fix(ai): omit tools field instead of sending empty array (#3650)

DashScope / Aliyun Qwen (OpenAI-compatible) rejects `tools: []`
with HTTP 400 `"[] is too short - 'tools'"`. Five providers used
a truthy check (`if (context.tools)`) that treated an empty array
as "send tools", so `pi --no-tools` produced `tools: []` in the
request body. Matching the Google provider's pattern, we now
guard on `context.tools.length > 0`:

- openai-completions.ts
- openai-responses.ts
- openai-codex-responses.ts
- azure-openai-responses.ts
- anthropic.ts

The openai-completions fallback that emits `tools: []` when the
conversation has tool history (required by LiteLLM / Anthropic
proxies) is preserved via the existing `else if (hasToolHistory)`
branch.

closes #3649

Co-authored-by: 槐聚 <huaiju@zbyte-inc.com>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Co-authored-by: Mario Zechner <badlogicgames@gmail.com>

packages/ai/CHANGELOG.md

@@ -2,6 +2,9 @@
 
 ## [Unreleased]
 
+### Fixed
+
+- Stopped sending `tools: []` on OpenAI-compatible, Anthropic, OpenAI Responses, OpenAI Codex Responses, and Azure OpenAI Responses requests when no tools are active (e.g. `pi --no-tools`). DashScope/Aliyun Qwen (OpenAI-compatible) rejects empty tools arrays with `"[] is too short - 'tools'"` (HTTP 400); the field is now omitted unless the conversation has tool history (the existing LiteLLM/Anthropic-proxy workaround).
 ## [0.70.2] - 2026-04-24
 
 ### Fixed

packages/ai/src/providers/anthropic.ts

@@ -877,7 +877,7 @@ function buildParams(
 		params.temperature = options.temperature;
 	}
 
-	if (context.tools) {
+	if (context.tools && context.tools.length > 0) {
 		params.tools = convertTools(
 			context.tools,
 			isOAuthToken,

packages/ai/src/providers/azure-openai-responses.ts

@@ -236,7 +236,7 @@ function buildParams(
 		params.temperature = options?.temperature;
 	}
 
-	if (context.tools) {
+	if (context.tools && context.tools.length > 0) {
 		params.tools = convertResponsesTools(context.tools);
 	}
 

packages/ai/src/providers/openai-codex-responses.ts

@@ -340,7 +340,7 @@ function buildRequestBody(
 		body.service_tier = options.serviceTier;
 	}
 
-	if (context.tools) {
+	if (context.tools && context.tools.length > 0) {
 		body.tools = convertResponsesTools(context.tools, { strict: null });
 	}
 

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

@@ -504,7 +504,7 @@ function buildParams(
 		params.temperature = options.temperature;
 	}
 
-	if (context.tools) {
+	if (context.tools && context.tools.length > 0) {
 		params.tools = convertTools(context.tools, compat);
 		if (compat.zaiToolStream) {
 			(params as any).tool_stream = true;

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

@@ -230,7 +230,7 @@ function buildParams(model: Model<"openai-responses">, context: Context, options
 		params.service_tier = options.serviceTier;
 	}
 
-	if (context.tools) {
+	if (context.tools && context.tools.length > 0) {
 		params.tools = convertResponsesTools(context.tools);
 	}
 

