feat: add direct MiniMax provider support

- Add MiniMax LLM handler with direct API integration (api.minimax.io/v1)
- Support models: MiniMax-M2.5, MiniMax-M2.5-highspeed
- Add MINIMAX_API_KEY environment variable support
- Add 'minimax' to allowed model prefixes
- Add MiniMax models to openrouter model config
- Add streaming and non-streaming request handling
- Add MiniMax per-token pricing for accurate billing
- Handle MiniMax temperature constraint (0.0, 1.0], default 1.0
- Route minimax/* models directly to MiniMax API before Fireworks fallback

Author: PR Bot

.env.example

@@ -6,6 +6,7 @@ ANTHROPIC_API_KEY=dummy_anthropic_key
 FIREWORKS_API_KEY=dummy_fireworks_key
 CANOPYWAVE_API_KEY=dummy_canopywave_key
 SILICONFLOW_API_KEY=dummy_siliconflow_key
+MINIMAX_API_KEY=dummy_minimax_key
 
 # Database & Server
 DATABASE_URL=postgresql://manicode_user_local:secretpassword_local@localhost:5432/manicode_db_local

common/src/constants/model-config.ts

@@ -6,6 +6,7 @@ export const ALLOWED_MODEL_PREFIXES = [
   'openai',
   'google',
   'x-ai',
+  'minimax',
 ] as const
 
 export const costModes = [
@@ -47,6 +48,8 @@ export const openrouterModels = {
   openrouter_gemini2_5_flash_thinking:
     'google/gemini-2.5-flash-preview:thinking',
   openrouter_grok_4: 'x-ai/grok-4-07-09',
+  openrouter_minimax_m2_5: 'minimax/minimax-m2.5',
+  openrouter_minimax_m2_5_highspeed: 'minimax/minimax-m2.5-highspeed',
 } as const
 export type openrouterModel =
   (typeof openrouterModels)[keyof typeof openrouterModels]
@@ -171,6 +174,7 @@ export const providerDomains = {
   openai: 'chatgpt.com',
   deepseek: 'deepseek.com',
   xai: 'x.ai',
+  minimax: 'minimax.io',
 } as const
 
 export function getLogoForModel(modelName: string): string | undefined {
@@ -182,6 +186,7 @@ export function getLogoForModel(modelName: string): string | undefined {
     domain = providerDomains.deepseek
   else if (modelName.includes('claude')) domain = providerDomains.anthropic
   else if (modelName.includes('grok')) domain = providerDomains.xai
+  else if (modelName.includes('minimax')) domain = providerDomains.minimax
 
   return domain
     ? `https://www.google.com/s2/favicons?domain=${domain}&sz=256`

packages/internal/src/env-schema.ts

@@ -9,6 +9,7 @@ export const serverEnvSchema = clientEnvSchema.extend({
   FIREWORKS_API_KEY: z.string().min(1),
   CANOPYWAVE_API_KEY: z.string().min(1).optional(),
   SILICONFLOW_API_KEY: z.string().min(1).optional(),
+  MINIMAX_API_KEY: z.string().min(1).optional(),
   LINKUP_API_KEY: z.string().min(1),
   CONTEXT7_API_KEY: z.string().optional(),
   GRAVITY_API_KEY: z.string().min(1),
@@ -54,6 +55,7 @@ export const serverProcessEnv: ServerInput = {
   FIREWORKS_API_KEY: process.env.FIREWORKS_API_KEY,
   CANOPYWAVE_API_KEY: process.env.CANOPYWAVE_API_KEY,
   SILICONFLOW_API_KEY: process.env.SILICONFLOW_API_KEY,
+  MINIMAX_API_KEY: process.env.MINIMAX_API_KEY,
   LINKUP_API_KEY: process.env.LINKUP_API_KEY,
   CONTEXT7_API_KEY: process.env.CONTEXT7_API_KEY,
   GRAVITY_API_KEY: process.env.GRAVITY_API_KEY,

