxTokensField ?? detected.maxTokensField,
 		requiresToolResultName: model.compat.requiresToolResultName ?? detected.requiresToolResultName,

packages/ai/src/types.ts

@@ -235,6 +235,8 @@ export interface OpenAICompletionsCompat {
 	supportsDeveloperRole?: boolean;
 	/** Whether the provider supports `reasoning_effort`. Default: auto-detected from URL. */
 	supportsReasoningEffort?: boolean;
+	/** Optional mapping from pi-ai reasoning levels to provider/model-specific `reasoning_effort` values. */
+	reasoningEffortMap?: Partial<Record<ThinkingLevel, string>>;
 	/** Whether the provider supports `stream_options: { include_usage: true }` for token usage in streaming responses. Default: true. */
 	supportsUsageInStreaming?: boolean;
 	/** Which field to use for max tokens. Default: auto-detected from URL. */

packages/ai/test/openai-completions-tool-choice.test.ts

@@ -119,4 +119,60 @@ describe("openai-completions tool_choice", () => {
 		expect(tool?.strict).toBeUndefined();
 		expect("strict" in (tool ?? {})).toBe(false);
 	});
+
+	it("maps groq qwen3 reasoning levels to default reasoning_effort", async () => {
+		const model = getModel("groq", "qwen/qwen3-32b")!;
+		let payload: unknown;
+
+		await streamSimple(
+			model,
+			{
+				messages: [
+					{
+						role: "user",
+						content: "Hi",
+						timestamp: Date.now(),
+					},
+				],
+			},
+			{
+				apiKey: "test",
+				reasoning: "medium",
+				onPayload: (params: unknown) => {
+					payload = params;
+				},
+			},
+		).result();
+
+		const params = (payload ?? mockState.lastParams) as { reasoning_effort?: string };
+		expect(params.reasoning_effort).toBe("default");
+	});
+
+	it("keeps normal reasoning_effort for groq models without compat mapping", async () => {
+		const model = getModel("groq", "openai/gpt-oss-20b")!;
+		let payload: unknown;
+
+		await streamSimple(
+			model,
+			{
+				messages: [
+					{
+						role: "user",
+						content: "Hi",
+						timestamp: Date.now(),
+					},
+				],
+			},
+			{
+				apiKey: "test",
+				reasoning: "medium",
+				onPayload: (params: unknown) => {
+					payload = params;
+				},
+			},
+		).result();
+
+		const params = (payload ?? mockState.lastParams) as { reasoning_effort?: string };
+		expect(params.reasoning_effort).toBe("medium");
+	});
 });

packages/ai/test/openai-completions-tool-result-images.test.ts

@@ -23,6 +23,7 @@ const compat: Required<OpenAICompletionsCompat> = {
 	supportsStore: true,
 	supportsDeveloperRole: true,
 	supportsReasoningEffort: true,
+	reasoningEffortMap: {},
 	supportsUsageInStreaming: true,
 	maxTokensField: "max_completion_tokens",
 	requiresToolResultName: false,

packages/coding-agent/docs/custom-provider.md

@@ -172,10 +172,17 @@ models: [{
   // ...
   compat: {
     supportsDeveloperRole: false,      // use "system" instead of "developer"
-    supportsReasoningEffort: false,    // disable reasoning_effort param
+    supportsReasoningEffort: true,
+    reasoningEffortMap: {              // map pi-ai levels to provider values
+      minimal: "default",
+      low: "default",
+      medium: "default",
+      high: "default",
+      xhigh: "default"
+    },
     maxTokensField: "max_tokens",      // instead of "max_completion_tokens"
     requiresToolResultName: true,      // tool results need name field
-    requiresMistralToolIds: true       // tool IDs must be 9 alphanumeric chars
+    requiresMistralToolIds: true,
     thinkingFormat: "qwen"             // uses enable_thinking: true
   }
 }]
@@ -568,6 +575,7 @@ interface ProviderModelConfig {
     supportsStore?: boolean;
     supportsDeveloperRole?: boolean;
     supportsReasoningEffort?: boolean;
+    reasoningEffortMap?: Partial<Record<"minimal" | "low" | "medium" | "high" | "xhigh", string>>;
     supportsUsageInStreaming?: boolean;
     maxTokensField?: "max_completion_tokens" | "max_tokens";
     requiresToolResultName?: boolean;