from pathlib import Path

from dotenv import load_dotenv

# Carica SEMPRE il .env del progetto e sovrascrive eventuali variabili già presenti

ENV_PATH = Path(__file__).resolve().parent / ".env"

load_dotenv(dotenv_path=ENV_PATH, override=True)

import argparse

import streamlit.web.cli as stcli

import sys

from scripts.evaluation.main import evaluation

from utils.config import Config

def parse_args() -> argparse.Namespace:

"""

Parse command-line arguments for the DQL application.

Provides options for configuring LLM providers, API credentials,

database connection strings, and processing behavior.

Returns:

argparse.Namespace: Object containing the parsed and validated arguments.

"""

parser = argparse.ArgumentParser(

description="DQL - Data Query Language",

formatter_class=argparse.RawTextHelpFormatter,

)

# Authentication argument for the AI model

parser.add_argument(

"-api",

action="store",

dest="api_key",

required=False,

help=(

"API key for the Gemini model.\n"

"If not specified, it will be read from '.env'."

),

)

# Persistence layer connection string

parser.add_argument(

"-uri_db",

action="store",

dest="uri_db",

required=False,

help=(

"MongoDB connection URI.\n"

"If not specified, it will be read from '.env'."

)

)

# Throttle control for API rate limits

parser.add_argument(

"-wait_seconds",

action="store",

dest="seconds",

type=int,

required=False,

default=0,

help="Number of seconds to wait after each LLM call (default: 0).",

)

# Optimization flag to reduce LLM overhead

parser.add_argument(

"-parsers",

action="store_true",

dest="parsers",

help=(

"Enable to avoid llm when possible"

),

)

# Specific model selection

parser.add_argument(

"-model_name",

action="store",

dest="model_name",

required=False,

default="gemini-2.5-flash",

help=(

"Specify the LLM model name.\n"

"Examples:\n"

" gemini-2.5-flash (default)\n"

" gpt-4o-mini\n"

" mistralai/Mistral-7B-Instruct-v0.2\n"

" claude-3-5-sonnet\n"

),

)

# AI engine provider selection

parser.add_argument(

"-provider",

action="store",

dest="provider",

required=False,

default="google_genai",

choices=["google_genai", "openai", "copilot", "huggingface"],

help=(

"Specify the LLM provider (default: google_genai).\n"

"Available options:\n"

" google_genai → Google Gemini\n"

" openai → OpenAI GPT models\n"

" copilot → GitHub Copilot API\n"

" huggingface → Hugging Face models\n"

),

)

# Evaluation Script

parser.add_argument(

"-evaluation_mode",

action="store_true",

dest="evaluation_mode",

help=(

"The streamlit application is not applied, but the 'evaluation' script."

),

)

# Optional future flags (commented for now)

# parser.add_argument(

# "-rag",

# action="store_true",

# dest="rag",

# help=(

# "Enable Retrieval-Augmented Generation (RAG) mode. "

# "Retrieves either entire documents or relevant chunks."

# ),

# )

#

# parser.add_argument(

# "-max_iterations",

# action="store",

# dest="max_iterations",

# type=int,

# required=False,

# help="Maximum number of query rewrite attempts.",

# )

#

# parser.add_argument(

# "-minimum_score",

# action="store",

# dest="minimum_score",

# type=float,

# required=False,

# help="Minimum rewrite score required to complete the process.",

# )

return parser.parse_args()

def _launch_streamlit() -> None:

"""

Launch the Streamlit user interface for DQL within the same process.

This helper overrides the system arguments to simulate a direct call

to the 'streamlit run' command pointing to the application's entry point.

"""

# Modify sys.argv to trigger Streamlit's CLI runner pointing to app.py

sys.argv = [

"streamlit",

"run",

"gui/app.py",

"--server.fileWatcherType=none", # Optimized performance by disabling file watcher

]

# Execute the Streamlit entry point and exit the parent process

sys.exit(stcli.main())

def main() -> None:

"""

Entry point for the DQL CLI application.

Workflow:

1. Capture runtime arguments from the shell.

2. Initialize the thread-safe Config Singleton with the parsed options.

3. Transition from the CLI environment to the Streamlit Web UI.

"""

# Parse CLI options

opts = parse_args()

# Bootstrap the configuration singleton before starting the UI

Config.get_instance(opts)

if opts:

if opts.evaluation_mode:

evaluation()

else:

# Hand off execution to the web interface

_launch_streamlit()

if __name__ == "__main__":

# Start the application lifecycle

main()