feat: Add support for separate LLM and embedding model endpoints

[sriramsowmithri9807]

Description (issue solved #1367)

This PR adds support for using different endpoints for LLM inference and embedding models, which is particularly useful when running separate llama.cpp servers for each function.

Changes Made

• Added LLM_ENDPOINT and EMBEDDING_ENDPOINT configuration options
• Updated GenericLLMProvider to handle custom base URLs for different providers
• Enhanced embedding initialization to support separate endpoints
• Improved configuration handling for both LLM and embedding providers
• Added proper environment variable support for endpoint configuration

How to Test

1. Set up your environment variables:

   export LLM_ENDPOINT="http://localhost:8080/v1"
   export EMBEDDING_ENDPOINT="http://localhost:8081/v1"
   export FAST_LLM="openai:llama3"  # or your preferred model
   export EMBEDDING="openai:llama-embed"  # or your preferred embedding model
   

Or update your config file:

python

{
    "LLM_ENDPOINT": "http://localhost:8080/v1",
    "EMBEDDING_ENDPOINT": "http://localhost:8081/v1",
    "FAST_LLM": "openai:llama3",
    "EMBEDDING": "openai:llama-embed"
}