hf_model.py

"""
HuggingFace Inference API Model Wrapper
Uses Gemma 3 27B for fintech orchestrator
"""

import os
from huggingface_hub import InferenceClient

# Initialize client
HF_TOKEN = os.getenv("HF_TOKEN")
MODEL_ID = "google/gemma-3-4b-it"  # Gemma 3 27B Instruct

client = InferenceClient(token=HF_TOKEN)


def generate_response(
    messages: list[dict],
    max_tokens: int = 1024,
    temperature: float = 0.7,
) -> str:
    """
    Generate response using HuggingFace Inference API.
    
    Args:
        messages: List of message dicts with 'role' and 'content'
        max_tokens: Maximum tokens to generate
        temperature: Sampling temperature
    
    Returns:
        Generated text response
    """
    try:
        response = client.chat.completions.create(
            model=MODEL_ID,
            messages=messages,
            max_tokens=max_tokens,
            temperature=temperature,
        )
        return response.choices[0].message.content
    except Exception as e:
        return f"Error: {str(e)}"


def calculate_expression(expression: str) -> str:
    """Simple calculator for financial expressions."""
    import re
    import math
    
    # Safe eval with limited functions
    allowed_names = {
        'abs': abs, 'round': round, 'min': min, 'max': max,
        'pow': pow, 'sqrt': math.sqrt, 'log': math.log,
        'exp': math.exp, 'pi': math.pi, 'e': math.e,
    }
    
    try:
        # Clean the expression
        expr = expression.strip()
        # Basic validation
        if not re.match(r'^[\d\s\+\-\*\/\.\(\)\^]+$', expr.replace('**', '^')):
            # Try to extract numbers and operators more flexibly
            pass
        
        result = eval(expr, {"__builtins__": {}}, allowed_names)
        return f"{result:,.2f}"
    except Exception as e:
        return f"Calculation error: {str(e)}"