Everything you need to build with MiniMax — from your first API call to production-scale deployment patterns with best practices and migration guides.
MiniMax documentation is organized into six sections — each designed for a specific stage of the developer journey from onboarding to production optimization.
The documentation lives at docs.minimax.gr.com with full-text search, a hierarchical sidebar, and in-page tables of contents. Every page follows a consistent template: a summary at the top, progressive disclosure from basic to advanced, code snippets in all supported languages, and a "Next Steps" section at the bottom linking to related pages. The sidebar reflects the documentation taxonomy — Getting Started, API Reference, Concept Guides, Best Practices, SDKs, and Migration — so you always know where you are in the knowledge graph.
Documentation source files live in the minimax-docs GitHub repository under a Creative Commons license. The site builds from Markdown using a static site generator with automated deployment on every merge to main. This means community contributions go live within minutes of merging a pull request, and MiniMax engineers can update documentation alongside code changes in the same workflow.
Start with Getting Started for your first API call, then jump to the API Reference for endpoint details. Concept Guides explain the model architecture. Best Practices cover production deployment patterns. SDK docs live in each language-specific section.
The Getting Started section walks you from account creation through your first API call in under 15 minutes, with language-specific quickstart paths.
New developers begin with the Quickstart Guide, which covers account creation, API key generation, and a curl-based first request. From there, language-specific quickstarts — Python, JavaScript, Go — show idiomatic integration with the official SDKs. Each quickstart produces a working mini-application: a chat bot that responds to terminal input, a semantic search script that indexes local files, or a video generator that accepts text prompts. The quickstarts link to deeper concept pages for each feature so you control how far down the rabbit hole you go.
The section also includes a "Choose Your Integration Path" decision tree for teams deciding between REST API direct calls, SDK usage, or CLI-based workflows. A glossary defines key terms — tokens, embeddings, temperature, top_p, RAG — with links to the relevant concept guides for deeper exploration.
The API Reference is the canonical specification for every MiniMax endpoint — request parameters, response shapes, error codes, and rate limits in one searchable resource.
Each endpoint page follows an identical template: endpoint URL and method at the top, a description paragraph, a request body table with every parameter (name, type, required, default, description), a response body table with every field, an example request and response in curl and each SDK language, and a collapsible section for error codes specific to that endpoint. The reference auto-generates from the API's OpenAPI specification, so it is always synchronized with the live API. A "Try It" button on each endpoint page opens the API playground with the endpoint pre-selected and authentication injected.
Rate limit information appears in a sidebar widget that adapts to your account tier — if you are logged in, the reference displays your current limits. Paginated endpoints include pagination parameter details and example loops in each SDK language. The reference also documents streaming SSE event formats and webhook payload structures.
Concept guides explain the principles behind MiniMax features — how models work, how to think about prompt engineering, and how to choose between architecture patterns.
Key concept guides cover: an introduction to large language models for developers new to AI, prompt engineering strategies with MiniMax-specific best practices, embeddings and vector search fundamentals, video generation parameters and trade-offs, function calling design patterns, and model selection — when to use miniMax-chat-v2 versus miniMax-reasoning versus miniMax-vision. Each guide includes diagrams, code examples, and a "Common Pitfalls" section drawn from real support tickets. The concept guides are versioned alongside the platform: when a new model capability ships, its concept guide ships simultaneously.
Advanced guides cover fine-tuning workflows, custom model evaluation, RAG pipeline architecture with MiniMax embeddings and popular vector databases, and multi-modal application patterns combining chat, vision, and video generation in a single user experience.
The Best Practices section distills production experience into actionable patterns for reliability, cost optimization, and security.
Topics include: implementing robust retry logic with exponential backoff and jitter, caching embeddings and chat responses to reduce API costs, streaming versus non-streaming trade-offs for different user experiences, prompt templating strategies for consistent outputs, testing AI-powered features with deterministic evaluation frameworks, monitoring API usage and setting budget alerts, and securing API keys across development, staging, and production environments. Each best practice article includes a "When to Apply" section that helps you decide whether the practice applies to your use case, and a "Gotchas" section documenting edge cases discovered by the MiniMax community.
Security best practices cover API key rotation automation, least-privilege key scoping, audit logging for sensitive operations, and data handling for regulated industries. A dedicated article addresses GDPR and data residency configuration using MiniMax regional endpoints.
Every MiniMax platform change comes with a migration guide showing before-and-after code — no guessing what broke between versions.
Migration guides follow a consistent structure: version comparison table, list of breaking changes with rationale, step-by-step upgrade instructions, before-and-after code snippets in all three SDK languages, and a checklist for validating the migration. Deprecated features remain functional for six months with deprecation warnings emitted in API response headers and SDK logging at the WARNING level. The changelog page aggregates every platform release — API changes, model updates, SDK releases, and infrastructure improvements — in reverse chronological order with filtering by category.