web/src/app/api/v1/chat/completions/_post.ts

@@ -53,6 +53,12 @@ import {
   handleSiliconFlowStream,
   isSiliconFlowModel,
 } from '@/llm-api/siliconflow'
+import {
+  MiniMaxError,
+  handleMiniMaxNonStream,
+  handleMiniMaxStream,
+  isMiniMaxModel,
+} from '@/llm-api/minimax'
 import {
   handleOpenAINonStream,
   OPENAI_SUPPORTED_MODELS,
@@ -366,11 +372,22 @@ export async function postChatCompletions(params: {
     // Handle streaming vs non-streaming
     try {
       if (bodyStream) {
-        // Streaming request — route to SiliconFlow/CanopyWave/Fireworks for supported models
+        // Streaming request — route to MiniMax/SiliconFlow/CanopyWave/Fireworks for supported models
+        const useMiniMax = isMiniMaxModel(typedBody.model)
         const useSiliconFlow = false // isSiliconFlowModel(typedBody.model)
         const useCanopyWave = false // isCanopyWaveModel(typedBody.model)
-        const useFireworks = isFireworksModel(typedBody.model)
-        const stream = useSiliconFlow
+        const useFireworks = !useMiniMax && isFireworksModel(typedBody.model)
+        const stream = useMiniMax
+          ? await handleMiniMaxStream({
+              body: typedBody,
+              userId,
+              stripeCustomerId,
+              agentId,
+              fetch,
+              logger,
+              insertMessageBigquery,
+            })
+          : useSiliconFlow
           ? await handleSiliconFlowStream({
               body: typedBody,
               userId,
@@ -430,12 +447,13 @@ export async function postChatCompletions(params: {
           },
         })
       } else {
-        // Non-streaming request — route to SiliconFlow/CanopyWave/Fireworks for supported models
+        // Non-streaming request — route to MiniMax/SiliconFlow/CanopyWave/Fireworks for supported models
         // TEMPORARILY DISABLED: route through OpenRouter
         const model = typedBody.model
+        const useMiniMaxDirect = isMiniMaxModel(model)
         const useSiliconFlow = false // isSiliconFlowModel(model)
         const useCanopyWave = false // isCanopyWaveModel(model)
-        const useFireworks = isFireworksModel(model)
+        const useFireworks = !useMiniMaxDirect && isFireworksModel(model)
         const modelParts = model.split('/')
         const shortModelName = modelParts.length > 1 ? modelParts[1] : model
         const isOpenAIDirectModel =
@@ -446,7 +464,17 @@ export async function postChatCompletions(params: {
         const shouldUseOpenAIEndpoint =
           isOpenAIDirectModel && typedBody.codebuff_metadata?.n !== undefined
 
-        const nonStreamRequest = useSiliconFlow
+        const nonStreamRequest = useMiniMaxDirect
+          ? handleMiniMaxNonStream({
+              body: typedBody,
+              userId,
+              stripeCustomerId,
+              agentId,
+              fetch,
+              logger,
+              insertMessageBigquery,
+            })
+          : useSiliconFlow
           ? handleSiliconFlowNonStream({
               body: typedBody,
               userId,
@@ -528,10 +556,14 @@ export async function postChatCompletions(params: {
       if (error instanceof SiliconFlowError) {
         siliconflowError = error
       }
+      let minimaxError: MiniMaxError | undefined
+      if (error instanceof MiniMaxError) {
+        minimaxError = error
+      }
 
       // Log detailed error information for debugging
       const errorDetails = openrouterError?.toJSON()
-      const providerLabel = siliconflowError ? 'SiliconFlow' : canopywaveError ? 'CanopyWave' : fireworksError ? 'Fireworks' : 'OpenRouter'
+      const providerLabel = minimaxError ? 'MiniMax' : siliconflowError ? 'SiliconFlow' : canopywaveError ? 'CanopyWave' : fireworksError ? 'Fireworks' : 'OpenRouter'
       logger.error(
         {
           error: getErrorObject(error),
@@ -545,8 +577,8 @@ export async function postChatCompletions(params: {
             ? typedBody.messages.length
             : 0,
           messages: typedBody.messages,
-          providerStatusCode: (openrouterError ?? fireworksError ?? canopywaveError ?? siliconflowError)?.statusCode,
-          providerStatusText: (openrouterError ?? fireworksError ?? canopywaveError ?? siliconflowError)?.statusText,
+          providerStatusCode: (openrouterError ?? fireworksError ?? canopywaveError ?? siliconflowError ?? minimaxError)?.statusCode,
+          providerStatusText: (openrouterError ?? fireworksError ?? canopywaveError ?? siliconflowError ?? minimaxError)?.statusText,
           openrouterErrorCode: errorDetails?.error?.code,
           openrouterErrorType: errorDetails?.error?.type,
           openrouterErrorMessage: errorDetails?.error?.message,
@@ -580,6 +612,9 @@ export async function postChatCompletions(params: {
       if (error instanceof SiliconFlowError) {
         return NextResponse.json(error.toJSON(), { status: error.statusCode })
       }
+      if (error instanceof MiniMaxError) {
+        return NextResponse.json(error.toJSON(), { status: error.statusCode })
+      }
 
       return NextResponse.json(
         { error: 'Failed to process request' },