idea: remote engines with embeddings, vision support

[louis-menlo]

Problem Statement

Client can just work with chat/completions endpoint for now.

Feature Idea

This is to enhance the app ability to interact with multiple OpenAI-compatible endpoints. Currently, we support chat/completions, allowing clients to send requests to a single endpoint. This update expands functionality to include support for additional endpoints and data types, specifically:

• Multiple Endpoint Support: Enables clients to seamlessly switch between and utilize various OpenAI-compatible endpoints beyond just chat/completions.
• Embeddings: Adds support for generating embeddings, empowering clients to leverage this powerful feature for tasks like semantic search and clustering.
• Image Upload: Introduces the capability to upload images, opening up possibilities for image-based interactions with supported models.
• Tool Use: With dynamic chat template support, also remote providers support using function calling.

This expanded support provides greater flexibility and allows clients to integrate a wider range of AI capabilities into their applications.