Connect Cherry Studio to EXAIV: Desktop Multi-Model Client Setup
Add EXAIV as a custom model provider in the Cherry Studio desktop app: API host format, auto-fetching the model list, chat/knowledge-base setup in one pass.
Connect Cherry Studio to EXAIV: Desktop Multi-Model Client Setup
Cherry Studio is a cross-platform (Windows/macOS/Linux) desktop AI client that aggregates multiple model services. EXAIV's OpenAI-compatible endpoint plugs in as a custom provider and supports auto-fetching the model list — no manual model names required.
Configuration steps
Open model services
Bottom-left Settings → Model Services.Add a provider
Pick "OpenAI" from the list or add a custom provider named EXAIV.Host and key
API Host:https://exaiv.com/api/ai/v1; API Key: your EXAIV key.Fetch models
Click "Manage" / "Fetch model list" — the client lists all available EXAIV models; check the ones you want.
Auto-fetch models
The client queries the endpoint's model-list endpoint. If it fails, add model names manually using the IDs on the models page.
Verify
Create a chat on any enabled model and send a message. Once it replies, try the knowledge-base Q&A or a translation assistant to confirm multi-scenario usage.
Common issues
| Symptom | Fix |
|---|---|
| Model list fetch fails | Does the API host end with /v1? Is the key valid? |
| 401 | Key incomplete or revoked — regenerate in the console |
| Chat timeouts | Check whether a system proxy interferes with direct access |
Advanced: knowledge base + multi-model division of labor
Cherry Studio's knowledge base vectorizes local documents for Q&A. Recommended division with EXAIV:
- Vectorization (embedding): pick an embedding model from the endpoint's supported list; done once per library
- Knowledge-base Q&A: use a long-context text model to reduce chunk loss
- Daily chat: a fast model keeps costs down
Bind different assistants to different model combinations and one client covers both "document research" and "high-frequency daily chat". All usage is visible as credits in the EXAIV console — spend stays transparent.
Data safety note
Knowledge-base documents stay local; chunk content is only sent to the selected model endpoint during vectorization. For sensitive documents, verify your environment's compliance requirements first.
Keeping models fresh
EXAIV adds new models continuously, and Cherry Studio picks them up without reinstalling anything: reopen the provider's model manager and fetch the list again — newly added models appear automatically and can be checked into your picker.
A practical rhythm: check the models page once a month, and when a model that fits your workload appears, enable it in Cherry Studio and switch the relevant assistants over. Because all models bill under the same key on the same credit balance, trying a new model carries zero configuration cost — if it does not fit, switch back in one click.
Community templates
Reading the real error before changing settings
When a request fails, resist the urge to toggle settings at random. Cherry Studio surfaces the upstream error body in the conversation — open the message details and read it. The three families you will see:
- 401 unauthorized: the key is wrong, truncated, or revoked — fix it at the provider settings, not the model picker
- 404 not found: the API host path is wrong for this provider type (the /v1 question) — compare against the exact working path in this guide
- 429 rate limited: the model is busy or your balance/limits need attention in the EXAIV console
Each family points at a different layer, and only the third one involves EXAIV-side state at all. This habit — read, classify, then change one thing — turns integration from guesswork into a five-minute exercise.