Cherry Studio · Client
Connect Cherry Studio to Tikway: custom API URL, model IDs and troubleshooting
Last updated：
A desktop client helps test model responses, but a connection test does not verify every feature. This guide configures Tikway using OpenAI-compatible Chat Completions. Menu names vary by Cherry Studio version; follow the settings in your installed version.
1. Prepare a key and model
Get a Tikway key from the console. In the catalog, select a text model supporting the OpenAI protocol and Chat Completions, and copy its complete ID. Begin with text before adding images, tools or optional parameters.
2. Add a custom provider
In the model-service or provider settings, add an OpenAI-compatible provider supporting a custom URL and name it Tikway. Choose a type that allows your own URL and key.
| Setting | Value to use |
|---|---|
| Provider type | OpenAI-compatible with a custom URL |
| API key | Your own Tikway API key |
| Base URL | Start with https://api.tikway.ai/v1, then check the version’s path-building rules |
| Model ID | Copy the complete Tikway catalog ID unchanged |
| Final text URL | https://api.tikway.ai/v1/chat/completions |
Some versions append /v1 automatically or offer a full-URL mode. Inspect the final URL and avoid /v1/v1/chat/completions or repeated /chat/completions. If inspecting network requests, record URLs, statuses and errors with authentication removed.
If fetching a model list is supported, try it and compare with the catalog. Failure to fetch a list does not prove text requests are unavailable; use manual model entry when the version allows it.
3. Send a minimal request
Enable the new provider/model, start a conversation and send “Reply only with OK”. Confirm model content returns, then match timestamp, model and cost in Tikway’s request records. Connection tests and ordinary conversations may use different requests; check both.
Voice, images and tools require suitable models and separate checks. This tutorial does not guarantee every version or plugin works.
4. Troubleshoot in order
401: Check for another platform’s key, whitespace or stale credentials. See authentication troubleshooting.
404 or missing model: Check the final URL, then the model ID and protocol; changing a display name is insufficient.
Website opens but desktop API times out: Check the desktop process’s network and proxy. The website and API use different domains, so homepage access does not verify API connectivity.
Streaming stalls or stops: Check termination events and use the streaming troubleshooting guide.
5. Costs and request records
Conversation history can be included in input each turn. Check actual tokens in response usage and request records. The cost calculator estimates ordinary text budgets; images, caches and extras need separate checks. For programmatic integration, see the OpenAI SDK guide.