tests/test_vision_hf_only.py

import asyncio
import base64
import sys
from pathlib import Path

import pytest
from PIL import Image

sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
from api import vision


class FakeResponse:
    def __init__(self, status_code=200, payload=None, content=b"png-bytes", text=""):
        self.status_code = status_code
        self._payload = payload
        self.content = content
        self.text = text

    def json(self):
        if isinstance(self._payload, Exception):
            raise self._payload
        return self._payload


class FakeClient:
    calls = []
    responses = []

    def __init__(self, *args, **kwargs):
        self.calls = []

    async def __aenter__(self):
        FakeClient.active = self
        return self

    async def __aexit__(self, *args):
        return False

    async def post(self, url, **kwargs):
        self.calls.append((url, kwargs))
        FakeClient.calls.append((url, kwargs))
        return FakeClient.responses.pop(0)


def test_analyze_uses_hf_vqa_without_openai(monkeypatch):
    FakeClient.calls = []
    FakeClient.responses = [FakeResponse(payload=[{"answer": "un gatto"}])]
    monkeypatch.setattr(vision.httpx, "AsyncClient", FakeClient)
    monkeypatch.delenv("OPENAI_API_KEY", raising=False)
    monkeypatch.delenv("OPENAI_BASE_URL", raising=False)
    monkeypatch.delenv("GROQ_API_KEY", raising=False)
    monkeypatch.delenv("GEMINI_API_KEY", raising=False)

    result = asyncio.run(vision.analyze_image(
        vision.AnalyzeImageRequest(base64_image=base64.b64encode(b"image").decode(), question="Cosa vedi?")
    ))

    assert result == {"ok": True, "description": "un gatto", "provider": "blip-vqa"}
    assert len(FakeClient.calls) == 1
    assert vision._HF_VQA_MODEL in FakeClient.calls[0][0]
    assert all("openai.com" not in call[0] for call in FakeClient.calls)


def test_generate_uses_huggingface(monkeypatch):
    class FakeInferenceClient:
        def __init__(self, **kwargs):
            self.kwargs = kwargs

        def text_to_image(self, **kwargs):
            assert kwargs["model"] == "stabilityai/stable-diffusion-3-medium-diffusers"
            return Image.new("RGB", (1, 1), (0, 120, 255))

    monkeypatch.setattr(vision, "InferenceClient", FakeInferenceClient)

    result = asyncio.run(vision.generate_image(vision.GenerateImageRequest(prompt="un paesaggio")))

    assert result["ok"] is True
    assert result["mime"] == "image/png"
    assert len(base64.b64decode(result["image_b64"])) > 0


def test_edit_uses_inference_client_hf(monkeypatch):
    class FakeInferenceClient:
        def __init__(self, **kwargs):
            self.kwargs = kwargs

        def image_to_image(self, **kwargs):
            assert kwargs["model"] == "black-forest-labs/FLUX.1-Kontext-dev"
            return Image.new("RGB", (1, 1), (0, 120, 255))

    monkeypatch.setattr(vision, "InferenceClient", FakeInferenceClient)

    result = asyncio.run(vision.edit_image(
        vision.EditImageRequest(prompt="rendi il cielo blu", base64_image="aW1hZ2U=")
    ))

    assert result["ok"] is True
    assert result["model"] == "FLUX.1-Kontext-dev"
    assert result["mime"] == "image/png"
    assert len(base64.b64decode(result["image_b64"])) > 0