Custom Provider Support

[334ms-dev]

Description

Quests.dev currently supports models via OpenRouter and Anthropic, but it does not allow developers to plug in their own custom model endpoints.

This limits flexibility for teams who want to use:

• Self-hosted models (e.g., Ollama, vLLM, LM Studio).
• Alternative providers (Mistral, Qwen, OpenAI-compatible APIs not routed through OpenRouter).
• Private endpoints for fine-tuned or proprietary models.

Expected Behavior

• Developers should be able to configure any LLM provider via a custom API endpoint.
• Support for OpenAI-compatible APIs as well as generic REST-based providers.
• Local/offline model usage should be possible.

Actual Behavior

• Only models from OpenRouter or Anthropic are selectable.
• No way to define custom endpoints or self-hosted models.

Benefits

• Unlocks offline / local-first workflows.
• Reduces vendor lock-in by enabling any LLM backend.
• Supports experimentation with open-source and fine-tuned models.

[mutewinter]

This feature is planned.

[Filippide8864]

Having the option to use compatible OpenAI providers would also be very useful for me, allowing me to use Quests with Groq for free and quickly.