ollama_chatbotTTS.py

import ollama
import json
from pydantic import BaseModel

class ChatResponse(BaseModel):
    response: str

class OllamaChat:
    response: str

    def __init__(self):
        """Initialize the Ollama model and pull it locally if needed."""
        # Modelo que queremos usar
        self.model = "hf.co/jnjj/vcvcvcv:Q4_0"
        self.system_prompt = (
            "You are a concise and natural-sounding assistant. "
            "Answer questions briefly, in one or two sentences at most, "
            "as if responding for text-to-speech (TTS). Keep it natural and conversational."
        )

        # Intentamos descargar (pull) el modelo
        try:
            print(f"Descargando el modelo {self.model}…")
            ollama.pull(model=self.model)
            print("Modelo descargado correctamente.")
        except Exception as e:
            print(f"No se pudo descargar el modelo: {e}")

    def get_response(self, user_input: str) -> str:
        """Processes the input text using Ollama and returns only the response string."""
        result = ollama.chat(
            model=self.model,
            messages=[
                {"role": "system", "content": self.system_prompt},
                {"role": "user", "content": user_input}
            ],
            format=ChatResponse.model_json_schema()
        )
        # Extraemos el campo "response" del JSON devuelto
        response = json.loads(result["message"]["content"])["response"]
        return response


# Ejemplo de uso
if __name__ == "__main__":
    ollama_chat = OllamaChat()
    user_text = input("Enter your question: ")
    print("\nOllama Response:\n", ollama_chat.get_response(user_text))