inference.py

import openai
from transformers import BertTokenizer
from evo_model import EvoTransformerForClassification
import torch

# Load Evo model
model = EvoTransformerForClassification.from_pretrained("trained_model")
model.eval()

# Tokenizer (BERT-compatible)
tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")

# Set OpenAI key (assumes you have it set as ENV VAR or replace directly)
import os
openai.api_key = os.getenv("OPENAI_API_KEY")

def query_gpt35(prompt):
    try:
        response = openai.ChatCompletion.create(
            model="gpt-3.5-turbo",
            messages=[{"role": "user", "content": prompt}],
            max_tokens=50,
            temperature=0.3,
        )
        return response['choices'][0]['message']['content'].strip()
    except Exception as e:
        return f"[GPT-3.5 Error] {e}"

def generate_response(goal, option1, option2):
    # Evo prediction
    prompt = f"Goal: {goal}\nOption 1: {option1}\nOption 2: {option2}\nWhich is better?"
    inputs = tokenizer([goal + " " + option1, goal + " " + option2],
                       return_tensors="pt", padding=True, truncation=True)
    with torch.no_grad():
        logits = model(**inputs)
    pred = torch.argmax(logits, dim=1).item()
    evo_result = option1 if pred == 0 else option2

    # GPT-3.5 prediction
    gpt_result = query_gpt35(prompt)

    return {
        "evo_suggestion": evo_result,
        "gpt_suggestion": gpt_result
    }