app_config.py

#!/usr/bin/env python3
"""
Embedded Configuration for LinguaCustodia API
Fallback configuration when clean architecture imports fail.
Updated for LinguaCustodia Pro Finance Suite models.
"""

import os
import torch
import gc
import logging
from pydantic import BaseModel, Field, field_validator, ConfigDict
from pydantic_settings import BaseSettings
from typing import Dict, List, Optional, Any, Literal
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
from huggingface_hub import login

logger = logging.getLogger(__name__)

# Model type definition for Pro Finance Suite
ModelType = Literal[
    "pro-finance-large", "pro-finance-medium", "pro-finance-small", 
    "pro-finance-mini", "llama-pro-finance-mini", "fin-pythia-1.4b"
]

class TokenizerConfig(BaseModel):
    """Tokenizer configuration for LinguaCustodia models."""
    eos_token: str = Field(..., description="End of sequence token")
    bos_token: Optional[str] = Field(None, description="Beginning of sequence token")
    pad_token: Optional[str] = Field(None, description="Padding token")
    unk_token: Optional[str] = Field(None, description="Unknown token")
    eos_token_id: int = Field(..., description="EOS token ID")
    bos_token_id: Optional[int] = Field(None, description="BOS token ID")
    pad_token_id: Optional[int] = Field(None, description="Pad token ID")
    vocab_size: int = Field(..., description="Vocabulary size")
    model_max_length: int = Field(131072, description="Maximum sequence length")

class GenerationConfig(BaseModel):
    """Generation configuration for LinguaCustodia models."""
    eos_tokens: List[int] = Field(..., description="List of EOS token IDs")
    bos_token_id: Optional[int] = Field(None, description="BOS token ID")
    temperature: float = Field(0.6, description="Sampling temperature")
    top_p: float = Field(0.9, description="Top-p sampling parameter")
    max_new_tokens: int = Field(150, description="Maximum new tokens to generate")
    repetition_penalty: float = Field(1.05, description="Repetition penalty")
    no_repeat_ngram_size: int = Field(2, description="No repeat n-gram size")
    early_stopping: bool = Field(False, description="Enable early stopping")
    min_length: int = Field(50, description="Minimum response length")

class ModelInfo(BaseModel):
    """Model information for LinguaCustodia models."""
    model_id: str = Field(..., description="HuggingFace model identifier")
    display_name: str = Field(..., description="Human-readable model name")
    architecture: str = Field(..., description="Model architecture class")
    parameters: str = Field(..., description="Model parameter count")
    memory_gb: int = Field(..., description="Required RAM in GB")
    vram_gb: int = Field(..., description="Required VRAM in GB")
    tokenizer: TokenizerConfig = Field(..., description="Tokenizer configuration")
    generation: GenerationConfig = Field(..., description="Generation configuration")

class AppSettings(BaseSettings):
    """Application settings."""
    model_name: ModelType = Field(default="pro-finance-small", description="Model to load")
    hf_token_lc: Optional[str] = Field(default=None, description="HuggingFace token for LinguaCustodia")
    hf_token: Optional[str] = Field(default=None, description="HuggingFace token")
    app_port: int = Field(default=7860, description="Application port")
    
    model_config = ConfigDict(
        env_file=".env",
        env_file_encoding="utf-8",
        case_sensitive=False,
        extra="ignore"
    )
    
    @field_validator('model_name')
    @classmethod
    def validate_model_name(cls, v):
        valid_models = [
            "pro-finance-large", "pro-finance-medium", "pro-finance-small", 
            "pro-finance-mini", "llama-pro-finance-mini", "fin-pythia-1.4b"
        ]
        if v not in valid_models:
            raise ValueError(f'Model name must be one of: {valid_models}')
        return v

