fix(ai): add Xiaomi reasoning replay compat

closes #4678
2026-05-18 11:14:54 +02:00
parent dbadffdfae
commit b8f51957a0
4 changed files with 103 additions and 4 deletions
--- a/packages/ai/CHANGELOG.md
+++ b/packages/ai/CHANGELOG.md
@@ -2,6 +2,10 @@

 ## [Unreleased]

+### Fixed
+
+- Fixed Xiaomi MiMo generated model metadata to replay assistant tool-call messages with `reasoning_content` for thinking-mode multi-turn requests ([#4678](https://github.com/earendil-works/pi/issues/4678)).
+
 ## [0.75.1] - 2026-05-18

 ### Fixed
--- a/packages/ai/scripts/generate-models.ts
+++ b/packages/ai/scripts/generate-models.ts
@@ -1073,6 +1073,10 @@ async function loadModelsDevData(): Promise<Model<any>[]> {
 		// Built-in `xiaomi` targets the API billing endpoint (single stable URL,
 		// keys from platform.xiaomimimo.com). The three `xiaomi-token-plan-*`
 		// providers cover prepaid Token Plan endpoints in cn / ams / sgp.
+		const xiaomiCompat: OpenAICompletionsCompat = {
+			requiresReasoningContentOnAssistantMessages: true,
+			thinkingFormat: "deepseek",
+		};
 		const xiaomiVariants = [
 			{ provider: "xiaomi", baseUrl: "https://api.xiaomimimo.com/v1" },
 			{ provider: "xiaomi-token-plan-cn", baseUrl: "https://token-plan-cn.xiaomimimo.com/v1" },
@@ -1092,6 +1096,7 @@ async function loadModelsDevData(): Promise<Model<any>[]> {
 						api: "openai-completions",
 						provider,
 						baseUrl,
+						compat: xiaomiCompat,
 						reasoning: m.reasoning === true,
 						input: m.modalities?.input?.includes("image") ? ["text", "image"] : ["text"],
 						cost: {
--- a/packages/ai/src/models.generated.ts
+++ b/packages/ai/src/models.generated.ts
@@ -7268,9 +7268,9 @@ export const MODELS = {
 		"big-pickle": {
 			id: "big-pickle",
 			name: "Big Pickle",
-			api: "openai-completions",
+			api: "anthropic-messages",
 			provider: "opencode",
-			baseUrl: "https://opencode.ai/zen/v1",
+			baseUrl: "https://opencode.ai/zen",
 			reasoning: true,
 			input: ["text"],
 			cost: {
@@ -7281,7 +7281,7 @@ export const MODELS = {
 			},
 			contextWindow: 200000,
 			maxTokens: 128000,
-		} satisfies Model<"openai-completions">,
+		} satisfies Model<"anthropic-messages">,
 		"claude-haiku-4-5": {
 			id: "claude-haiku-4-5",
 			name: "Claude Haiku 4.5",
@@ -15879,6 +15879,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi",
 			baseUrl: "https://api.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text"],
 			cost: {
@@ -15896,6 +15897,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi",
 			baseUrl: "https://api.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text", "image"],
 			cost: {
@@ -15913,6 +15915,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi",
 			baseUrl: "https://api.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text"],
 			cost: {
@@ -15930,6 +15933,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi",
 			baseUrl: "https://api.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text", "image"],
 			cost: {
@@ -15947,6 +15951,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi",
 			baseUrl: "https://api.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text"],
 			cost: {
@@ -15966,6 +15971,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-ams",
 			baseUrl: "https://token-plan-ams.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text"],
 			cost: {
@@ -15983,6 +15989,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-ams",
 			baseUrl: "https://token-plan-ams.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text", "image"],
 			cost: {
@@ -16000,6 +16007,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-ams",
 			baseUrl: "https://token-plan-ams.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text"],
 			cost: {
@@ -16017,6 +16025,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-ams",
 			baseUrl: "https://token-plan-ams.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text", "image"],
 			cost: {
@@ -16034,6 +16043,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-ams",
 			baseUrl: "https://token-plan-ams.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text"],
 			cost: {
@@ -16053,6 +16063,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-cn",
 			baseUrl: "https://token-plan-cn.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text"],
 			cost: {
@@ -16070,6 +16081,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-cn",
 			baseUrl: "https://token-plan-cn.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text", "image"],
 			cost: {
@@ -16087,6 +16099,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-cn",
 			baseUrl: "https://token-plan-cn.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text"],
 			cost: {
@@ -16104,6 +16117,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-cn",
 			baseUrl: "https://token-plan-cn.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text", "image"],
 			cost: {
@@ -16121,6 +16135,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-cn",
 			baseUrl: "https://token-plan-cn.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text"],
 			cost: {
@@ -16140,6 +16155,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-sgp",
 			baseUrl: "https://token-plan-sgp.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text"],
 			cost: {
@@ -16157,6 +16173,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-sgp",
 			baseUrl: "https://token-plan-sgp.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text", "image"],
 			cost: {
@@ -16174,6 +16191,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-sgp",
 			baseUrl: "https://token-plan-sgp.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text"],
 			cost: {
@@ -16191,6 +16209,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-sgp",
 			baseUrl: "https://token-plan-sgp.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text", "image"],
 			cost: {
@@ -16208,6 +16227,7 @@ export const MODELS = {
 			api: "openai-completions",
 			provider: "xiaomi-token-plan-sgp",
 			baseUrl: "https://token-plan-sgp.xiaomimimo.com/v1",
+			compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
 			reasoning: true,
 			input: ["text"],
 			cost: {
--- a/packages/ai/test/openai-completions-tool-choice.test.ts
+++ b/packages/ai/test/openai-completions-tool-choice.test.ts
@@ -3,7 +3,7 @@ import { beforeEach, describe, expect, it, vi } from "vitest";
 import { getModel } from "../src/models.js";
 import { convertMessages } from "../src/providers/openai-completions.js";
 import { streamSimple } from "../src/stream.js";
-import type { Model, Tool } from "../src/types.js";
+import type { AssistantMessage, Model, Tool, ToolResultMessage } from "../src/types.js";

