rasahq--rasa/tests/core/test_nlg.py

import uuid
from typing import Text, Any

import jsonschema
import pytest
from sanic import Sanic, response

from rasa.core.nlg.callback import CallbackNaturalLanguageGenerator
from rasa.core.nlg.response import TemplatedNaturalLanguageGenerator
from rasa.utils.endpoints import EndpointConfig, read_endpoint_config
from rasa.core.agent import Agent


def nlg_app(base_url="/"):

    app = Sanic("test_nlg")

    @app.route(base_url, methods=["POST"])
    async def generate(request):
        """Simple HTTP NLG generator, checks that the incoming request
        is format according to the spec."""

        nlg_request_format_spec = {
            "type": "object",
            "properties": {
                "response": {"type": "string"},
                "id": {"type": ["string", "null"]},
                "arguments": {"type": "object"},
                "tracker": {
                    "type": "object",
                    "properties": {
                        "sender_id": {"type": "string"},
                        "slots": {"type": "object"},
                        "latest_message": {"type": "object"},
                        "latest_event_time": {"type": "number"},
                        "paused": {"type": "boolean"},
                        "events": {"type": "array"},
                    },
                },
                "channel": {
                    "type": "object",
                    "properties": {"name": {"type": "string"}},
                },
            },
        }

        nlg_call = request.json

        jsonschema.validate(nlg_call, nlg_request_format_spec)

        if nlg_call.get("response") == "utter_greet":
            response_dict = {"text": "Hey there!"}
        else:
            response_dict = {"text": "Sorry, didn't get that."}
        return response.json(response_dict)

    return app


# noinspection PyShadowingNames
@pytest.fixture()
def http_nlg(loop, sanic_client):
    return loop.run_until_complete(sanic_client(nlg_app()))


async def test_nlg(http_nlg, trained_rasa_model: Text):
    sender = str(uuid.uuid1())

    nlg_endpoint = EndpointConfig.from_dict({"url": http_nlg.make_url("/")})
    agent = Agent.load(trained_rasa_model, generator=nlg_endpoint)

    response = await agent.handle_text("/greet", sender_id=sender)
    assert len(response) == 1
    assert response[0] == {"text": "Hey there!", "recipient_id": sender}


def test_nlg_endpoint_config_loading(endpoints_path: Text):
    cfg = read_endpoint_config(endpoints_path, "nlg")

    assert cfg == EndpointConfig.from_dict({"url": "http://localhost:5055/nlg"})


def test_nlg_schema_validation():
    content = {"text": "Hey there!"}
    assert CallbackNaturalLanguageGenerator.validate_response(content)


def test_nlg_schema_validation_empty_buttons():
    content = {"text": "Hey there!", "buttons": []}
    assert CallbackNaturalLanguageGenerator.validate_response(content)


def test_nlg_schema_validation_empty_image():
    content = {"text": "Hey there!", "image": None}
    assert CallbackNaturalLanguageGenerator.validate_response(content)


def test_nlg_schema_validation_empty_custom_dict():
    content = {"custom": {}}
    assert CallbackNaturalLanguageGenerator.validate_response(content)


@pytest.mark.parametrize(
    "slot_name, slot_value",
    [
        ("tag_w_underscore", "a"),
        ("tag with space", "bacon"),
        ("tag.with.dot", "chocolate"),
        ("tag-w-dash", "apple pie"),
        ("tag-w-$", "banana"),
        ("tag-w-@", "one"),
        ("tagCamelCase", "two"),
        ("tag-w-*", "three"),
        ("tag_w_underscore", "a"),
        ("tag.with.float.val", 1.3),
        ("tag-w-$", "banana"),
        ("tagCamelCase", "two"),
        ("empty_string", ""),
        ("null", None),
    ],
)
def test_nlg_fill_response_text(slot_name: Text, slot_value: Any):
    response = {"text": f"{{{slot_name}}}"}
    t = TemplatedNaturalLanguageGenerator(responses=dict())
    result = t._fill_response(response=response, filled_slots={slot_name: slot_value})
    assert result == {"text": str(slot_value)}


@pytest.mark.parametrize(
    "img_slot_name, img_slot_value",
    [("url", "https://www.exampleimg.com"), ("img1", "https://www.appleimg.com")],
)
def test_nlg_fill_response_image(img_slot_name: Text, img_slot_value: Text):
    response = {"image": f"{{{img_slot_name}}}"}
    t = TemplatedNaturalLanguageGenerator(responses=dict())
    result = t._fill_response(
        response=response, filled_slots={img_slot_name: img_slot_value}
    )
    assert result == {"image": str(img_slot_value)}


@pytest.mark.parametrize(
    "slot_name, slot_value",
    [
        ("tag_w_underscore", "a"),
        ("tag with space", "bacon"),
        ("tag.with.dot", "chocolate"),
        ("tag-w-dash", "apple pie"),
        ("tag-w-$", "banana"),
        ("tag-w-@", "one"),
        ("tagCamelCase", "two"),
        ("tag-w-*", "three"),
        ("tag_w_underscore", "a"),
        ("tag.with.float.val", 1.3),
        ("tag-w-$", "banana"),
        ("tagCamelCase", "two"),
        ("empty_string", ""),
        ("null", None),
    ],
)
def test_nlg_fill_response_custom(slot_name: Text, slot_value: Any):
    response = {
        "custom": {
            "field": f"{{{slot_name}}}",
            "properties": {"field_prefixed": f"prefix_{{{slot_name}}}"},
            "bool_field": True,
            "int_field:": 42,
            "empty_field": None,
        }
    }
    t = TemplatedNaturalLanguageGenerator(responses=dict())
    result = t._fill_response(response=response, filled_slots={slot_name: slot_value})

    assert result == {
        "custom": {
            "field": str(slot_value),
            "properties": {"field_prefixed": f"prefix_{slot_value}"},
            "bool_field": True,
            "int_field:": 42,
            "empty_field": None,
        }
    }


def test_nlg_fill_response_custom_with_list():
    response = {
        "custom": {
            "blocks": [{"fields": [{"text": "*Departure date:*\n{test}"}]}],
            "other": ["{test}"],
        }
    }
    t = TemplatedNaturalLanguageGenerator(responses=dict())
    result = t._fill_response(response=response, filled_slots={"test": 5})
    assert result == {
        "custom": {
            "blocks": [{"fields": [{"text": "*Departure date:*\n5"}]}],
            "other": ["5"],
        }
    }


