Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.mirobody.ai/llms.txt

Use this file to discover all available pages before exploring further.

Aliyun Model Studio (DashScope) is Mirobody’s recommended LLM backend for the China cluster. One key gives you access to Alibaba’s Qwen family, plus DeepSeek / Kimi / GLM / MiniMax, plus embeddings and rerankers.
Official docs: What is Model Studio · Model list · API Key console

​
1. Register and get a key

1

Open Model Studio

Sign in to the Bailian console. Accounts are auto-enrolled; some models require pay-as-you-go activation.
2

Create an API Key

On the API Key page, create a key and copy it — shown once.
3

Confirm billing

Under Billing, enable paid usage for the models you want; some new models include free tier.

​
2. Endpoint info

DashScope offers both the native protocol and an OpenAI-compatible mode. Mirobody talks to the native protocol — you only need to pass the API key via BYOK headers; no base URL configuration is required on your side.
FieldValue
OpenAI-compatible base URL (reference)https://dashscope.aliyuncs.com/compatible-mode/v1
Native base URL (used by Mirobody)https://dashscope.aliyuncs.com/api/v1
AuthAuthorization: Bearer <DASHSCOPE_API_KEY>
RegionsHangzhou (default) · Virginia · Singapore · Frankfurt

​
3. Available models

The Mirobody /api/chat provider field routes to one of these. Refer to the official model list for the live catalog.

​
Text generation

Mirobody providerUnderlying model idWhen to use
DeepSeekFlashdeepseek-v4-flashDefault — fast, low cost
DeepSeekProdeepseek-v4-proComplex reasoning, medical interpretation
kimikimi-k2.6Long-context document analysis
minimaxMiniMax-M2.7Natural Chinese expression
— (direct via DashScope)qwen3.7-maxQwen flagship for long, multi-step tasks
— (direct via DashScope)qwen3.6-plusBalanced cost/performance — official default pick
— (direct via DashScope)qwen3.6-flashHigh throughput, low latency
— (direct via DashScope)glm-5.1Zhipu GLM
— (direct via DashScope)mimo-v2.5-proXiaomi MiMo

​
Embeddings & rerank

ModelUse case
text-embedding-v4Text embedding
tongyi-embedding-vision-plusMultimodal embedding
qwen3-rerankReranker

​
4. Use it in Mirobody

curl -N -X POST https://mcp.thetahealth.cn/api/chat \
  -H "Authorization: Bearer $TOKEN" \
  -H "X-LLM-Provider: dashscope" \
  -H "X-LLM-Api-Key: sk-xxx-from-dashscope" \
  -H "Content-Type: application/json" \
  -d '{"question":"hi","agent":"App","provider":"DeepSeekFlash"}'

​
Troubleshooting

  • Upstream InvalidApiKey → check that the Bailian key is active and the chosen model is enabled.
  • Upstream Throttling.User → DashScope-side RPM/TPM limit; adjust quotas in API Key Usage.
  • Model ID not found → some newer models require explicit activation in the “Model Plaza”.
  • No VPN needed — DashScope is reachable directly from mainland China.