Add Alibaba Cloud (Model Studio / DashScope) LLM provider (#4738)

Author: georgi

.claude/skills/nodetool-chat-cli/SKILL.md

@@ -54,7 +54,7 @@ nodetool chat --sandbox
 | `--trace-stdout [pretty\|json]` | Stream spans to stdout |
 
 Providers: `anthropic`, `claude_agent_sdk`, `openai`, `codex`, `gemini`, `xai`,
-`groq`, `mistral`, `deepseek`, `moonshot`, `minimax`, `cerebras`, `gmi`,
+`groq`, `mistral`, `deepseek`, `moonshot`, `minimax`, `cerebras`, `alibaba`, `gmi`,
 `together`, `openrouter`, `huggingface`, `replicate`, `kie`, `aki`, `ollama`,
 `lmstudio`, `mlx`. Any other registered provider id (e.g. `vllm`) also works when
 passed explicitly.

.claude/skills/nodetool-model-provider-config/SKILL.md

@@ -22,7 +22,8 @@ You help users configure AI model providers and select the right models for thei
 | **MLX** | Local | — | Apple Silicon optimized |
 
 Other registered chat providers (any of these is valid for `-p/--provider`):
-`groq`, `mistral`, `deepseek`, `moonshot`, `minimax`, `cerebras`, `together`,
+`groq`, `mistral`, `deepseek`, `moonshot`, `minimax`, `cerebras`, `alibaba`
+(`DASHSCOPE_API_KEY`, Qwen via Alibaba Cloud Model Studio), `together`,
 `openrouter`, `codex`, `claude_agent_sdk`, `lmstudio`. Run `nodetool models
 providers` to see configured providers and `nodetool models recommended` for the
 curated model list.

