Alibaba Cloud Model Studio node added

DawidMyslak

·Apr 9, 2026·#27928feat(Alibaba Cloud Model Studio Node): Add new node

n8n now includes native support for Alibaba Cloud's Model Studio, letting workflows chat with Qwen models, generate images, and create videos.

n8n workflows can now interact directly with Alibaba Cloud's Model Studio API. A new vendor node brings three operation types: text chat completions with Qwen models, image analysis and generation, and video creation from text or existing images. The text operation supports tool and function calling, allowing Qwen models to integrate with n8n's AI Agent system. Generated images download automatically as binary data, while video generation uses asynchronous polling with automatic retry logic. The node follows the established vendor pattern in the langchain package, using the shared AlibabaCloudApi credential with regional endpoints including Frankfurt workspace support. Model lists are currently static, though the architecture supports dynamic loading via the OpenAI-compatible models endpoint in a future update.

View Original GitHub Description

Summary

Add the Alibaba Cloud Model Studio standalone node, based on community PR #27814 by @dmolenaars, integrated into core with review feedback applied.

Loom recording

Features

Text: Chat completions with Qwen models (Qwen3 Max, Qwen3.5 Plus/Flash, MoE variants). Supports multi-turn conversations, web search, and configurable parameters. Includes tool connector for AI Agent integration.
Image: Vision-language analysis (Qwen-VL), text-to-image generation with auto-download (Z-Image Turbo, Wan 2.6 T2I, Qwen Image).
Video: Text-to-video and image-to-video generation (Wan 2.6 T2V/I2V) with async polling, plus video download.

Architecture notes

Model lists are currently static, matching the original community PR. The DashScope native API does not expose a unified /models endpoint filtered by capability. However, the OpenAI-compatible API (/compatible-mode/v1/models) is available and can be used in a follow-up to switch to dynamic model loading — this is fully backward compatible since the parameter name (modelId) and type (options) won't change.
Telemetry: The text model parameter uses modelId (matching OpenAI/Gemini/Anthropic convention) and the node is registered in AI_VENDOR_NODE_TYPES for analytics tracking.

Changes from community PR

Relocated node from nodes-base to nodes-langchain/nodes/vendors/ (matching OpenAI, Gemini, Anthropic pattern)
Unified credential to shared AlibabaCloudApi in nodes-langchain/credentials/ (matching PR #27882's pattern) with region selector and workspace ID for Frankfurt
Fixed codex identifier, removed unused requestDefaults and routing blocks
Renamed snake_case parameter names to camelCase per n8n conventions
Replaced throw new Error() with NodeOperationError
Removed extends IDataObject from request body interface
Set usableAsTool: true, human-readable model display names
Extracted versionDescription.ts following vendor node pattern
Added dynamic inputs for Tools connector on text/message operation
Auto-download generated images as binary data (matching Gemini behavior)
Removed redundant Download Image operation (generate now handles it)
Renamed text model parameter to modelId for telemetry compatibility
Added Frankfurt workspace ID validation with UserError
Unified simplifyOutput and pairedItem formats across all operations
Added node type constant to AI_VENDOR_NODE_TYPES for telemetry

Related Linear tickets, Github issues, and Community forum posts

https://linear.app/n8n/issue/NODE-4759

Based on community contribution: https://github.com/n8n-io/n8n/pull/27814

Review / Merge checklist

PR title and summary are descriptive. (conventions)
Docs updated or follow-up ticket created.
Tests included.
PR Labeled with release/backport (if the PR is an urgent fix that needs to be backported)