# LinguaCustodia Pro Finance Suite model configurations
LINGUACUSTODIA_MODELS = {
    "pro-finance-large": ModelInfo(
        model_id="LinguaCustodia/Llama-Pro-Finance-Large",
        display_name="Llama Pro Finance Large",
        architecture="LlamaForCausalLM",
        parameters="70B",
        memory_gb=140,
        vram_gb=80,
        tokenizer=TokenizerConfig(
            eos_token="<|eot_id|>",
            bos_token="<|begin_of_text|>",
            pad_token="<|eot_id|>",
            unk_token=None,
            eos_token_id=128009,
            bos_token_id=128000,
            pad_token_id=128009,
            vocab_size=128000,
            model_max_length=131072
        ),
        generation=GenerationConfig(
            eos_tokens=[128001, 128008, 128009],
            bos_token_id=128000
        )
    ),
    "pro-finance-medium": ModelInfo(
        model_id="LinguaCustodia/LLM-Pro-Finance-Medium",
        display_name="LLM Pro Finance Medium",
        architecture="LlamaForCausalLM",
        parameters="32B",
        memory_gb=64,
        vram_gb=32,
        tokenizer=TokenizerConfig(
            eos_token="<|eot_id|>",
            bos_token="<|begin_of_text|>",
            pad_token="<|eot_id|>",
            unk_token=None,
            eos_token_id=128009,
            bos_token_id=128000,
            pad_token_id=128009,
            vocab_size=128000,
            model_max_length=131072
        ),
        generation=GenerationConfig(
            eos_tokens=[128001, 128008, 128009],
            bos_token_id=128000
        )
    ),
    "pro-finance-small": ModelInfo(
        model_id="LinguaCustodia/LLM-Pro-Finance-Small",
        display_name="LLM Pro Finance Small",
        architecture="LlamaForCausalLM",
        parameters="8B",
        memory_gb=16,
        vram_gb=8,
        tokenizer=TokenizerConfig(
            eos_token="<|eot_id|>",
            bos_token="<|begin_of_text|>",
            pad_token="<|eot_id|>",
            unk_token=None,
            eos_token_id=128009,
            bos_token_id=128000,
            pad_token_id=128009,
            vocab_size=128000,
            model_max_length=131072
        ),
        generation=GenerationConfig(
            eos_tokens=[128001, 128008, 128009],
            bos_token_id=128000
        )
    ),
    "pro-finance-mini": ModelInfo(
        model_id="LinguaCustodia/LLM-Pro-Finance-Mini",
        display_name="LLM Pro Finance Mini",
        architecture="LlamaForCausalLM",
        parameters="3B",
        memory_gb=6,
        vram_gb=3,
        tokenizer=TokenizerConfig(
            eos_token="<|eot_id|>",
            bos_token="<|begin_of_text|>",
            pad_token="<|eot_id|>",
            unk_token=None,
            eos_token_id=128009,
            bos_token_id=128000,
            pad_token_id=128009,
            vocab_size=128000,
            model_max_length=131072
        ),
        generation=GenerationConfig(
            eos_tokens=[128001, 128008, 128009],
            bos_token_id=128000
        )
    ),
    "llama-pro-finance-mini": ModelInfo(
        model_id="LinguaCustodia/Llama-Pro-Finance-Mini",
        display_name="Llama Pro Finance Mini",
        architecture="LlamaForCausalLM", 
        parameters="1B",
        memory_gb=3,
        vram_gb=2,
        tokenizer=TokenizerConfig(
            eos_token="<|eot_id|>",
            bos_token="<|begin_of_text|>",
            pad_token="<|eot_id|>",
            unk_token=None,
            eos_token_id=128009,
            bos_token_id=128000,
            pad_token_id=128009,
            vocab_size=128000,
            model_max_length=131072
        ),
        generation=GenerationConfig(
            eos_tokens=[128001, 128008, 128009],
            bos_token_id=128000
        )
    ),
    "fin-pythia-1.4b": ModelInfo(
        model_id="LinguaCustodia/fin-pythia-1.4b",
        display_name="Fin-Pythia 1.4B Financial",
        architecture="GPTNeoXForCausalLM",
        parameters="1.4B",
        memory_gb=3,
        vram_gb=2,
        tokenizer=TokenizerConfig(
            eos_token="<|endoftext|>",
            bos_token="<|endoftext|>",
            pad_token=None,
            unk_token="<|endoftext|>",
            eos_token_id=0,
            bos_token_id=0,
            pad_token_id=None,
            vocab_size=50304,
            model_max_length=2048
        ),
        generation=GenerationConfig(
            eos_tokens=[0],
            bos_token_id=0
        )
    )
}

# Default model configuration
DEFAULT_MODEL = "pro-finance-small"

def get_model_config(model_name: str) -> ModelInfo:
    """Get model configuration by name."""
    if model_name not in LINGUACUSTODIA_MODELS:
        raise ValueError(f"Model '{model_name}' not found. Available models: {list(LINGUACUSTODIA_MODELS.keys())}")
    return LINGUACUSTODIA_MODELS[model_name]

def get_app_settings() -> AppSettings:
    """Get application settings."""
    return AppSettings()

def authenticate_huggingface(token: str) -> bool:
    """Authenticate with HuggingFace."""
    try:
        login(token=token, add_to_git_credential=False)
        logger.info("✅ Successfully authenticated with HuggingFace")
        return True
    except Exception as e:
        logger.error(f"❌ HuggingFace authentication failed: {e}")
        return False

def setup_gpu_environment():
    """Setup GPU environment."""
    if torch.cuda.is_available():
        torch.cuda.empty_cache()
        gc.collect()
        logger.info(f"🚀 GPU available: {torch.cuda.get_device_name(0)}")
        logger.info(f"📊 GPU memory: {torch.cuda.get_device_properties(0).total_memory / 1e9:.1f}GB")
        return True
    else:
        logger.warning("⚠️ No GPU available, using CPU")
        return False

def load_model_and_tokenizer(model_info: ModelInfo, use_auth_token: Optional[str] = None):
    """Load model and tokenizer with proper configuration."""
    try:
        logger.info(f"🔄 Loading model: {model_info.model_id}")
        
        # Load tokenizer
        tokenizer = AutoTokenizer.from_pretrained(
            model_info.model_id,
            token=use_auth_token,
            trust_remote_code=True
        )
        
        # Configure special tokens
        if tokenizer.pad_token is None:
            tokenizer.pad_token = tokenizer.eos_token
            
        # Load model
        model = AutoModelForCausalLM.from_pretrained(
            model_info.model_id,
            token=use_auth_token,
            trust_remote_code=True,
            torch_dtype=torch.bfloat16,
            device_map="auto"
        )
        
        logger.info(f"✅ Model loaded successfully: {model_info.display_name}")
        return model, tokenizer
        
    except Exception as e:
        logger.error(f"❌ Failed to load model {model_info.model_id}: {e}")
        raise

def create_pipeline(model, tokenizer, model_info: ModelInfo):
    """Create inference pipeline."""
    try:
        pipe = pipeline(
            "text-generation",
            model=model,
            tokenizer=tokenizer,
            torch_dtype=torch.bfloat16,
            device_map="auto",
            **model_info.generation.model_dump()
        )
        logger.info("✅ Pipeline created successfully")
        return pipe
    except Exception as e:
        logger.error(f"❌ Failed to create pipeline: {e}")
        raise