Suanlimao
← Back to all tutorials

Connect your tools

RAGFlow

Add an OpenAI-API-compatible provider under RAGFlow's Model providers.

About 6 minFor:Admins who self-host RAGFlow for document Q&A

Before you start

  • RAGFlow is deployed and you can sign in
  • A key and a model name

Step 1 · STEP 0x01

Choose a provider

User settings → Model providers
1

Available models

OpenAI-API-compatible
2

Instance name

Anything you like
Diagram only — the real screen in your version may differ

Go to User settings → Model providers, pick the OpenAI-API-compatible provider under Available models and name the instance. The Chinese UI uses the equivalent labels; follow your version.

Create a key →

Step 2 · STEP 0x02

Enter the key and Base URL, then add a model

OpenAI-API-compatible
1

API Key

<YOUR_API_KEY>
2

Base URL

https://api.suanlimao.shop/v1
3

Model name (custom model)

<MODEL_NAME>
4

Model type

chat
5

Max context length (optional)

Per the model's details
6

Tool calling (optional)

On if the model supports it
7
Verify and save
Diagram only — the real screen in your version may differ
  • Base URL is https://api.suanlimao.shop/v1. The docs say to “confirm whether the path must include /v1” for compatible APIs — ours does.
  • When adding a custom model the docs require the name to match the model ID exposed by the provider API — the name on the plaza.
  • Saving runs a verification; on failure a warning marker appears — open it for the error.
Don't guess the model name: open the model plaza, click “Copy model id” on a model card and paste it. A wrong model name is the most common error.

Open the model plaza →

Step 3 · STEP 0x03

Set as default and test

Model providers → System model settings
1

Chat model

<MODEL_NAME>
2

Chat assistant → test

Summarise this document in one sentence
Diagram only — the real screen in your version may differ
  • Point the Chat model in system model settings to it and try a question in a chat assistant.
Embedding and rerank models are added the same way with the matching model type; for types the plaza does not offer, use RAGFlow's built-in models.

✓ How to tell it worked

  • The provider verifies, chat assistants answer and the Usage page shows the request.

FAQ

Verification fails?

The docs' order: credentials, endpoint (is /v1 present), model name, service availability, quota. Check each and verify again.

Chat works but parsing documents is slow?

Parsing uses embedding and OCR models, not the chat model; check those configurations.

Related tutorials

Still stuck?

Read the troubleshooting tutorials first. If that does not help, open a ticket with the request_id, time, model and the exact error (never paste your key).