[Feature] Support openai compatible LLM running in private host

Summary

Allow users to configure to use local LLM like llama.cpp/vllm/sglang running on private hosts with openai compatible API

Why This Matters

Users with mid-range hardware may use llama.cpp/vllm/sglang inference engine which support openai compatible API to minimize cost

Scope

• LLM/provider change:

Proposed Behavior

A configuration in .env for like OPENAI_BASE_URL to configure to point to a private host like http://10.0.0.X:8000/v1