 const mockState = vi.hoisted(() => ({
 	lastParams: undefined as unknown,
@@ -817,6 +817,76 @@ describe("openai-completions tool_choice", () => {
 		expect(writeCall).not.toHaveProperty("partialArgs");
 	});

+	it("stores Xiaomi MiMo reasoning replay compat in built-in metadata", () => {
+		const providers = ["xiaomi", "xiaomi-token-plan-cn", "xiaomi-token-plan-ams", "xiaomi-token-plan-sgp"] as const;
+
+		for (const provider of providers) {
+			const model = getModel(provider, "mimo-v2.5-pro")!;
+			expect(model.compat?.requiresReasoningContentOnAssistantMessages).toBe(true);
+			expect(model.compat?.thinkingFormat).toBe("deepseek");
+			expect(model.compat?.maxTokensField).toBeUndefined();
+			expect(model.compat?.supportsDeveloperRole).toBeUndefined();
+		}
+	});
+
+	it("replays Xiaomi MiMo assistant tool calls with empty reasoning_content when thinking is missing", async () => {
+		const model = getModel("xiaomi", "mimo-v2.5-pro")!;
+		const assistantMessage: AssistantMessage = {
+			role: "assistant",
+			api: "openai-completions",
+			provider: "xiaomi",
+			model: "mimo-v2.5-pro",
+			content: [{ type: "toolCall", id: "call_1", name: "read", arguments: { path: "README.md" } }],
+			usage: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+				totalTokens: 0,
+				cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+			},
+			stopReason: "toolUse",
+			timestamp: Date.now(),
+		};
+		const toolResult: ToolResultMessage = {
+			role: "toolResult",
+			toolCallId: "call_1",
+			toolName: "read",
+			content: [{ type: "text", text: "contents" }],
+			isError: false,
+			timestamp: Date.now(),
+		};
+		let payload: unknown;
+
+		await streamSimple(
+			model,
+			{
+				messages: [
+					{ role: "user", content: "Read README.md", timestamp: Date.now() },
+					assistantMessage,
+					toolResult,
+				],
+			},
+			{
+				apiKey: "test",
+				reasoning: "high",
+				onPayload: (params: unknown) => {
+					payload = params;
+				},
+			},
+		).result();
+
+		const params = (payload ?? mockState.lastParams) as {
+			messages?: Array<Record<string, unknown>>;
+			thinking?: { type?: string };
+			reasoning_effort?: string;
+		};
+		const replayedAssistant = params.messages?.find((message) => message.role === "assistant");
+		expect(replayedAssistant).toMatchObject({ role: "assistant", reasoning_content: "" });
+		expect(params.thinking).toEqual({ type: "enabled" });
+		expect(params.reasoning_effort).toBe("high");
+	});
+
 	it("normalizes OpenCode Go reasoning deltas to reasoning_content for replay", async () => {
 		mockState.chunks = [
 			{