feat(ai): add Xiaomi MiMo provider (#4005)

* fix(ai): include minimax-cn in cross-provider-handoff matrix

* feat(ai): add Xiaomi MiMo provider

Adds Xiaomi MiMo as an openai-completions-compatible provider.

- packages/ai: register provider in types/KnownProvider, env-api-keys (XIAOMI_API_KEY), generate-models, models.generated.ts, overflow util, README, CHANGELOG
- packages/ai/test: extend stream, tokens, abort, empty, context-overflow, overflow, image-tool-result, tool-call-without-result, total-tokens, unicode-surrogate, cross-provider-handoff matrices with Xiaomi
- packages/coding-agent: default model (mimo-v2.5-pro), display name (Xiaomi MiMo), CLI env var docs, README, docs/providers.md

closes #3912

---------

Co-authored-by: Mario Zechner <badlogicgames@gmail.com>

packages/ai/src/env-api-keys.ts

@@ -122,6 +122,7 @@ function getApiKeyEnvVars(provider: string): readonly string[] | undefined {
 		"kimi-coding": "KIMI_API_KEY",
 		"cloudflare-workers-ai": "CLOUDFLARE_API_KEY",
 		"cloudflare-ai-gateway": "CLOUDFLARE_API_KEY",
+		xiaomi: "XIAOMI_API_KEY",
 	};
 
 	const envVar = envMap[provider];

packages/ai/src/models.generated.ts

@@ -16231,6 +16231,93 @@ export const MODELS = {
 			maxTokens: 4096,
 		} satisfies Model<"openai-completions">,
 	},
+	"xiaomi": {
+		"mimo-v2-flash": {
+			id: "mimo-v2-flash",
+			name: "MiMo-V2-Flash",
+			api: "openai-completions",
+			provider: "xiaomi",
+			baseUrl: "https://api.xiaomimimo.com/v1",
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 0.1,
+				output: 0.3,
+				cacheRead: 0.01,
+				cacheWrite: 0,
+			},
+			contextWindow: 256000,
+			maxTokens: 64000,
+		} satisfies Model<"openai-completions">,
+		"mimo-v2-omni": {
+			id: "mimo-v2-omni",
+			name: "MiMo-V2-Omni",
+			api: "openai-completions",
+			provider: "xiaomi",
+			baseUrl: "https://api.xiaomimimo.com/v1",
+			reasoning: true,
+			input: ["text", "image"],
+			cost: {
+				input: 0.4,
+				output: 2,
+				cacheRead: 0.08,
+				cacheWrite: 0,
+			},
+			contextWindow: 256000,
+			maxTokens: 128000,
+		} satisfies Model<"openai-completions">,
+		"mimo-v2-pro": {
+			id: "mimo-v2-pro",
+			name: "MiMo-V2-Pro",
+			api: "openai-completions",
+			provider: "xiaomi",
+			baseUrl: "https://api.xiaomimimo.com/v1",
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 1,
+				output: 3,
+				cacheRead: 0.2,
+				cacheWrite: 0,
+			},
+			contextWindow: 1000000,
+			maxTokens: 128000,
+		} satisfies Model<"openai-completions">,
+		"mimo-v2.5": {
+			id: "mimo-v2.5",
+			name: "MiMo-V2.5",
+			api: "openai-completions",
+			provider: "xiaomi",
+			baseUrl: "https://api.xiaomimimo.com/v1",
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 0.4,
+				output: 2,
+				cacheRead: 0.08,
+				cacheWrite: 0,
+			},
+			contextWindow: 1048576,
+			maxTokens: 131072,
+		} satisfies Model<"openai-completions">,
+		"mimo-v2.5-pro": {
+			id: "mimo-v2.5-pro",
+			name: "MiMo-V2.5-Pro",
+			api: "openai-completions",
+			provider: "xiaomi",
+			baseUrl: "https://api.xiaomimimo.com/v1",
+			reasoning: true,
+			input: ["text", "image"],
+			cost: {
+				input: 1,
+				output: 3,
+				cacheRead: 0.2,
+				cacheWrite: 0,
+			},
+			contextWindow: 1048576,
+			maxTokens: 131072,
+		} satisfies Model<"openai-completions">,
+	},
 	"zai": {
 		"glm-4.5-air": {
 			id: "glm-4.5-air",

packages/ai/src/types.ts

@@ -42,7 +42,8 @@ export type KnownProvider =
 	| "opencode-go"
 	| "kimi-coding"
 	| "cloudflare-workers-ai"
-	| "cloudflare-ai-gateway";
+	| "cloudflare-ai-gateway"
+	| "xiaomi";
 export type Provider = KnownProvider | string;
 
 export type ThinkingLevel = "minimal" | "low" | "medium" | "high" | "xhigh";

packages/ai/src/utils/overflow.ts

@@ -23,6 +23,9 @@ import type { AssistantMessage } from "../types.js";
  * - Cerebras: "400/413 status code (no body)"
  * - Mistral: "Prompt contains X tokens ... too large for model with Y maximum context length"
  * - z.ai: Does NOT error, accepts overflow silently - handled via usage.input > contextWindow
+ * - Xiaomi MiMo: Truncates input to fill contextWindow exactly, then returns finish_reason "length"
+ *   with output=0 (no room left to generate). Detected via stopReason "length" + zero output +
+ *   input filling the context window.
  * - Ollama: Some deployments truncate silently, others return errors like "prompt too long; exceeded max context length by X tokens"
  */
 const OVERFLOW_PATTERNS = [
@@ -90,6 +93,8 @@ const NON_OVERFLOW_PATTERNS = [
  * **Unreliable detection:**
  * - z.ai: Sometimes accepts overflow silently (detectable via usage.input > contextWindow),
  *   sometimes returns rate limit errors. Pass contextWindow param to detect silent overflow.
+ * - Xiaomi MiMo: Truncates input to fit contextWindow then returns stopReason "length" with
+ *   output=0. Pass contextWindow param to detect via the "filled context + zero output" signal.
  * - Ollama: May truncate input silently for some setups, but may also return explicit
  *   overflow errors that match the patterns above. Silent truncation still cannot be
  *   detected here because we do not know the expected token count.
@@ -127,6 +132,16 @@ export function isContextOverflow(message: AssistantMessage, contextWindow?: num
 		}
 	}
 
+	// Case 3: Length-stop overflow (Xiaomi MiMo style) - server truncates oversized input
+	// to fit the context window, leaving no room for output. Returns stopReason "length"
+	// with output=0 and input+cacheRead filling the context window.
+	if (contextWindow && message.stopReason === "length" && message.usage.output === 0) {
+		const inputTokens = message.usage.input + message.usage.cacheRead;
+		if (inputTokens >= contextWindow * 0.99) {
+			return true;
+		}
+	}
+
 	return false;
 }