File size: 8,002 Bytes
1696c32
 
 
1b88635
 
 
c8a1687
1b88635
1696c32
 
 
 
 
1b88635
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c8a1687
 
 
 
 
 
 
 
1696c32
 
 
 
 
c8a1687
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1696c32
c8a1687
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1696c32
c8a1687
 
 
 
 
 
 
 
 
 
 
 
 
1696c32
c8a1687
 
1696c32
1b88635
 
c8a1687
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1b88635
c8a1687
 
 
 
1b88635
c8a1687
 
 
 
1b88635
 
 
 
 
 
c8a1687
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1b88635
c8a1687
 
1b88635
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
import os
from pathlib import Path

import numpy as np
import pandas as pd

from buster.buster import Buster, BusterConfig
from buster.documents import DocumentsManager

TEST_DATA_DIR = Path(__file__).resolve().parent / "data"
DOCUMENTS_FILE = os.path.join(str(TEST_DATA_DIR), "document_embeddings_huggingface_subset.tar.gz")


def get_fake_embedding(length=1536):
    rng = np.random.default_rng()
    return list(rng.random(length, dtype=np.float32))


class DocumentsMock(DocumentsManager):
    def __init__(self, filepath):
        self.filepath = filepath

        n_samples = 100
        self.documents = pd.DataFrame.from_dict(
            {
                "title": ["test"] * n_samples,
                "url": ["http://url.com"] * n_samples,
                "content": ["cool text"] * n_samples,
                "embedding": [get_fake_embedding()] * n_samples,
                "n_tokens": [10] * n_samples,
                "source": ["fake source"] * n_samples,
            }
        )

    def add(self, documents):
        pass

    def get_documents(self, source):
        return self.documents


def test_chatbot_mock_data(tmp_path, monkeypatch):
    gpt_expected_answer = "this is GPT answer"
    monkeypatch.setattr("buster.buster.get_documents_manager_from_extension", lambda filepath: DocumentsMock)
    monkeypatch.setattr("buster.buster.get_embedding", lambda x, engine: get_fake_embedding())
    monkeypatch.setattr("openai.Completion.create", lambda **kwargs: {"choices": [{"text": gpt_expected_answer}]})

    hf_transformers_cfg = BusterConfig(
        documents_file=tmp_path / "not_a_real_file.tar.gz",
        unknown_prompt="This doesn't seem to be related to the huggingface library. I am not sure how to answer.",
        embedding_model="text-embedding-ada-002",
        top_k=3,
        thresh=0.7,
        max_words=3000,
        response_format="slack",
        completer_cfg={
            "name": "GPT3",
            "text_before_prompt": (
                """You are a slack chatbot assistant answering technical questions about huggingface transformers, a library to train transformers in python.\n"""
                """Make sure to format your answers in Markdown format, including code block and snippets.\n"""
                """Do not include any links to urls or hyperlinks in your answers.\n\n"""
                """Now answer the following question:\n"""
            ),
            "text_before_documents": "",
            "completion_kwargs": {
                "engine": "text-davinci-003",
                "max_tokens": 200,
                "temperature": None,
                "top_p": None,
                "frequency_penalty": 1,
                "presence_penalty": 1,
            },
        },
    )
    buster = Buster(hf_transformers_cfg)
    answer = buster.process_input("What is a transformer?")
    assert isinstance(answer, str)
    assert answer.startswith(gpt_expected_answer)


def test_chatbot_real_data__chatGPT():
    hf_transformers_cfg = BusterConfig(
        documents_file=DOCUMENTS_FILE,
        unknown_prompt="I'm sorry, but I am an AI language model trained to assist with questions related to the huggingface transformers library. I cannot answer that question as it is not relevant to the library or its usage. Is there anything else I can assist you with?",
        embedding_model="text-embedding-ada-002",
        top_k=3,
        thresh=0.7,
        max_words=3000,
        response_format="slack",
        completer_cfg={
            "name": "ChatGPT",
            "text_before_prompt": (
                """You are a slack chatbot assistant answering technical questions about huggingface transformers, a library to train transformers in python.\n"""
                """Make sure to format your answers in Markdown format, including code block and snippets.\n"""
                """Do not include any links to urls or hyperlinks in your answers.\n\n"""
                """Now answer the following question:\n"""
            ),
            "text_before_documents": "Only use these documents as reference:\n",
            "completion_kwargs": {
                "model": "gpt-3.5-turbo",
            },
        },
    )
    buster = Buster(hf_transformers_cfg)
    answer = buster.process_input("What is a transformer?")
    assert isinstance(answer, str)


def test_chatbot_real_data__chatGPT_OOD():
    buster_cfg = BusterConfig(
        documents_file=DOCUMENTS_FILE,
        unknown_prompt="I'm sorry, but I am an AI language model trained to assist with questions related to the huggingface transformers library. I cannot answer that question as it is not relevant to the library or its usage. Is there anything else I can assist you with?",
        embedding_model="text-embedding-ada-002",
        top_k=3,
        thresh=0.7,
        max_words=3000,
        completer_cfg={
            "name": "ChatGPT",
            "text_before_prompt": (
                """You are a slack chatbot assistant answering technical questions about huggingface transformers, a library to train transformers in python. """
                """Make sure to format your answers in Markdown format, including code block and snippets. """
                """Do not include any links to urls or hyperlinks in your answers. """
                """If you do not know the answer to a question, or if it is completely irrelevant to the library usage, let the user know you cannot answer. """
                """Use this response: """
                """I'm sorry, but I am an AI language model trained to assist with questions related to the huggingface transformers library. I cannot answer that question as it is not relevant to the library or its usage. Is there anything else I can assist you with?"""
                """For example:\n"""
                """What is the meaning of life for huggingface?\n"""
                """I'm sorry, but I am an AI language model trained to assist with questions related to the huggingface transformers library. I cannot answer that question as it is not relevant to the library or its usage. Is there anything else I can assist you with?"""
                """Now answer the following question:\n"""
            ),
            "text_before_documents": "Only use these documents as reference:\n",
            "completion_kwargs": {
                "model": "gpt-3.5-turbo",
            },
        },
        response_format="gradio",
    )
    buster = Buster(buster_cfg)
    answer = buster.process_input("What is a good recipe for brocolli soup?")
    assert isinstance(answer, str)
    assert buster_cfg.unknown_prompt in answer


def test_chatbot_real_data__GPT():
    hf_transformers_cfg = BusterConfig(
        documents_file=DOCUMENTS_FILE,
        unknown_prompt="This doesn't seem to be related to the huggingface library. I am not sure how to answer.",
        embedding_model="text-embedding-ada-002",
        top_k=3,
        thresh=0.7,
        max_words=3000,
        response_format="slack",
        completer_cfg={
            "name": "GPT3",
            "text_before_prompt": (
                """You are a slack chatbot assistant answering technical questions about huggingface transformers, a library to train transformers in python.\n"""
                """Make sure to format your answers in Markdown format, including code block and snippets.\n"""
                """Do not include any links to urls or hyperlinks in your answers.\n\n"""
                """Now answer the following question:\n"""
            ),
            "text_before_documents": "",
            "completion_kwargs": {
                "engine": "text-davinci-003",
                "max_tokens": 200,
                "temperature": None,
                "top_p": None,
                "frequency_penalty": 1,
                "presence_penalty": 1,
            },
        },
    )
    buster = Buster(hf_transformers_cfg)
    answer = buster.process_input("What is a transformer?")
    assert isinstance(answer, str)