Custom Base URL Support for All Model Providers (Especially Image Generation Endpoints)

Many model providers operate region-specific API endpoints with different base URLs depending on the user's region. For example:

MiniMax uses minimaxi.com for mainland China and minimax.io for overseas users GLM, Moonshot, and other providers similarly maintain separate domestic and international endpoints Cloud platforms like Azure OpenAI, AWS Bedrock, and GCP Vertex AI host models behind their own custom API endpoints Currently, TypingMind only supports custom base URL configuration for OpenAI, Anthropic, and Gemini. This is a great start, but far too limited — many other integrated providers (Moonshot, GLM, MiniMax, etc.) also need this flexibility, as they serve different regions with entirely different endpoints.

The Image-2 plugin is the most critical gap right now.

The Image-2 plugin (which handles /v1/images/generations and /v1/images/edit) has its base URL hardcoded to OpenAI's official API (api.openai.com) with no option to override it. This means:

Users who access OpenAI-compatible image APIs through Azure OpenAI Service cannot use the Image-2 plugin at all. Users who rely on region-specific endpoints or any other OpenAI-compatible image API hosting are completely locked out. For chat models, at least OpenAI's base URL can be customized — but the Image-2 plugin, which also calls OpenAI-format endpoints, offers zero flexibility. This inconsistency is frustrating. There is currently no workaround. The plugin simply does not work unless you can directly reach api.openai.com.

Why this matters:

Large affected user base: China has a massive community of AI power users and TypingMind customers. Providers like MiniMax, Moonshot, and GLM explicitly serve this market with dedicated domestic endpoints — TypingMind should be able to reach them. Not just a China issue: Users worldwide may need to route requests through Azure OpenAI, AWS, GCP, or enterprise gateways that expose the same OpenAI-compatible API under a different base URL. Consistency: If OpenAI/Anthropic/Gemini already support custom base URLs, there's no reason other providers and the Image-2 plugin shouldn't. Proposed solution:

(High priority) Allow the Image-2 plugin to accept a custom base URL instead of being hardcoded to api.openai.com. This is the most impactful single change, as there is currently zero workaround for affected users. (General) Extend custom base URL support to all integrated providers — not just OpenAI, Anthropic, and Gemini. As TypingMind integrates more providers with regional endpoint variations, this will only become more important. These are relatively small configuration changes but would immediately unblock a significant portion of your user base. Thank you for considering this! 🙏