Bug: OpenAI-compatible providers overlap and hide existing models/keys in settings

Description

There is a major issue in the jcode repository regarding how API keys and models are managed under the "jcode ani model provider settings".

Currently, the system groups all OpenAI-compatible API keys and models under a single generic tag: "openai compatible". Mixing all different providers into one single category causes severe overlapping issues.

Environment

• JCode Version: v0.75.5

Current Behavior

• All unique OpenAI-compatible providers are mixed together under one tag.
• Adding or modifying a provider often causes previously saved keys, custom endpoints, or models from other providers to completely disappear or become hidden.
• The settings overwrite or mask existing configurations instead of keeping them isolated.

Expected Behavior

Each OpenAI-compatible provider configuration should be handled independently, or the system should accurately persist and display multiple distinct endpoints/keys without wiping out existing data.

This area of the codebase requires a deep review and refactoring to ensure data persistence and a proper UI/UX separation for custom providers.

Start by tracing how model-provider settings group and persist OpenAI-compatible keys, endpoints, and models, then reproduce the issue by adding configurations for multiple providers. Done means each provider remains independently visible and existing keys, endpoints, and models are preserved after adding or modifying another provider.