[Feature Request] Dynamic Model Routing & Multi-Model Support

[lichengwu]

Motivation

Currently, in local deployment scenarios, Eigent only supports running tasks with a single, explicitly configured default model. Even if a task requires multimodal capabilities (e.g., generating images) or would benefit from a specialized model (e.g., a code-generation specific model), the system will not automatically select an appropriate alternative (such as a multimodal version of qwen) if a text-only model is configured.

Solution

Core Requirements:

• Intelligent Task-Model Routing: The system should be able to analyze task requirements (e.g., code writing, image generation) and automatically route them to the most suitable configured model (e.g., multimodal models for image tasks) rather than relying solely on a default configuration.
• Multi-Provider & Personalized Configuration: Support for configuring multiple providers and allowing different Agents to bind to specific, personalized models to improve task accuracy and efficiency.
• Future Roadmap: As mentioned by the development team, this functionality is planned for future support, and a more detailed roadmap is expected.

Alternatives

No response

Additional context

No response