feat(providers): add MiniMax provider support

.changeset/add-minimax-provider.md

@@ -0,0 +1,6 @@
+---
+'@open-codesign/shared': minor
+'@open-codesign/providers': patch
+---
+
+Add MiniMax as a built-in provider with static M3 and M2.7 model choices, regional endpoint guidance, and Codex environment-key import support.

MINIMAX.md

@@ -0,0 +1,43 @@
+# MiniMax Provider
+
+Open CoDesign includes MiniMax as a built-in provider. The default entry uses the global English OpenAI-compatible endpoint and exposes `MiniMax-M3` and `MiniMax-M2.7` as static model choices.
+
+## Configuration
+
+- Provider ID: `minimax`
+- API key environment variable: `MINIMAX_API_KEY`
+- Default model: `MiniMax-M3`
+- Model choices: `MiniMax-M3`, `MiniMax-M2.7`
+- Official documentation: `https://platform.minimax.io/docs/api-reference/api-overview`
+- China documentation: `https://platform.minimaxi.com/docs/api-reference/api-overview`
+
+The built-in provider uses `https://api.minimax.io/v1`. To use another region or wire format, add a custom provider in Settings and use one of these public bases:
+
+| Region | Wire | Base URL | Documentation |
+| --- | --- | --- | --- |
+| `global_en` | OpenAI-compatible | `https://api.minimax.io/v1` | `https://platform.minimax.io/docs` |
+| `global_en` | Anthropic-compatible | `https://api.minimax.io/anthropic` | `https://platform.minimax.io/docs` |
+| `cn_zh` | OpenAI-compatible | `https://api.minimaxi.com/v1` | `https://platform.minimaxi.com/docs` |
+| `cn_zh` | Anthropic-compatible | `https://api.minimaxi.com/anthropic` | `https://platform.minimaxi.com/docs` |
+
+For an Anthropic-compatible entry, enter the base directly without adding `/v1`. Open CoDesign derives the versioned `/v1/messages` request path internally. For an OpenAI-compatible entry, keep `/v1` in the public base; the client appends the chat-completions path.
+
+## Model catalog
+
+Prices are USD per million tokens. The model metadata below mirrors the shared provider registry.
+
+| Model | Context | Input | Output | Cache read | Cache write | Input modalities | Thinking |
+| --- | ---: | ---: | ---: | ---: | ---: | --- | --- |
+| `MiniMax-M3` | 1,000,000 | 0.30 | 1.20 | 0.06 | null | Text, image, video | Adaptive, disabled |
+| `MiniMax-M2.7` | 204,800 | 0.30 | 1.20 | 0.06 | 0.375 | Text | Always on |
+
+`MiniMax-M3` has tiered pricing based on input length and service tier:
+
+| Service tier | Input length | Input | Output | Cache read |
+| --- | --- | ---: | ---: | ---: |
+| Standard | Up to 512,000 | 0.30 | 1.20 | 0.06 |
+| Standard | Above 512,000 | 0.60 | 2.40 | 0.12 |
+| Priority | Up to 512,000 | 0.45 | 1.80 | 0.09 |
+| Priority | Above 512,000 | 0.90 | 3.60 | 0.18 |
+
+The desktop provider adapter keeps MiniMax OpenAI-compatible models on the chat-completions path and does not enable the OpenAI developer-role reasoning flag for these model IDs.

README.md

@@ -225,6 +225,7 @@ Add a `SKILL.md` to any project to teach the model your own taste.
 - **Unified provider model** — Anthropic, OpenAI, Gemini, DeepSeek, OpenRouter, SiliconFlow, local Ollama, or any OpenAI-compatible relay; keyless (IP-allowlisted) proxies supported
 - **One-click import and sign-in** — bring Claude Code / Codex API-key provider configs across, or sign in with ChatGPT subscription for Codex models
 - **Dynamic model picker** — every provider exposes its real model catalogue, not a hardcoded shortlist
+- **MiniMax support** - built-in MiniMax access with regional OpenAI-compatible and Anthropic-compatible endpoint guidance ([provider guide](./MINIMAX.md))
 
