MiniMax Integration Tools & SDKs

Available SDKs

MiniMax ships three official SDKs — Python, JavaScript, Go — each wrapping the REST API in idiomatic language interfaces with streaming, retry, and type safety built in.

Python SDK (minimax-sdk): Requires Python 3.8+. Install with pip install minimax-sdk. Provides synchronous and asynchronous clients via MiniMax and AsyncMiniMax classes. Handles automatic retry with exponential backoff, streaming chat completions, and file uploads for fine-tuning and video reference. Full type hints throughout. The async client integrates with asyncio, FastAPI, and any async Python framework. Test suite includes mocked HTTP transport for offline testing and integration tests against the live API.

JavaScript SDK (@minimax/sdk): Targets Node.js 18+ with full TypeScript declarations. Install via npm install @minimax/sdk. Ships ESM and CommonJS bundles. The fetch-based HTTP client supports streaming via async iterators. A pluggable middleware system lets you inject custom logging, retry policies, and rate limit handlers. Browser-compatible builds are included for frontend use behind an API proxy. The package size is under 50KB gzipped with zero polyfill dependencies.

Go SDK (minimax-go): Requires Go 1.21+. Import as github.com/minimax/minimax-go. Idiomatic Go patterns: context-aware cancellation, structured error types with errors.As support, and zero dependencies beyond the standard library. Automatic pagination for list endpoints. Request and response structs mirror the API reference exactly. The client is safe for concurrent use across goroutines.

CLI Tool

The MiniMax CLI puts the entire API in your terminal — chat with models, generate embeddings, submit video tasks, and manage keys without opening a code editor.

Install the CLI via Homebrew: brew install minimax/tap/minimax-cli, or download prebuilt binaries for macOS, Linux, and Windows from GitHub Releases. First-run setup prompts for your API key (stored encrypted in ~/.minimax/config.yaml). Key commands include minimax chat for interactive conversations with any MiniMax model, minimax embed "your text" for vector generation, minimax video submit --prompt "..." for video tasks with status polling, and minimax models list to browse available models with pricing. The CLI supports JSON output mode for shell scripting and CI/CD pipelines.

Advanced features include: piping stdin directly into chat prompts (cat meeting_notes.txt | minimax chat --system "Summarize this"), batch embedding with wildcard file patterns, and a plugin system for custom subcommands. The CLI auto-updates on launch and respects the MINIMAX_API_KEY environment variable for containerized environments.

VS Code Extension

The MiniMax VS Code extension brings AI assistance into your editor — inline chat, code completion, and prompt management without leaving your workspace.

Install from the VS Code Marketplace (search "MiniMax") or the Open VSX Registry. After authenticating with your API key, the extension provides a side-panel chat interface for generating and refining code. The inline completion feature suggests function bodies, docstrings, and test cases based on context. A prompt snippet library ships with common templates for code generation, refactoring, and explanation tasks — and you can add your own. The extension respects .gitignore and sends only the active file context to the API by default. Language support includes Python, JavaScript, TypeScript, Go, Rust, and Java.

Configuration options let you set the default model, control context window size, enable or disable telemetry, and map custom keyboard shortcuts to extension commands. The extension is open source on GitHub under the MIT license.

LangChain Integration

The langchain-minimax package exposes MiniMax chat models, embeddings, and video generation as native LangChain components.

Install with pip install langchain-minimax. The package provides ChatMiniMax (implements LangChain's BaseChatModel), MiniMaxEmbeddings (implements Embeddings base class), and MiniMaxVideoGen (custom chain for video generation workflows). These components work with LangChain's standard patterns: prompt templates, chain composition, agent tool use, and memory management. The embeddings integration supports all LangChain vector stores — Pinecone, Weaviate, Chroma, Milvus, FAISS — enabling RAG pipelines with MiniMax embeddings as the retrieval backbone.

The package is community-maintained with MiniMax engineering support. It tracks LangChain releases within one week and supports both the stable and experimental LangChain APIs. Documentation includes a full RAG tutorial that builds a document Q&A system using MiniMax embeddings, Pinecone, and MiniMax chat in under 100 lines of Python.

Vector Database Connectors

MiniMax embeddings integrate with every major vector database through native connectors or LangChain abstraction — choose based on your infrastructure preferences.

Pinecone: The MiniMax Python SDK includes a MiniMaxPinecone helper that handles embedding generation and upsertion in a single call. Configure with your Pinecone API key and index name. Weaviate: A Weaviate module (minimax-weaviate) registers MiniMax as a vectorizer in your Weaviate schema. Milvus & Chroma: Use via the LangChain integration — the embeddings class plugs directly into Milvus and Chroma vector stores. Pgvector: A PostgreSQL extension recipe in the MiniMax examples repository shows how to store and query MiniMax embeddings in Postgres.

Webhook Support

MiniMax webhooks deliver asynchronous operation results — video generation completions, fine-tuning job status — directly to your server via signed HTTP callbacks.

Configure webhook URLs in the platform hub under Settings > Webhooks. Each webhook receives POST requests with a JSON payload containing the task ID, event type (video.completed, fine_tune.completed, fine_tune.failed), and a signed result URL for downloading outputs. Webhook payloads are signed with HMAC-SHA256 using a shared secret you configure — verify the X-MiniMax-Signature header on every incoming request to prevent spoofing. Failed deliveries retry with exponential backoff (5s, 25s, 125s, 625s, 3125s) before marking the webhook endpoint as degraded.

A webhook testing tool in the platform hub lets you send simulated events to verify your endpoint configuration before production deployment. Webhook delivery logs are available in the dashboard for the last 30 days, including request and response bodies for debugging.

Integration Matrix

This table catalogs every MiniMax integration tool with its language, maturity level, and key features to help you pick the right tool for your stack.

Tool	Language	Maturity	Key Features
minimax-sdk (Python)	Python 3.8+	Stable	Sync/async clients, streaming, retry, file upload, type hints
@minimax/sdk (JS)	TypeScript/Node.js 18+	Stable	ESM/CJS, fetch-based, middleware, browser build, async iterators
minimax-go	Go 1.21+	Stable	Context support, zero deps, structured errors, pagination
minimax-cli	Go (binary)	Stable	Chat, embeddings, video, key mgmt, JSON output, plugins
VS Code Extension	TypeScript	Beta	Inline chat, code completion, prompt library, key management
langchain-minimax	Python	Beta	Chat, embeddings, video gen chains, RAG pipeline support
minimax-weaviate	Python	Beta	Weaviate vectorizer module, schema integration
Webhooks	HTTP/JSON	Stable	Signed payloads, retry, testing tool, 30-day delivery logs