The changelog RSS feed and a developer mailing list provide push notifications for updates. Critical changes — security patches, breaking API changes, model deprecations — also appear as banner notifications in the platform hub.
MiniMax documentation search indexes every word on every page — including code snippets, API parameter names, and error messages — with instant results as you type.
Press / or Ctrl+K to focus the search bar from any documentation page. Search supports: exact phrase matching with double quotes, endpoint path matching (type /v1/chat to jump directly to the chat completions reference), and language filtering (append lang:python to see only Python examples). The search index includes not just documentation pages but also GitHub Issues, changelog entries, and community discussion threads — all surfaced through a unified results page. Use the sidebar filter to narrow results to a specific documentation section.
This table maps each documentation section to its contents and target audience so you can navigate directly to what you need.
| Section | Contents | Audience |
|---|---|---|
| Getting Started | Quickstart guides, account setup, first API call, glossary | New developers, evaluators |
| API Reference | Endpoint specs, parameters, response shapes, error codes, rate limits | All developers integrating MiniMax |
| Concept Guides | LLM fundamentals, prompt engineering, embeddings, video gen, model selection | Developers designing AI features |
| Best Practices | Retry logic, caching, streaming, prompt templating, monitoring, security | Production engineers, SREs |
| SDKs & Tools | Python/JS/Go SDK guides, CLI reference, VS Code extension docs | SDK users, tooling adopters |
| Migration & Changelog | Version migration guides, breaking change notices, full changelog | Maintainers upgrading integrations |
| Community | GitHub discussions, issue tracker guide, contribution docs, code of conduct | Open source contributors |
MiniMax provides comprehensive developer documentation organized into six sections. The Getting Started section offers language-specific quickstart guides that produce working applications in under 15 minutes. The API Reference contains the canonical specification for every endpoint with request parameters, response fields, and interactive "Try It" buttons. Concept Guides explain model architecture, prompt engineering, embeddings, and video generation principles. Best Practices cover production deployment patterns including retry logic, caching, security, and cost optimization. SDK documentation provides complete API surfaces for Python, JavaScript, and Go clients. Migration Guides include before-and-after code for every breaking change. All documentation is hosted at docs.minimax.gr.com with full-text search, hierarchical navigation, and community-contributable source files on GitHub.
The documentation site provides multiple navigation paths. The full-text search bar — accessible from any page with / or Ctrl+K — indexes all content including code snippets, API parameter names, error messages, and community discussion threads. The left sidebar offers hierarchical topic navigation with collapsible sections. In-page tables of contents appear on the right for longer articles. Search supports exact phrase matching, direct endpoint path lookup (type /v1/chat), and language filtering (append lang:python). The API Reference includes a filtering toolbar for endpoint category, SDK language, and model version. If you still cannot find what you need, the community GitHub Discussions forum lets you ask questions that feed back into documentation improvements.
Documentation updates are continuous. API reference pages regenerate automatically from the OpenAPI specification whenever new endpoints or parameters deploy — usually within minutes of a platform release. Concept guides and best-practice articles follow a scheduled review cycle: quarterly editorial reviews ensure accuracy and freshness, while ad-hoc updates address community feedback and support trends. The changelog page updates with every platform release, typically weekly for patches and monthly for minor releases. SDK documentation updates alongside SDK releases, which follow the same cadence. All documentation source files are publicly accessible in the minimax-docs GitHub repository, where you can track update history through commit logs and watch the repository for notifications.
Every MiniMax API version change ships with a comprehensive migration guide. Each guide opens with a version comparison table and a summary of breaking changes along with the technical rationale for each change. Step-by-step upgrade instructions walk through the migration process. Before-and-after code snippets appear in Python, JavaScript, and Go — not pseudo-code but runnable examples you can copy into your project. A validation checklist helps you confirm the migration is complete. Deprecated features remain functional during a six-month window, and deprecation warnings appear in API response headers and SDK logging. Migration guides are linked from the changelog, the API reference pages for affected endpoints, and the banner notification system in the platform hub.
Documentation contributions are welcome through the standard fork-and-pull-request workflow on the minimax-docs GitHub repository. The contribution guide covers the writing style (active voice, present tense, direct address), Markdown conventions (headings, code blocks, callouts, tables), and code snippet formatting (language tags, environment variable placeholders, SDK-specific tabs). Pull requests can fix errors, clarify ambiguous explanations, add code examples in any supported language, or translate content. Documentation PRs receive review within two business days. The repository uses standard CI checks — markdownlint for style consistency, link checking for broken URLs, and a preview deployment that builds a staging version of the docs for visual review. All documentation is licensed under Creative Commons Attribution 4.0.