 ### Generation and editing
 - **Prompt → HTML or JSX/React component** prototype, rendered in a sandboxed iframe (vendored React 18 + Babel on-device)

apps/desktop/src/main/imports/codex-config.test.ts

@@ -262,6 +262,7 @@ env_key  = "${envKey}"
     'DEEPSEEK_API_KEY',
     'XAI_API_KEY',
     'GROQ_API_KEY',
+    'MINIMAX_API_KEY',
   ])('accepts known provider env_key=%s', async (envKey) => {
     const toml = `
 [model_providers.p]

apps/desktop/src/main/imports/codex-config.ts

@@ -49,6 +49,7 @@ export const ALLOWED_IMPORT_ENV_KEYS: ReadonlySet<string> = new Set([
   'GOOGLE_API_KEY',
   'GROQ_API_KEY',
   'MISTRAL_API_KEY',
+  'MINIMAX_API_KEY',
   'OPENAI_API_KEY',
   'OPENROUTER_API_KEY',
   'PERPLEXITY_API_KEY',

apps/desktop/src/main/onboarding/provider-parsers.ts

@@ -218,7 +218,7 @@ export function parseValidateKey(raw: unknown): ValidateKeyInput {
   const providerId = provider.trim();
   if (!isSupportedOnboardingProvider(providerId)) {
     throw new CodesignError(
-      `Provider "${providerId}" is not supported. Only anthropic, openai, openrouter, ollama.`,
+      `Provider "${providerId}" is not supported. Only anthropic, openai, openrouter, minimax, ollama.`,
       ERROR_CODES.PROVIDER_NOT_SUPPORTED,
     );
   }

packages/providers/src/validate.test.ts

@@ -1,4 +1,4 @@
-import { CodesignError } from '@open-codesign/shared';
+import { CodesignError, MINIMAX_ENDPOINTS } from '@open-codesign/shared';
 import { afterEach, describe, expect, it, vi } from 'vitest';
 import { pingProvider } from './validate';
 
@@ -127,6 +127,23 @@ describe('pingProvider', () => {
     expect(result).toEqual({ ok: true, modelCount: 0 });
   });
 
+  it('uses the configured global MiniMax OpenAI endpoint and Bearer auth', async () => {
+    mockFetch(async (url, init) => {
+      expect(url).toBe(`${MINIMAX_ENDPOINTS.global_en.openaiBaseUrl}/models`);
+      const headers = (init?.headers ?? {}) as Record<string, string>;
+      expect(headers['authorization']).toBe('Bearer minimax-test-key');
+      return new Response(
+        JSON.stringify({ data: [{ id: 'MiniMax-M3' }, { id: 'MiniMax-M2.7' }] }),
+        {
+          status: 200,
+          headers: { 'content-type': 'application/json' },
+        },
+      );
+    });
+    const result = await pingProvider('minimax', 'minimax-test-key');
+    expect(result).toEqual({ ok: true, modelCount: 2 });
+  });
+
   it('respects custom baseUrl without /v1 suffix', async () => {
     mockFetch(async (url) => {
       expect(url).toBe('https://proxy.example/v1/models');

packages/providers/src/validate.ts

@@ -3,6 +3,7 @@ import {
   CodesignError,
   ERROR_CODES,
   isSupportedOnboardingProvider,
+  MINIMAX_ENDPOINTS,
   type SupportedOnboardingProvider,
   stripInferenceEndpointSuffix,
 } from '@open-codesign/shared';
@@ -57,6 +58,15 @@ function endpoint(provider: SupportedOnboardingProvider, baseUrl?: string): Prov
         headers: (apiKey) => ({ authorization: `Bearer ${apiKey}` }),
       };
     }
+    case 'minimax': {
+      const root = baseUrl
+        ? normalizeValidateBaseUrl(baseUrl)
+        : normalizeValidateBaseUrl(MINIMAX_ENDPOINTS.global_en.openaiBaseUrl);
+      return {
+        url: `${root}/v1/models`,
+        headers: (apiKey) => ({ authorization: `Bearer ${apiKey}` }),
+      };
+    }
     case 'ollama': {
       // Local Ollama — OpenAI-compat endpoint at /v1. No auth header; the
       // caller (renderer) may still pass a non-empty apiKey as a sentinel
@@ -97,7 +107,7 @@ export async function pingProvider(
 ): Promise<ValidateResult> {
   if (!isSupportedOnboardingProvider(provider)) {
     throw new CodesignError(
-      `Provider "${provider}" is not supported by the first-run provider shortcut. Supported: anthropic, openai, openrouter, ollama. Add custom providers in Settings, or use ChatGPT subscription sign-in for chatgpt-codex.`,
+      `Provider "${provider}" is not supported by the first-run provider shortcut. Supported: anthropic, openai, openrouter, minimax, ollama. Add custom providers in Settings, or use ChatGPT subscription sign-in for chatgpt-codex.`,
       ERROR_CODES.PROVIDER_NOT_SUPPORTED,
     );
   }

