import unittest

from pipecat.frames.frames import LLMMessagesAppendFrame, UserImageRawFrame
from services.pipecat.pipeline import _multimodal_user_input_frame
from services.pipecat.processors import UserInputError, parse_user_input


class UserInputParserTests(unittest.TestCase):
    def test_parses_text_and_current_camera_frame(self):
        value = parse_user_input(
            {
                "type": "user-input",
                "schema_version": 1,
                "input_id": "input_1",
                "parts": [
                    {"type": "input_text", "text": "帮我看看"},
                    {
                        "type": "input_image",
                        "source": {
                            "type": "camera_frame",
                            "frame": "current",
                        },
                    },
                ],
                "options": {
                    "run_immediately": True,
                    "interrupt": True,
                },
            }
        )

        self.assertIsNotNone(value)
        self.assertEqual(value.text, "帮我看看")
        self.assertTrue(value.has_camera_frame)
        self.assertEqual(value.transcript_text, "帮我看看\n已发送一张图片")

    def test_rejects_legacy_and_unsupported_image_sources(self):
        self.assertIsNone(parse_user_input({"type": "user-text", "text": "旧协议"}))
        with self.assertRaisesRegex(UserInputError, "当前摄像头"):
            parse_user_input(
                {
                    "type": "user-input",
                    "schema_version": 1,
                    "input_id": "input_2",
                    "parts": [
                        {
                            "type": "input_image",
                            "source": {
                                "type": "uploaded_asset",
                                "asset_id": "asset_1",
                            },
                        }
                    ],
                }
            )

    def test_native_image_uses_the_standard_multimodal_user_turn_path(self):
        image = UserImageRawFrame(
            image=bytes([220, 40, 40] * 16 * 16),
            size=(16, 16),
            format="RGB",
        )

        frame = _multimodal_user_input_frame(
            image,
            "请根据用户刚提交的图片进行回复。",
        )

        self.assertIsInstance(frame, LLMMessagesAppendFrame)
        self.assertTrue(frame.run_llm)
        self.assertEqual(frame.messages[0]["role"], "user")
        content = frame.messages[0]["content"]
        self.assertEqual(
            content[0],
            {"type": "text", "text": "请根据用户刚提交的图片进行回复。"},
        )
        self.assertEqual(content[1]["type"], "image_url")
        self.assertTrue(
            content[1]["image_url"]["url"].startswith(
                "data:image/jpeg;base64,"
            )
        )


if __name__ == "__main__":
    unittest.main()