from huggingface_hub import InferenceClient
import os

# HF automatically injects token
client = InferenceClient(
    token=os.getenv("HF_TOKEN")
)

MODEL = "HuggingFaceH4/zephyr-7b-beta"


def call_llm(prompt):

    response = client.text_generation(
        model=MODEL,
        prompt=prompt,
        max_new_tokens=400,
        temperature=0.7
    )

    return response


def market_agent(problem, memory, prompt):
    return call_llm(prompt.format(problem=problem, memory=memory))


def finance_agent(problem, memory, prompt):
    return call_llm(prompt.format(problem=problem, memory=memory))


def risk_agent(problem, memory, prompt):
    return call_llm(prompt.format(problem=problem, memory=memory))


def ethics_agent(problem, memory, prompt):
    return call_llm(prompt.format(problem=problem, memory=memory))


def synthesis_agent(problem, memory, prompt):
    return call_llm(prompt.format(problem=problem, memory=memory))