Skip to main content
Cherry Studio is an open-source cross-platform desktop AI client. It can hold configs for several LLM providers at once, supports a local knowledge base and agents, and connects to RuAPI through the OpenAI-compatible protocol.

​
Prerequisites

1

RuAPI account + API key

Console → API Keys → create a sk-... key.

​
Install Cherry Studio

Download from cherry-ai.com or GitHub Releases:
Download Cherry-Studio-Setup-x.y.z.exe and run the installer.

​
Add RuAPI as a provider

1

Open Settings → Model Services

Once Cherry Studio is running, click the ⚙️ icon at the bottom-left → Settings → Model Services in the left panel.
2

Click 'Add'

Find the + Add button at the bottom of the provider list.
3

Basic info

Click OK.
4

API key and base URL

In the RuAPI provider detail panel:
Cherry Studio appends /v1/chat/completions automatically, so https://www.ruapi.ai is enough. https://www.ruapi.ai/v1 usually works too (Cherry Studio dedupes).
Toggle Enable on.
5

Add models

Below the same panel, click + Add Model and enter model IDs:
  • gpt-5.4
  • gpt-5.4-mini
  • claude-sonnet-5
  • claude-opus-4-8
  • claude-haiku-4-5-20251001
  • deepseek-v4-pro
  • gemini-2.5-pro
Full list at the pricing page.
Click Fetch Model List and Cherry Studio will pull the full catalog via /v1/models and add them all.

​
First call

Back on the main screen, New chat in the top-left, click the model name in the top-right and pick RuAPI → your model (or any model you added). Send:
A normal reply plus a matching RuAPI Console → Usage Logs entry = success.

​
Default model choices

Settings → Default Models:
  • Default assistant model: everyday chat. Recommend gpt-5.4 or Claude Sonnet.
  • Topic naming model: auto-titles conversations. Recommend a cheap small model like Claude Haiku.
  • Translation model: gpt-5.4-mini or Claude Haiku work well.

​
Troubleshooting

  • The API key must be a RuAPI key (sk-...), not an OpenAI key.
  • Watch for trailing whitespace.
  • Check the key in RuAPI Console → API Keys.
  • Model ID typo. Hand-entered names must match RuAPI exactly — use Fetch Model List to autoload them.
  • Don’t append /chat/completions to the API host — https://www.ruapi.ai is enough.
Streaming (SSE) is enabled by default on both sides. If it stutters:
  • Check network stability.
  • Try disabling streaming in settings to compare.
Cherry Studio’s MCP relies on model-level function calling. GPT and Claude models handle it reliably — see function calling.
Make sure the selected model supports vision. On RuAPI, GPT, Claude, Gemini and GLM-5V-Turbo models understand images — see vision.
Cherry Studio supports any number of providers. Add each one the same way. The model dropdown in the top-right of each chat switches between them.

​
Advanced

  • Local knowledge base: Cherry Studio’s built-in RAG needs an embedding model. RuAPI doesn’t offer embedding models right now — use another provider’s or a local embedding model for the knowledge base, and keep chat on RuAPI.
  • Agents: Cherry Studio’s agent system runs client-side and calls your configured provider. Claude Sonnet or gpt-5.4 are solid agent backbones.
  • Multiple keys: if one key isn’t enough, create several with separate limits in the RuAPI Console and bind them to different Cherry Studio profiles.