config/settings.py

"""Application settings using pydantic-settings."""
from pydantic_settings import BaseSettings, SettingsConfigDict
from typing import Optional, Dict


class Settings(BaseSettings):
    """Application settings loaded from environment variables."""
    
    # API Keys
    openai_api_key: str
    mistralai_api_key: str
    
    # JWT Security
    jwt_secret_key: str
    jwt_algorithm: str = "HS256"
    jwt_expiration_minutes: int = 60
    
    # API Config
    api_title: str = "CAPL Routeur IA API"
    api_version: str = "1.0.0"
    environment: str = "development"
    
    # LangSmith (optional)
    langchain_tracing_v2: bool = False
    langchain_api_key: Optional[str] = None
    langchain_project: str = "routeur-ia"

    # Supabase / RAG
    supabase_url: Optional[str] = None
    supabase_key: Optional[str] = None
    supabase_table: str = "documents"
    supabase_match_fn: str = "match_documents"
    rag_top_k: int = 5

    # Multi-index (logical index name -> Supabase table + match function)
    # Can be overridden via .env using a JSON structure if desired
    vector_indexes: Dict[str, Dict[str, str]] = {
        "default": {"table": "documents", "query_name": "match_documents"},
        # Example additional indexes (create corresponding table and function in Supabase)
        "projects": {"table": "project_documents", "query_name": "match_project_documents"},
        # "formations": {"table": "formations", "query_name": "match_formations"},
    }
    
    # Upload & chunking (PDF ingestion)
    max_upload_mb_pdf: int = 50
    chunk_size: int = 1000
    chunk_overlap: int = 100
    doc_default_type: str = "project_doc"
    
    # Post-processing pipeline configuration
    postprocessors_enabled: list[str] = [
        "carbon_impact",
        "pricing",
        "equivalences",
    ]
    
    currency: str = "USD"

    # Pricing per 1,000,000 tokens (input/output)
    pricing: dict = {
        # Complete or override in .env via nested JSON if needed
        "mistral-medium-latest": {"input_per_1m": 0.40, "output_per_1m": 2.00},
        "mistral-small-latest": {"input_per_1m": 0.10, "output_per_1m": 0.30},
        "mistral-large-latest": {"input_per_1m": 2.00, "output_per_1m": 6.00},
        "magistral-small-latest": {"input_per_1m": 0.50, "output_per_1m": 1.50},
        "magistral-medium-latest": {"input_per_1m": 2.00, "output_per_1m": 5.00},
    }

    # Equivalence ratios using kgCO2eq as input
    equivalence_ratios: dict = {
        # https://impactco2.fr/outils/comparateur
        # Ratios expressed as UNITS per kgCO2eq (invert of kgCO2 per unit)
        # smartphone: 85.9 kgCO2 per unit → 1 / 85.9 ≈ 0.011643 smartphones per kgCO2
        "smartphone_per_kgCO2eq": 0.011643,
        # car (km): 219 g CO2 per km → 0.219 kg per km → 1 / 0.219 ≈ 4.566210 km per kgCO2
        "car_km_per_kgCO2eq": 4.566210,
        # tgv (km): 2.93 g CO2 per km → 0.00293 kg per km → 1 / 0.00293 ≈ 341.296928 km per kgCO2
        "tgv_km_per_kgCO2eq": 341.296928,
        # water (l): 321 g CO2 per liter → 0.321 kg per liter → 1 / 0.321 ≈ 3.115265 liters per kgCO2
        "water_l_per_kgCO2eq": 3.115265,
    }
    
    # Voice / Pipecat
    voice_enabled: bool = True
    voice_stt_model: str = "whisper-1"
    voice_stt_language: str = "fr"
    voice_tts_voice: str = "alloy"
    voice_tts_model: str = "tts-1"
    voice_default_model: str = "mistral-large-latest"
    voice_max_response_length: int = 500

    # VAD (Voice Activity Detection) defaults - overridable via client request_data
    voice_vad_stop_secs: float = 0.2
    voice_vad_start_secs: float = 0.2
    voice_vad_confidence: float = 0.7
    voice_vad_min_volume: float = 0.6

    # Voice / WebRTC ICE servers (TURN/STUN for NAT traversal)
    # Option 1: Static JSON array in VOICE_ICE_SERVERS (e.g. from Metered dashboard)
    voice_ice_servers: Optional[str] = None
    # Option 2: Twilio TURN - often works on HF Spaces (streamlit-webrtc examples)
    twilio_account_sid: Optional[str] = None
    twilio_auth_token: Optional[str] = None
    # Option 3: Metered API - fetch dynamically at startup
    metered_app_name: Optional[str] = None
    metered_api_key: Optional[str] = None

    # Voice / Daily.co (alternative transport for HF Spaces, NAT traversal)
    daily_api_key: Optional[str] = None

    model_config = SettingsConfigDict(
        env_file=".env",
        env_file_encoding="utf-8",
        case_sensitive=False,
        extra="ignore"
    )


# Singleton instance
settings = Settings()