[Feature]: LLM – Support for Multiple Model

[wnghdcjfe]

Current Situation/Problem

• The llm script has no abstraction layer for model backends.
• All requests assume an OpenAI-style API (prompt → completion).
• Users running self-hosted models (e.g. LLaMA + Ollama, vLLM, LocalAI) cannot integrate easily.

Proposed Feature

• Backend Abstraction Layer: Introduce a pluggable model provider interface (e.g., provider: "openai" | "llama" | "custom").
• Configurable Model Selection: Allow setting the model provider and model name in config or via CLI flags:

Alternatives (Optional)

No response

Additional Information

No response