Separate client fees from Link API usage
When adding Link in Chatbox or Cherry Studio, use a Link key and https://www.aifast.link/v1, and copy the ID from the current model directory. Client subscriptions, model-vendor memberships and Link API balances are separate products.
Test short text first, then attachments and tool calls separately. With a Link plan, check whether the selected model belongs to its group and inspect the charge log. An answer in the chat UI does not prove that the request used plan allowance.
Review Link funding and monthly-plan rules
Common configuration parameters
| Field | Recommended value |
|---|---|
| Supplier type | OpenAI Compatible / Custom OpenAI |
| API Key | an independent API key created in the AIFast console |
| Base URL | https://www.aifast.link/v1 |
| Model ID | Copy data[].id from the /v1/models response |
Verify before configuring:
curl https://www.aifast.link/v1/models \
-H "Authorization: Bearer $AIFAST_API_KEY"Chatbox configuration ideas
- Open the model provider or model settings.
- Choose OpenAI API or OpenAI Compatible custom service.
- Fill in the API Key and Base URL.
- Add the real model ID manually, or refresh the model list.
- After saving, create a new session for minimal testing.
If Chatbox will automatically append /v1, the address should be adjusted based on connection testing or request logs to avoid /v1/v1.
Cherry Studio configuration ideas
- Go to "Settings/Model Services".
- Add custom suppliers.
- Select the OpenAI Compatible method.
- Fill in the API Key and API address.
- Use the Manage or Add Model function to enter the real model ID.
- Perform a connection check and set the model as the assistant default model.
The client interface will be updated. If the menu name is different from this article, you should look for similar fields such as "Model Service", "Supplier", "API Address", "Custom Model", etc.
Connection test passed but unable to chat
Connection tests may only check domain names, certifications, or model lists. Keep checking:
- Whether the model actually selected in the session is consistent with the added model.
- Whether the model ID contains extra spaces or a Chinese name.
- Request whether to enter
/v1/chat/completions. - Whether the current model supports images, tool calls, or long contexts.
401, 404 and model not found
- 401: Check API Key and Bearer authentication.
- 404: Check Base URL and client automatic splicing rules.
- model not found: Requery
/v1/models, copy the realid. - Return HTML: The request did not enter the API route.
Multi-device use
Different devices use independent keys to facilitate statistics, quotas and revocation. Do not pass keys via public sync services, chat logs, or screenshots.
Client versions and request-path checks
Official release source was reviewed on 11 October 2026: Chatbox v1.23.5 and Cherry Studio v2.1.6. This is a review of those specific versions, not a full desktop-app acceptance test. Menus, path normalization and endpoint settings may differ in other versions. The earlier UI verification date is not replaced with a new desktop-test date.
Chatbox: avoid duplicating the version path across API Host and API Path
The normalizeOpenAIApiHostAndPath pure function from v1.23.5 was executed independently, and the custom OpenAI request code was checked for API Host + API Path concatenation. The following results are actual function outputs; no key or paid inference was involved.
| API Host | API Path | URL |
|---|---|---|
| https://www.aifast.link/v1 | /chat/completions | https://www.aifast.link/v1/chat/completions |
| https://www.aifast.link/v1 | /v1/chat/completions | https://www.aifast.link/v1/v1/chat/completions |
The first row uses the default chat path. The second demonstrates a duplicated /v1: change API Path to /chat/completions and check again. A correct path still requires model-access and response tests. For the Responses protocol, verify /v1/responses rather than copying the chat endpoint.
Cherry Studio: confirm the provider’s primary endpoint
The connection settings in v2.1.6 distinguish the generic API Host from Anthropic API Host according to the provider’s primary endpoint and expose a request-configuration control. Confirm the provider type and final request URL. Changing only the model name does not make OpenAI Chat, Responses and Anthropic Messages interchangeable.
For an actual integration test, save the provider configuration without secrets, client version, selected conversation model, final request path, short response and corresponding Link usage log. A connection check covers only part of this process; source screenshots are not substitutes for successful-chat screenshots.
Chatbox v1.23.5 · Chatbox Custom OpenAI · Cherry Studio v2.1.6