tests/test_tokenize.py

import os
os.environ["MOCK_INFERENCE"] = "true"

from fastapi.testclient import TestClient
from app import app

client = TestClient(app)

def test_tokenize():
    payload = {
        "model": "plain-llama-3-8b",
        "text": "This is a test sentence."
    }
    
    response = client.post("/v1/tokenize", json=payload)
    assert response.status_code == 200
    json_data = response.json()
    assert json_data["model"] == "plain-llama-3-8b"
    assert "token_count" in json_data
    assert json_data["token_count"] > 0