Understand what the Base URL controls
The Base URL tells a client where to send API requests. Some clients expect the service root, some require an OpenAI-compatible version path, and others append the version automatically. Mixing formats can duplicate or omit part of the URL.
Use the current guide for the client you are configuring. Avoid copying an old screenshot or combining fields from different providers.
Web login and API key access are different
A successful website login proves that a browser session exists. API calls depend on a separate key that must exist, remain enabled, have sufficient quota, not be expired, and include access to the target group.
Create keys by purpose: one for a local coding assistant, one for image testing, and another for automation. Dedicated keys are easier to revoke and make usage attribution possible.
Use the current model identifier
The API model name is not always the same as a product label in a client. Providers add and remove models, and visibility may change by group, protocol, cost, or availability.
Check the model marketplace immediately before configuration. If a model is visible but the real request fails, continue with key permissions, balance, endpoint type, and request format instead of assuming the model is healthy.
Follow one repeatable diagnostic sequence
Verify the Base URL, API key, model name, group access, account balance, usage limits, and channel status. Do not change several fields at once, because that makes the cause difficult to reproduce.
Finish with one minimal real request. A models endpoint can prove that a key is recognized and a model is visible, but it cannot prove that chat, image generation, or a long-context task will complete.