CLAUDE.md

@@ -198,7 +198,7 @@ npm run dev:chat -- --url ws://localhost:7777/ws
 Chat flags:
 ```
 -p, --provider <name>    anthropic, openai, gemini, xai, groq, mistral, deepseek,
-                         moonshot, minimax, cerebras, together, openrouter,
+                         moonshot, minimax, cerebras, alibaba, together, openrouter,
                          huggingface, replicate, kie, aki, ollama, lmstudio,
                          claude_agent_sdk, codex, gmi, mlx, node_llama_cpp
                          (any registry provider id also works, e.g. vllm, llama_cpp)

docs/architecture.md

@@ -187,6 +187,7 @@ NodeTool supports 20+ AI model providers through a unified provider interface:
 | **Mistral** | Text generation |
 | **Together** | Text, embeddings |
 | **Cerebras** | Fast text inference |
+| **Alibaba Cloud** | Text (Qwen models) |
 | **GMI Cloud** | Open-weight text inference |
 | **OpenRouter** | Multi-provider routing |
 | **vLLM** | Self-hosted inference |

docs/developer/providers/index.md

@@ -14,7 +14,7 @@ NodeTool spans many providers but only a few distinct mechanisms. Find your prov
 
 | Mechanism | What "add a model" means | Providers |
 | :-------- | :----------------------- | :-------- |
-| **Dynamic fetch** | Nothing — models appear from the provider's live `/models` (or local daemon) endpoint. Code only for capability/cost overrides. | [OpenAI](openai.md) (chat), [Anthropic](anthropic.md), [Gemini](gemini.md) (text), [xAI](xai.md), [OpenAI-compatible](openai-compatible.md) (Groq, Mistral, DeepSeek, Moonshot, Cerebras, Cohere, OpenRouter), [Ollama](ollama.md), [Local inference](local-inference.md) (LM Studio, llama.cpp, vLLM) |
+| **Dynamic fetch** | Nothing — models appear from the provider's live `/models` (or local daemon) endpoint. Code only for capability/cost overrides. | [OpenAI](openai.md) (chat), [Anthropic](anthropic.md), [Gemini](gemini.md) (text), [xAI](xai.md), [OpenAI-compatible](openai-compatible.md) (Groq, Mistral, DeepSeek, Moonshot, Cerebras, Alibaba Cloud, Cohere, OpenRouter), [Ollama](ollama.md), [Local inference](local-inference.md) (LM Studio, llama.cpp, vLLM) |
 | **Static model list** | A code edit to a hardcoded array in the provider (plus cost entry). | [OpenAI](openai.md) (image), [Gemini](gemini.md) (Imagen/Veo) |
 | **Codegen from upstream schemas** | Add an endpoint id to a config, run the generator — the manifest is regenerated, never hand-edited. | [FAL](fal.md), [Replicate](replicate.md), [KIE](kie.md) |
 | **Hand-maintained manifest** | Add a JSON entry to the package manifest. | [AtlasCloud](atlascloud.md), [Topaz](topaz.md), [Together](together.md) |
@@ -26,7 +26,7 @@ NodeTool spans many providers but only a few distinct mechanisms. Find your prov
 - **[Anthropic (Claude)](anthropic.md)** — models fetched live from the Anthropic API.
 - **[Google Gemini](gemini.md)** — text auto-fetches; Imagen/Veo are static lists.
 - **[xAI (Grok)](xai.md)** — chat, image, and video classified from `/v1/models`.
-- **[OpenAI-compatible providers](openai-compatible.md)** — Groq, Mistral, DeepSeek, Moonshot, Cerebras, Cohere, OpenRouter, and how to add a new one.
+- **[OpenAI-compatible providers](openai-compatible.md)** — Groq, Mistral, DeepSeek, Moonshot, Cerebras, Alibaba Cloud, Cohere, OpenRouter, and how to add a new one.
 - **[FAL](fal.md)** — nodes generated from FAL OpenAPI schemas via codegen.
 - **[Replicate](replicate.md)** — nodes generated from Replicate model schemas via codegen.
 - **[KIE](kie.md)** — nodes generated from per-model configs into a manifest.

docs/developer/providers/openai-compatible.md

@@ -6,8 +6,8 @@ description: "How to add a new OpenAI-compatible cloud provider (Groq, Mistral,
 
 > **Audience:** Coding agents and contributors who want to add a new OpenAI-compatible cloud provider or change how an existing one exposes models.
 
-This guide covers seven providers that share one implementation pattern:
-**Groq**, **Mistral**, **DeepSeek**, **Moonshot (Kimi)**, **Cerebras**, **Cohere**, and **OpenRouter**.
+This guide covers eight providers that share one implementation pattern:
+**Groq**, **Mistral**, **DeepSeek**, **Moonshot (Kimi)**, **Cerebras**, **Alibaba Cloud (Qwen)**, **Cohere**, and **OpenRouter**.
 
 ---
 
@@ -120,7 +120,7 @@ export class DeepSeekProvider extends OpenAIProvider {
 }
 ```
 
-Groq, Mistral, Cerebras, and OpenRouter follow this pattern exactly. OpenRouter adds a few extras: `defaultHeaders` on the `OpenAI` client (HTTP-Referer, X-Title), `textToImage`, and a `hasToolSupport` override that returns `false` for o1/o3 model IDs.
+Groq, Mistral, Cerebras, Alibaba Cloud, and OpenRouter follow this pattern exactly. Alibaba Cloud reads `DASHSCOPE_API_KEY` and points at Model Studio's international DashScope endpoint, `https://dashscope-intl.aliyuncs.com/compatible-mode/v1`. OpenRouter adds a few extras: `defaultHeaders` on the `OpenAI` client (HTTP-Referer, X-Title), `textToImage`, and a `hasToolSupport` override that returns `false` for o1/o3 model IDs.
 
 ---
 
@@ -189,7 +189,7 @@ The empty string for `ACME_API_KEY` is intentional — it forces the registry to
 
 ### Chat models are fetched dynamically — no change needed
 
-If Groq, DeepSeek, Mistral, Cerebras, or OpenRouter adds a new chat model, it appears in the model picker automatically on the next call to `getAvailableLanguageModels()`. No code change is required.
+If Groq, DeepSeek, Mistral, Cerebras, Alibaba Cloud, or OpenRouter adds a new chat model, it appears in the model picker automatically on the next call to `getAvailableLanguageModels()`. No code change is required.
 
