app/config.py

"""Centralized configuration with dual LLM backend support."""

from __future__ import annotations

import os
from dotenv import load_dotenv

load_dotenv()


class Config:
    """All settings read from environment variables with sensible defaults."""

    # Backend selection
    LLM_BACKEND:          str   = os.getenv("LLM_BACKEND", "ollama")

    # Hugging Face backend
    HF_MODEL_NAME:        str   = os.getenv("HF_MODEL_NAME", "Qwen/Qwen2-7B-Instruct")
    HF_DEVICE:            str   = os.getenv("HF_DEVICE", "auto")
    HF_MAX_NEW_TOKENS:    int   = int(os.getenv("HF_MAX_NEW_TOKENS", 2048))

    # Ollama backend
    OLLAMA_HOST:          str   = os.getenv("OLLAMA_HOST", "http://localhost:11434")
    OLLAMA_MODEL:         str   = os.getenv("OLLAMA_MODEL", "llama2")

    # GGUF backend (llama-cpp-python)
    GGUF_MODEL_PATH:      str   = os.getenv("GGUF_MODEL_PATH", "")
    GGUF_N_CTX:           int   = int(os.getenv("GGUF_N_CTX", 4096))
    GGUF_N_GPU_LAYERS:    int   = int(os.getenv("GGUF_N_GPU_LAYERS", -1))

    # LM Studio backend
    LMSTUDIO_URL:         str   = os.getenv("LMSTUDIO_URL", "http://localhost:1234")
    LMSTUDIO_MODEL:       str   = os.getenv("LMSTUDIO_MODEL", "")

    # Embedding model
    EMBED_MODEL:          str   = os.getenv("EMBED_MODEL", "intfloat/multilingual-e5-large")

    # Index & data
    FAISS_INDEX:          str   = os.getenv("FAISS_INDEX", "QModel.index")
    METADATA_FILE:        str   = os.getenv("METADATA_FILE", "metadata.json")

    # Retrieval
    TOP_K_SEARCH:         int   = int(os.getenv("TOP_K_SEARCH", 20))
    TOP_K_RETURN:         int   = int(os.getenv("TOP_K_RETURN", 5))

    # Generation
    TEMPERATURE:          float = float(os.getenv("TEMPERATURE", 0.2))
    MAX_TOKENS:           int   = int(os.getenv("MAX_TOKENS", 2048))

    # Caching
    CACHE_SIZE:           int   = int(os.getenv("CACHE_SIZE", 512))
    CACHE_TTL:            int   = int(os.getenv("CACHE_TTL", 3600))

    # Ranking
    RERANK_ALPHA:         float = float(os.getenv("RERANK_ALPHA", 0.6))
    HADITH_BOOST:         float = float(os.getenv("HADITH_BOOST", 0.08))

    # Safety
    CONFIDENCE_THRESHOLD: float = float(os.getenv("CONFIDENCE_THRESHOLD", 0.30))

    # CORS
    ALLOWED_ORIGINS:      str   = os.getenv("ALLOWED_ORIGINS", "*")

    MAX_EXAMPLES:         int   = int(os.getenv("MAX_EXAMPLES", 3))


cfg = Config()