File size: 4,317 Bytes
a5686cb
 
 
 
 
 
 
6a89c2d
 
 
 
a5686cb
6a89c2d
a5686cb
 
6a89c2d
a5686cb
 
 
 
 
 
 
 
5522fae
a5686cb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6a89c2d
 
 
a5686cb
6a89c2d
a5686cb
 
 
 
 
 
 
 
 
 
 
 
 
 
6a89c2d
a5686cb
6a89c2d
a5686cb
 
 
 
 
 
 
 
2983354
 
 
 
 
 
a6ba9ec
a5686cb
2983354
 
a5686cb
2983354
 
 
 
 
 
a5686cb
2983354
 
 
 
 
 
 
 
a5686cb
 
6a89c2d
a5686cb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2983354
 
 
a5686cb
f9aee46
6a89c2d
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
import gradio as gr
from transformers import pipeline
from haystack.document_stores import FAISSDocumentStore
from haystack.nodes import EmbeddingRetriever
import numpy as np
import openai

document_store = FAISSDocumentStore.load(
    index_path=f"./documents/climate_gpt.faiss",
    config_path=f"./documents/climate_gpt.json",
)

classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
system_template = {
    "role": "system",
    "content": "You have been a climate change expert for 30 years. You answer questions about climate change in an educationnal and concise manner. Whenever possible your answers are backed up by facts and numbers from scientific reports.",
}

dense = EmbeddingRetriever(
    document_store=document_store,
    embedding_model="sentence-transformers/multi-qa-mpnet-base-dot-v1",
    model_format="sentence_transformers",
)

openai.api_key=""

def is_climate_change_related(sentence: str) -> bool:
    results = classifier(
        sequences=sentence,
        candidate_labels=["climate change related", "non climate change related"],
    )
    return results["labels"][np.argmax(results["scores"])] == "climate change related"


def make_pairs(lst):
    """from a list of even lenght, make tupple pairs"""
    return [(lst[i], lst[i + 1]) for i in range(0, len(lst), 2)]


def gen_conv(query: str, history=[system_template], ipcc=True):
    """return (answer:str, history:list[dict], sources:str)"""
    retrieve = ipcc and is_climate_change_related(query)
    sources = ""
    messages = history + [
        {"role": "user", "content": query},
    ]

    if retrieve:
        docs = dense.retrieve(query=query, top_k=5)
        sources = "\n\n".join(
            [
                "If relevant, use those extracts in your answer and give the reference of the information you used."
            ]
            + [
                f"{d.meta['file_name']} Page {d.meta['page_number']}\n{d.content}"
                for d in docs
            ]
        )
        messages.append({"role": "system", "content": sources})

    answer = openai.ChatCompletion.create(
        model="gpt-3.5-turbo",
        messages=messages,
        temperature=0.2,
        #         max_tokens=200,
    )["choices"][0]["message"]["content"]

    if retrieve:
        messages.pop()
        # answer = "(top 5 documents retrieved) " + answer
        sources = "\n\n".join(
            f"{d.meta['file_name']} Page {d.meta['page_number']}:\n{d.content}"
            for d in docs
        )
    messages.append({"role": "assistant", "content": answer})
    gradio_format = make_pairs([a["content"] for a in messages[1:]])

    return gradio_format, messages, sources


def set_openai_api_key(api_key):
    """Set the api key and return chain.
    If no api_key, then None is returned.
    """
    os.environ["OPENAI_API_KEY"] = api_key
    openai.api_key = api_key
    return f"You're all set: this is your api key: {openai.api_key}"

    
# Gradio
with gr.Blocks(title="Eki IPCC Explorer") as demo:
    gr.Markdown(
        """
        # Add your OPENAI api key First
        """
    )
    
    with gr.Row():
            openai_api_key_textbox = gr.Textbox(
                placeholder="Paste your OpenAI API key (sk-...) and hit Enter",
                show_label=False,
                lines=1,
                type="password",
            )
    
    
    gr.Markdown(
        """
        # Ask me anything, I'm a climate expert
        """
    )

    with gr.Row():
        with gr.Column(scale=2):
            chatbot = gr.Chatbot()
            state = gr.State([system_template])

            with gr.Row():
                ask = gr.Textbox(
                    show_label=False, placeholder="Enter text and press enter"
                ).style(container=False)

        with gr.Column(scale=1, variant="panel"):

            gr.Markdown("### Sources")
            sources_textbox = gr.Textbox(
                interactive=False, show_label=False, max_lines=50
            )

    ask.submit(
        fn=gen_conv, inputs=[ask, state], outputs=[chatbot, state, sources_textbox]
    )
    
    openai_api_key_textbox.change(set_openai_api_key, inputs=[openai_api_key_textbox])
    openai_api_key_textbox.submit(set_openai_api_key, inputs=[openai_api_key_textbox])

demo.launch()