Skip to content

CLAUDE.md

Latest commit

 

History

History
111 lines (81 loc) · 4.62 KB

CLAUDE.md

File metadata and controls

111 lines (81 loc) · 4.62 KB

CLAUDE.md -- MiniMax SDK

Overview

Auto-generated C# SDK for the MiniMax Platform -- covers Hailuo video generation, music generation, text-to-speech (T2A v2), voice cloning, and file management.

Scope:

  • Video -- Hailuo 2.3 / 2.3-Fast / 02 text-to-video, image-to-video, start-end, subject reference (async task pattern: POST + poll).
  • Music -- music-2.6 / music-cover / music-2.0 generation.
  • Speech -- T2A v2 text-to-speech (speech-2.8-hd, speech-2.8-turbo, speech-2.6-*, speech-02-*, speech-01-*) + voice clone.
  • Files -- upload/retrieve/list/delete (used as inputs for voice clone, music cover, video start frames).

Excluded (use another SDK):

  • LLM chat completions -- use tryAGI.OpenAI.CustomProviders.Minimax(...) for the OpenAI-compatible chat endpoint.
  • Embeddings -- MiniMax embeddings are chat-adjacent; use an LLM-focused SDK if needed.
  • WebSocket TTS streaming -- not yet modelled; REST T2A v2 stream=true still works.

No public OpenAPI spec exists -- openapi.yaml is handcrafted from platform.minimax.io/docs.

Build & Test

dotnet build MiniMax.slnx
dotnet test src/tests/IntegrationTests/

Tests skip (AssertInconclusiveException) when MINIMAX_API_KEY is missing.

Auth

Standard Bearer token. Get an API key from https://platform.minimax.io/user-center/basic-information/interface-key.

using var client = new MiniMaxClient(apiKey); // MINIMAX_API_KEY env var

Base URL

  • Default: https://api.minimax.io (global)
  • China mainland: https://api.minimaxi.chat

Switch via client.TrySelectServer("https-api-minimaxi-chat") or pass baseUri to the constructor.

Sub-Client Layout

Sub-client Endpoints Purpose
client.Video POST /v1/video_generation, GET /v1/query/video_generation Hailuo video tasks
client.Music POST /v1/music_generation Music generation
client.Speech POST /v1/t2a_v2, POST /v1/voice_clone TTS + voice cloning
client.Files POST/GET /v1/files/{upload,retrieve,list,delete} File management

Async Video Task Pattern

// 1. Submit
var create = await client.Video.CreateVideoGenerationTaskAsync(
    model: "MiniMax-Hailuo-2.3",
    prompt: "A drone shot of mountains at sunrise",
    duration: 6,
    resolution: VideoGenerationRequestResolution.x1080p);

// 2. Poll (Preparing -> Queueing -> Processing -> Success)
while (true)
{
    var status = await client.Video.GetVideoGenerationTaskAsync(create.TaskId!);
    if (status.Status is VideoTaskStatus.Success or VideoTaskStatus.Fail)
        break;
    await Task.Delay(TimeSpan.FromSeconds(10));
}

// 3. Retrieve file URL
var file = await client.Files.RetrieveFileAsync(status.FileId!);
// file.File.DownloadUrl

MEAI Integration

AIFunction tools for use with any IChatClient (via MiniMaxToolExtensions):

  • AsCreateVideoTaskTool() -- submit a Hailuo video task
  • AsGetVideoTaskTool() -- poll a video task by ID
  • AsGenerateMusicTool() -- generate music from prompt/lyrics
  • AsSynthesizeSpeechTool() -- T2A v2 speech synthesis
  • AsCloneVoiceTool() -- create a voice clone from an uploaded file

No dedicated IChatClient / IEmbeddingGenerator / ISpeechToTextClient implementation -- T2A v2 does synthesis only, not transcription. Chat goes through tryAGI.OpenAI.CustomProviders.Minimax(...).

Key Files

  • src/libs/MiniMax/openapi.yaml -- Handcrafted OpenAPI spec
  • src/libs/MiniMax/generate.sh -- Regen script (runs autosdk on local spec)
  • src/libs/MiniMax/Generated/ -- Never edit -- auto-generated code
  • src/libs/MiniMax/Extensions/MiniMaxClient.Tools.cs -- MEAI AIFunction tools
  • src/tests/IntegrationTests/Examples/ -- Example tests (also generate docs)

Notes & Gotchas

  • Enum naming: VideoGenerationRequestResolution.x1080p / x768p / x512p (AutoSDK prepends x when enum values start with digits).
  • Resolution per model: Hailuo-2.3 supports 512P/768P/1080P, Hailuo-2.3-Fast is 768P only, Hailuo-02 supports 768P/1080P -- check the platform docs for the exact matrix before each release.
  • Audio output: default is hex (inline hex-encoded bytes). Pass outputFormat: *OutputFormat.Url for a downloadable link.
  • Voice clone prerequisite: upload the reference audio (MP3/M4A/WAV, 10s-5min) via client.Files.UploadFileAsync(purpose: "voice_clone", ...) first, then pass the returned file_id to CreateVoiceCloneAsync.
  • CLS compliance: extension methods returning AIFunction are marked [CLSCompliant(false)] because Microsoft.Extensions.AI.AIFunction is not CLS-compliant.