config.py

import os
from pathlib import Path

# Get the current directory
BASE_DIR = Path(__file__).parent.absolute()

class Config:
    # Data paths (relative to project directory)
    DATA_DIR = BASE_DIR / "data"
    VECTORSTORE_DIR = BASE_DIR / "vectorstore"
    MODELS_DIR = BASE_DIR / "models"
    
    # Ensure directories exist
    DATA_DIR.mkdir(exist_ok=True)
    VECTORSTORE_DIR.mkdir(exist_ok=True)
    MODELS_DIR.mkdir(exist_ok=True)
    
    # Model configurations
    EMBEDDING_MODEL = "sentence-transformers/all-MiniLM-L6-v2"
    
    # Free LLM options (choose one)
    LLM_PROVIDERS = {
        "flan_t5_small": {
            "model_name": "google/flan-t5-small",
            "task": "text2text-generation",
            "max_new_tokens": 256,
            "temperature": 0.3,
            "do_sample": True
        },
        "flan_t5_base": {
            "model_name": "google/flan-t5-base", 
            "task": "text2text-generation",
            "max_new_tokens": 512,
            "temperature": 0.3,
            "do_sample": True
        },
        "local_llm": {
            "model_name": "microsoft/DialoGPT-small",  # Fallback option
            "task": "text-generation",
            "max_new_tokens": 128,
            "temperature": 0.6,
            "do_sample": True
        },
        "simple_qa": {
            "model_name": "distilbert-base-uncased-distilled-squad",
            "task": "question-answering",
            "max_new_tokens": 256,
            "temperature": 0.3
        }
    }
    
    # Default LLM provider (FLAN-T5 is better for instruction following)
    DEFAULT_LLM_PROVIDER = "flan_t5_small"
    
    # Text processing
    CHUNK_SIZE = 500
    CHUNK_OVERLAP = 50
    
    # Retrieval settings
    RETRIEVAL_K = 3  # Number of documents to retrieve
    
    # Flask settings
    FLASK_HOST = "0.0.0.0"
    FLASK_PORT = 8080
    FLASK_DEBUG = True