packages/shared/src/config.test.ts

@@ -5,6 +5,9 @@ import {
   defaultProviderCapabilities,
   detectWireFromBaseUrl,
   hydrateConfig,
+  MINIMAX_ENDPOINTS,
+  MINIMAX_MODEL_CATALOG,
+  MINIMAX_MODEL_IDS,
   migrateLegacyToV3,
   parseConfigFlexible,
   resolveProviderCapabilities,
@@ -13,6 +16,50 @@ import {
 } from './config';
 
 describe('config v3 schema', () => {
+  it('registers MiniMax models and all configured endpoint variants', () => {
+    expect(MINIMAX_MODEL_IDS).toEqual(['MiniMax-M3', 'MiniMax-M2.7']);
+    expect(BUILTIN_PROVIDERS.minimax).toMatchObject({
+      id: 'minimax',
+      wire: 'openai-chat',
+      baseUrl: MINIMAX_ENDPOINTS.global_en.openaiBaseUrl,
+      envKey: 'MINIMAX_API_KEY',
+      defaultModel: 'MiniMax-M3',
+      modelsHint: ['MiniMax-M3', 'MiniMax-M2.7'],
+    });
+    expect(
+      Object.values(MINIMAX_ENDPOINTS).flatMap((endpoint) => [
+        endpoint.openaiBaseUrl,
+        endpoint.anthropicBaseUrl,
+      ]),
+    ).toEqual([
+      'https://api.minimax.io/v1',
+      'https://api.minimax.io/anthropic',
+      'https://api.minimaxi.com/v1',
+      'https://api.minimaxi.com/anthropic',
+    ]);
+    expect(
+      Object.values(MINIMAX_ENDPOINTS).every((endpoint) =>
+        endpoint.anthropicBaseUrl.endsWith('/anthropic'),
+      ),
+    ).toBe(true);
+    expect(MINIMAX_MODEL_CATALOG).toMatchObject([
+      {
+        modelId: 'MiniMax-M3',
+        contextWindow: 1_000_000,
+        inputModalities: ['text', 'image', 'video'],
+        thinking: ['adaptive', 'disabled'],
+      },
+      {
+        modelId: 'MiniMax-M2.7',
+        contextWindow: 204_800,
+        inputModalities: ['text'],
+        thinking: ['always_on'],
+      },
+    ]);
+    expect(MINIMAX_MODEL_CATALOG[0]?.pricingTiersUsdPerMillionTokens).toHaveLength(4);
+    expect(MINIMAX_MODEL_CATALOG[1]?.pricingUsdPerMillionTokens.cacheWrite).toBe(0.375);
+  });
+
   it('parses a minimal v3 config', () => {
     const raw = {
       version: 3,
@@ -302,7 +349,7 @@ describe('config v3 schema', () => {
 });
 
 describe('migrateLegacyToV3', () => {
-  it('seeds three builtin providers from an empty v2', () => {
+  it('seeds all supported builtin providers from an empty v2', () => {
     const legacy = {
       version: 2 as const,
       provider: 'anthropic' as const,

packages/shared/src/config.ts

@@ -7,6 +7,7 @@ const ProviderIdEnum = z.enum([
   'openai',
   'google',
   'openrouter',
+  'minimax',
   'groq',
   'cerebras',
   'xai',
@@ -20,6 +21,7 @@ export const SUPPORTED_ONBOARDING_PROVIDERS = [
   'anthropic',
   'openai',
   'openrouter',
+  'minimax',
   'ollama',
 ] as const;
 export type SupportedOnboardingProvider = (typeof SUPPORTED_ONBOARDING_PROVIDERS)[number];
@@ -29,6 +31,84 @@ export type SupportedOnboardingProvider = (typeof SUPPORTED_ONBOARDING_PROVIDERS
 export const OLLAMA_DEFAULT_BASE_URL = 'http://localhost:11434/v1';
 export const OLLAMA_DEFAULT_MODEL = 'llama3.2';
 