 ### Tool-support overrides
 

docs/models-and-providers.md

@@ -54,6 +54,7 @@ The generic `nodetool.agents.Agent` and chat nodes route to whichever provider o
 | Groq | Llama, Qwen, GPT-OSS, Kimi K2 on LPU inference |
 | Mistral | Mistral Large, Mistral Medium, Mistral Small, Mixtral, Codestral, Magistral |
 | Cerebras | Llama, Qwen, GPT-OSS on wafer-scale inference |
+| Alibaba Cloud | Qwen3-Max, Qwen-Plus, Qwen-Flash, Qwen-Turbo, Qwen3-VL |
 | GMI Cloud | Llama, DeepSeek, Qwen (open-weight) |
 | Moonshot | Kimi K2, Kimi latest |
 | <img src="assets/icons/minimax.svg" width="16" height="16" style="vertical-align: middle;" alt="" /> MiniMax | MiniMax-Text, MiniMax M2 |

docs/providers.md

@@ -24,6 +24,7 @@ Checked against each provider's implementation in `packages/runtime/src/provider
 | Groq | ✅ | | | | | | |
 | Mistral | ✅ | | | | | ✅ | |
 | Cerebras | ✅ | | | | | | |
+| Alibaba Cloud | ✅ | | | | | | |
 | GMI Cloud | ✅ | | | | | | |
 | OpenRouter | ✅ | ✅ | | | | | |
 | Together AI | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | |
@@ -92,6 +93,10 @@ Mistral runs chat models (Mistral, Mixtral) plus one embedding model, `mistral-e
 
 Cerebras runs chat models on its high-throughput inference hardware. Text only. Cloud only, keyed by `CEREBRAS_API_KEY`. See the [OpenAI-compatible providers guide](developer/providers/openai-compatible.md).
 
+## Alibaba Cloud
+
+Alibaba Cloud Model Studio serves the Qwen chat models — qwen3-max, qwen-plus, qwen-flash, qwen-turbo, and the qwen3-vl vision line — with streaming and tool calling. Text only. Cloud only, keyed by `DASHSCOPE_API_KEY`, over the DashScope OpenAI-compatible endpoint (default `https://dashscope-intl.aliyuncs.com/compatible-mode/v1`, the international/Singapore region); models are fetched live from its `/models` endpoint. Model Studio keys are region-scoped — set `DASHSCOPE_BASE_URL` to your region's `/compatible-mode/v1` endpoint when the key was created outside Singapore. Tool calling is gated per model family: Qwen-Math, QVQ, OCR, translation, and audio models are outside Alibaba's function-calling list and are not sent tools. Get a key at [Model Studio](https://modelstudio.console.alibabacloud.com/). See the [OpenAI-compatible providers guide](developer/providers/openai-compatible.md).
+
 ## GMI Cloud
 
 GMI Cloud is an OpenAI-compatible chat endpoint for open-weight models — Llama, DeepSeek, and Qwen variants. Text only. Cloud only, keyed by `GMI_API_KEY`.

electron/src/components/BootMessage.tsx

@@ -76,6 +76,12 @@ const PROVIDERS: Provider[] = [
     signupUrl: "https://cloud.cerebras.ai/",
     capabilities: ["Fast Inference", "Chat", "Text Generation"],
   },
+  {
+    name: "Alibaba Cloud",
+    description: "Qwen models via Alibaba Cloud Model Studio's OpenAI-compatible API.",
+    signupUrl: "https://modelstudio.console.alibabacloud.com/",
+    capabilities: ["Chat", "Text Generation", "Reasoning"],
+  },
   {
     name: "GMI Cloud",
     description: "OpenAI-compatible inference for open-weight LLMs like Llama, DeepSeek, and Qwen.",

marketing/scripts/fetch-provider-logos.mjs

@@ -40,6 +40,7 @@ const PROVIDER_DOMAINS = {
   openrouter: "openrouter.ai",
   huggingface: "huggingface.co",
   cerebras: "cerebras.ai",
+  alibaba: "alibabacloud.com",
   moonshot: "moonshot.ai",
   // media specialists
   elevenlabs: "elevenlabs.io",