@pytest.mark.parametrize(
    "response_text, expected",
    [
        ('{{"variable":"{slot_1}"}}', '{"variable":"foo"}'),
        ("{slot_1} and {slot_2}", "foo and bar"),
        ("{{{slot_1}, {slot_2}!}}", "{foo, bar!}"),
        ("{{{slot_1}}}", "{foo}"),
        ("{{slot_1}}", "{slot_1}"),
    ],
)
def test_nlg_fill_response_text_with_json(response_text, expected):
    response = {"text": response_text}
    t = TemplatedNaturalLanguageGenerator(responses=dict())
    result = t._fill_response(
        response=response, filled_slots={"slot_1": "foo", "slot_2": "bar"}
    )
    assert result == {"text": expected}


@pytest.mark.parametrize("slot_name, slot_value", [("tag_w_\n", "a")])
def test_nlg_fill_response_with_bad_slot_name(slot_name, slot_value):
    response_text = f"{{{slot_name}}}"
    t = TemplatedNaturalLanguageGenerator(responses=dict())
    result = t._fill_response(
        response={"text": response_text}, filled_slots={slot_name: slot_value}
    )
    assert result["text"] == response_text


@pytest.mark.parametrize(
    "text_slot_name, text_slot_value, img_slot_name, img_slot_value",
    [
        ("tag_w_underscore", "a", "url", "https://www.exampleimg.com"),
        ("tag with space", "bacon", "img1", "https://www.appleimg.com"),
    ],
)
def test_nlg_fill_response_image_and_text(
    text_slot_name, text_slot_value, img_slot_name, img_slot_value
):
    response = {"text": f"{{{text_slot_name}}}", "image": f"{{{img_slot_name}}}"}
    t = TemplatedNaturalLanguageGenerator(responses=dict())
    result = t._fill_response(
        response=response,
        filled_slots={text_slot_name: text_slot_value, img_slot_name: img_slot_value},
    )
    assert result == {"text": str(text_slot_value), "image": str(img_slot_value)}


@pytest.mark.parametrize(
    "text_slot_name, text_slot_value, cust_slot_name, cust_slot_value",
    [
        ("tag_w_underscore", "a", "tag.with.dot", "chocolate"),
        ("tag with space", "bacon", "tag-w-dash", "apple pie"),
    ],
)
def test_nlg_fill_response_text_and_custom(
    text_slot_name, text_slot_value, cust_slot_name, cust_slot_value
):
    response = {
        "text": f"{{{text_slot_name}}}",
        "custom": {
            "field": f"{{{cust_slot_name}}}",
            "properties": {"field_prefixed": f"prefix_{{{cust_slot_name}}}"},
        },
    }
    t = TemplatedNaturalLanguageGenerator(responses=dict())
    result = t._fill_response(
        response=response,
        filled_slots={text_slot_name: text_slot_value, cust_slot_name: cust_slot_value},
    )
    assert result == {
        "text": str(text_slot_value),
        "custom": {
            "field": str(cust_slot_value),
            "properties": {"field_prefixed": f"prefix_{str(cust_slot_value)}"},
        },
    }


@pytest.mark.parametrize(
    "attach_slot_name, attach_slot_value", [("attach_file", "https://attach.pdf")]
)
def test_nlg_fill_response_attachment(attach_slot_name, attach_slot_value):
    response = {"attachment": "{" + attach_slot_name + "}"}
    t = TemplatedNaturalLanguageGenerator(responses=dict())
    result = t._fill_response(
        response=response, filled_slots={attach_slot_name: attach_slot_value}
    )
    assert result == {"attachment": str(attach_slot_value)}


@pytest.mark.parametrize(
    "button_slot_name, button_slot_value", [("button_1", "button_1")]
)
def test_nlg_fill_response_button(button_slot_name, button_slot_value):
    response = {
        "buttons": [
            {
                "payload": f'/choose{{{{"some_slot": "{{{button_slot_name}}}"}}}}',
                "title": f"{{{button_slot_name}}}",
            }
        ]
    }
    t = TemplatedNaturalLanguageGenerator(responses=dict())
    result = t._fill_response(
        response=response, filled_slots={button_slot_name: button_slot_value}
    )
    assert result == {
        "buttons": [
            {
                "payload": f'/choose{{"some_slot": "{button_slot_value}"}}',
                "title": f"{button_slot_value}",
            }
        ]
    }


@pytest.mark.parametrize(
    "quick_replies_slot_name, quick_replies_slot_value", [("qreply", "reply 1")]
)
def test_nlg_fill_response_quick_replies(
    quick_replies_slot_name, quick_replies_slot_value
):
    response = {"quick_replies": f"{{{quick_replies_slot_name}}}"}
    t = TemplatedNaturalLanguageGenerator(responses=dict())
    result = t._fill_response(
        response=response,
        filled_slots={quick_replies_slot_name: quick_replies_slot_value},
    )
    assert result == {"quick_replies": str(quick_replies_slot_value)}