+export const MINIMAX_ENDPOINTS = {
+  global_en: {
+    region: 'global_en',
+    openaiBaseUrl: 'https://api.minimax.io/v1',
+    anthropicBaseUrl: 'https://api.minimax.io/anthropic',
+    docsRoot: 'https://platform.minimax.io/docs',
+  },
+  cn_zh: {
+    region: 'cn_zh',
+    openaiBaseUrl: 'https://api.minimaxi.com/v1',
+    anthropicBaseUrl: 'https://api.minimaxi.com/anthropic',
+    docsRoot: 'https://platform.minimaxi.com/docs',
+  },
+} as const;
+
+export const MINIMAX_MODEL_CATALOG = [
+  {
+    modelId: 'MiniMax-M3',
+    contextWindow: 1_000_000,
+    pricingUsdPerMillionTokens: {
+      input: 0.3,
+      output: 1.2,
+      cacheRead: 0.06,
+      cacheWrite: null,
+    },
+    pricingTiersUsdPerMillionTokens: [
+      {
+        serviceTier: 'standard',
+        inputTokensLte: 512_000,
+        input: 0.3,
+        output: 1.2,
+        cacheRead: 0.06,
+        cacheWrite: null,
+      },
+      {
+        serviceTier: 'standard',
+        inputTokensGt: 512_000,
+        input: 0.6,
+        output: 2.4,
+        cacheRead: 0.12,
+        cacheWrite: null,
+      },
+      {
+        serviceTier: 'priority',
+        inputTokensLte: 512_000,
+        input: 0.45,
+        output: 1.8,
+        cacheRead: 0.09,
+        cacheWrite: null,
+      },
+      {
+        serviceTier: 'priority',
+        inputTokensGt: 512_000,
+        input: 0.9,
+        output: 3.6,
+        cacheRead: 0.18,
+        cacheWrite: null,
+      },
+    ],
+    inputModalities: ['text', 'image', 'video'],
+    thinking: ['adaptive', 'disabled'],
+  },
+  {
+    modelId: 'MiniMax-M2.7',
+    contextWindow: 204_800,
+    pricingUsdPerMillionTokens: {
+      input: 0.3,
+      output: 1.2,
+      cacheRead: 0.06,
+      cacheWrite: 0.375,
+    },
+    inputModalities: ['text'],
+    thinking: ['always_on'],
+  },
+] as const;
+
+export const MINIMAX_MODEL_IDS = MINIMAX_MODEL_CATALOG.map((model) => model.modelId);
+
 // ── Wire types (v3) ──────────────────────────────────────────────────────────
 
 export const WireApiSchema = z.enum([
@@ -278,6 +358,23 @@ export const BUILTIN_PROVIDERS: Readonly<Record<SupportedOnboardingProvider, Pro
       modelDiscoveryMode: 'models',
     },
   },
+  minimax: {
+    id: 'minimax',
+    name: 'MiniMax',
+    builtin: true,
+    wire: 'openai-chat',
+    baseUrl: MINIMAX_ENDPOINTS.global_en.openaiBaseUrl,
+    envKey: 'MINIMAX_API_KEY',
+    defaultModel: MINIMAX_MODEL_CATALOG[0].modelId,
+    modelsHint: MINIMAX_MODEL_IDS,
+    capabilities: {
+      supportsKeyless: false,
+      supportsModelsEndpoint: false,
+      supportsReasoning: false,
+      requiresClaudeCodeIdentity: false,
+      modelDiscoveryMode: 'static-hint',
+    },
+  },
   ollama: {
     id: 'ollama',
     name: 'Ollama (local)',
@@ -384,7 +481,7 @@ function cloneBuiltin(id: SupportedOnboardingProvider): ProviderEntry {
 }
 
 /**
- * Pure: migrate a validated v1/v2 config to v3. Seeds the three builtin
+ * Pure: migrate a validated v1/v2 config to v3. Seeds the supported builtin
  * providers and overlays any stored baseUrls onto them.
  */
 export function migrateLegacyToV3(legacy: LegacyConfig): ConfigV3 {
@@ -507,6 +604,13 @@ export const PROVIDER_SHORTLIST: Record<SupportedOnboardingProvider, ProviderSho
     primary: ['anthropic/claude-sonnet-4.6', 'openai/gpt-4o'],
     defaultPrimary: 'anthropic/claude-sonnet-4.6',
   },
+  minimax: {
+    provider: 'minimax',
+    label: 'MiniMax',
+    keyHelpUrl: 'https://platform.minimax.io/docs/api-reference/api-overview',
+    primary: MINIMAX_MODEL_IDS,
+    defaultPrimary: MINIMAX_MODEL_CATALOG[0].modelId,
+  },
   ollama: {
     provider: 'ollama',
     label: 'Ollama (local)',

packages/shared/src/index.ts

@@ -5,6 +5,7 @@ export const ProviderId = z.enum([
   'openai',
   'google',
   'openrouter',
+  'minimax',
   'groq',
   'cerebras',
   'xai',
@@ -319,6 +320,9 @@ export {
   ImageGenerationSettingsSchema,
   ImageGenerationSizeSchema,
   isSupportedOnboardingProvider,
+  MINIMAX_ENDPOINTS,
+  MINIMAX_MODEL_CATALOG,
+  MINIMAX_MODEL_IDS,
   migrateLegacyToV3,
   PROVIDER_SHORTLIST,
   ProviderCapabilitiesSchema,