packages/ai/test/openai-completions-empty-tools.test.ts

@@ -0,0 +1,141 @@
+import { beforeEach, describe, expect, it, vi } from "vitest";
+import { getModel } from "../src/models.js";
+import { streamSimple } from "../src/stream.js";
+
+// Empty tools arrays must NOT be serialized as `tools: []` — some OpenAI-compatible
+// backends (e.g. DashScope / Aliyun Qwen via compatible-mode) reject the request with
+// `"[] is too short - 'tools'"` (HTTP 400) when `--no-tools` produces an empty array.
+// Regression for https://github.com/badlogic/pi-mono/issues/<issue-number>
+
+const mockState = vi.hoisted(() => ({
+	lastParams: undefined as unknown,
+}));
+
+vi.mock("openai", () => {
+	class FakeOpenAI {
+		chat = {
+			completions: {
+				create: (params: unknown) => {
+					mockState.lastParams = params;
+					const stream = {
+						async *[Symbol.asyncIterator]() {
+							yield {
+								choices: [{ delta: {}, finish_reason: "stop" }],
+								usage: {
+									prompt_tokens: 1,
+									completion_tokens: 1,
+									prompt_tokens_details: { cached_tokens: 0 },
+									completion_tokens_details: { reasoning_tokens: 0 },
+								},
+							};
+						},
+					};
+					const promise = Promise.resolve(stream) as Promise<typeof stream> & {
+						withResponse: () => Promise<{
+							data: typeof stream;
+							response: { status: number; headers: Headers };
+						}>;
+					};
+					promise.withResponse = async () => ({
+						data: stream,
+						response: { status: 200, headers: new Headers() },
+					});
+					return promise;
+				},
+			},
+		};
+	}
+
+	return { default: FakeOpenAI };
+});
+
+describe("openai-completions empty tools handling", () => {
+	beforeEach(() => {
+		mockState.lastParams = undefined;
+	});
+
+	it("omits tools field when context.tools is an empty array", async () => {
+		const { compat: _compat, ...baseModel } = getModel("openai", "gpt-4o-mini")!;
+		const model = { ...baseModel, api: "openai-completions" } as const;
+
+		await streamSimple(
+			model,
+			{
+				messages: [{ role: "user", content: "hi", timestamp: Date.now() }],
+				tools: [],
+			},
+			{ apiKey: "test" },
+		).result();
+
+		const params = mockState.lastParams as { tools?: unknown };
+		expect("tools" in (params as object)).toBe(false);
+	});
+
+	it("omits tools field when context.tools is undefined", async () => {
+		const { compat: _compat, ...baseModel } = getModel("openai", "gpt-4o-mini")!;
+		const model = { ...baseModel, api: "openai-completions" } as const;
+
+		await streamSimple(
+			model,
+			{
+				messages: [{ role: "user", content: "hi", timestamp: Date.now() }],
+			},
+			{ apiKey: "test" },
+		).result();
+
+		const params = mockState.lastParams as { tools?: unknown };
+		expect("tools" in (params as object)).toBe(false);
+	});
+
+	it("still emits tools: [] for Anthropic/LiteLLM proxy when conversation has tool history", async () => {
+		const { compat: _compat, ...baseModel } = getModel("openai", "gpt-4o-mini")!;
+		const model = { ...baseModel, api: "openai-completions" } as const;
+
+		await streamSimple(
+			model,
+			{
+				messages: [
+					{ role: "user", content: "use the tool", timestamp: Date.now() },
+					{
+						role: "assistant",
+						content: [
+							{
+								type: "toolCall",
+								id: "t1",
+								name: "noop",
+								arguments: {},
+							},
+						],
+						stopReason: "toolUse",
+						usage: {
+							input: 0,
+							output: 0,
+							cacheRead: 0,
+							cacheWrite: 0,
+							totalTokens: 0,
+							cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+						},
+						api: "openai-completions",
+						provider: "openai",
+						model: "gpt-4o-mini",
+						timestamp: Date.now(),
+					},
+					{
+						role: "toolResult",
+						toolCallId: "t1",
+						toolName: "noop",
+						content: [{ type: "text", text: "done" }],
+						isError: false,
+						timestamp: Date.now(),
+					},
+				],
+				tools: [],
+			},
+			{ apiKey: "test" },
+		).result();
+
+		const params = mockState.lastParams as { tools?: unknown[] };
+		expect(Array.isArray(params.tools)).toBe(true);
+		expect(params.tools).toEqual([]);
+	});
+});