Hugging Face's logo

Spaces:
zliang
/
ClimateChat
Running

App Files Community
ClimateChat
File size: 6,736 Bytes 
e247edf
ed7625a
 
 
 
 
 
 
 
 
 
 
 
e247edf
 
ed7625a
e247edf
ed7625a
 
 
 
 
 
 
e247edf
ed7625a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e247edf
 
 
 
 
ed7625a
 
 
 
e247edf
ed7625a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e247edf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ed7625a
 
 
 
 
 
 
 
e247edf
 
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
 
import openai
import streamlit as st
from langchain.llms import OpenAI
from langchain.chat_models import ChatOpenAI
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.chains import RetrievalQA

from langchain.prompts.prompt import PromptTemplate

from langchain.vectorstores import FAISS
import re
import time


# import e5-large-v2 embedding model
model_name = "intfloat/e5-large-v2"
model_kwargs = {'device': 'cuda'}
encode_kwargs = {'normalize_embeddings': False}
embeddings = HuggingFaceEmbeddings(
    model_name=model_name,
    model_kwargs=model_kwargs,
    encode_kwargs=encode_kwargs
)

# load IPCC database
db = FAISS.load_local("IPCC_index_e5_1000_pdf", embeddings)



def generate_response(input_text):
    docs = db.similarity_search(input_text,k=5)

    json1 = docs[0].metadata
    json2 = docs[1].metadata
    json3 = docs[2].metadata
    json4 = docs[3].metadata
    json5 = docs[4].metadata
    #st.write({"source1":json1["source"], "source2":json2["source"],"source3":json3["source"]})


    climate_TEMPLATE = """ You are ChatClimate, take a deep breath and provide an answer to educated general audience based on the context, and Format your answer in Markdown. :"
    
    Context: {context}

    Question: {question}
    
    Answer:
    

    check if you use the info below, if you used please add used source for in-text reference, if not used, do not add them .
   

    [{source1} page {page1}]
    [{source2} page {page2}]
    [{source3} page {page3}]
    [{source4} page {page4}]
    [{source5} page {page5}]

   Check if you use the source in your ansewer, make sure list used sources you refer to  and their hyperlinks as below in a section named "sources":

    [{source1} page {page1}](https://www.ipcc.ch/report/ar6/wg3/downloads/report/{source1}.pdf#page={page1})
    [{source2} page {page2}](https://www.ipcc.ch/report/ar6/wg3/downloads/report/{source2}.pdf#page={page2})
    [{source3} page {page3}](https://www.ipcc.ch/report/ar6/wg3/downloads/report/{source3}.pdf#page={page3})
    [{source4} page {page4}](https://www.ipcc.ch/report/ar6/wg3/downloads/report/{source4}.pdf#page={page4})
    [{source5} page {page5}](https://www.ipcc.ch/report/ar6/wg3/downloads/report/{source5}.pdf#page={page5})

   

    at the end of your answer, make sure to add a short highlight  of your answer in humor and make sure no more than 5 words.

    Highlight:  
    """
    climate_PROMPT = PromptTemplate(input_variables=["question", "context"], 
                                    partial_variables={"source1":json1["source"], "source2":json2["source"],
                                                       "source3":json3["source"],"source4":json4["source"],"source5":json5["source"],"page1":json1["page"],
                                                       "page2":json2["page"],"page3":json3["page"],"page4":json4["page"],"page5":json5["page"]},
                                    template=climate_TEMPLATE, )
    
    #climate_PROMPT.partial(source = docs[0].metadata)

    llm = ChatOpenAI(
        model_name="gpt-3.5-turbo",
        temperature=0.1,
        max_tokens=2000, 
        openai_api_key=openai_api_key
    )

# Define retriever
    retriever = db.as_retriever(search_kwargs={"k": 5})

    qa_chain = RetrievalQA.from_chain_type(llm, 
                                            retriever=retriever,
                                            chain_type="stuff", #"stuff", "map_reduce","refine", "map_rerank"
                                            return_source_documents=True,
                                            verbose=True,
                                            chain_type_kwargs={"prompt": climate_PROMPT}
                                            )
    
    return qa_chain({'query': input_text})


with st.sidebar:
    openai_api_key = st.text_input("OpenAI API Key", key="chatbot_api_key", type="password")
    "[Get an OpenAI API key](https://platform.openai.com/account/api-keys)"
    st.markdown("## 🌍 Welcome to ClimateChat! 🌍")
    st.markdown("ClimateChat Harnesses the latest [IPCC reports](https://www.ipcc.ch/report/ar6/wg3/) and the power of Large Language Models to answer your questions about climate change. When you interact with ClimateChat not only will you receive clear, concise, and accurate answers, but each response is coupled with sources and hyperlinks for further exploration and verification.\
                Our objective is to make climate change information accessible, understandable, and actionable for everyone, everywhere.")
st.title("💬🌍🌡️ClimateChat")
st.caption("💬 A Climate Change chatbot powered by OpenAI LLM and IPCC documents")
#col1, col2,  = st.columns(2)


if "messages" not in st.session_state:
    st.session_state["messages"] = [{"role": "assistant", "content": "Any question about the climate change?"}]

for msg in st.session_state.messages:
    st.chat_message(msg["role"]).write(msg["content"])

if prompt := st.chat_input():
    if not openai_api_key:
        st.info("Please add your OpenAI API key to continue.")
        st.stop()

    st.session_state.messages.append({"role": "user", "content": prompt})
    st.chat_message("user").write(prompt)
    result = generate_response(prompt)
    result_r = result["result"]
    index = result_r.find("Highlight:")

    # Extract everything after "Highlight:"
    match = re.search(r"Highlight: (.+)", result_r)
    if match:
        highlighted_text = match.group(1)
    else:
        highlighted_text="hello world"




        # Display assistant response in chat message container
    with st.chat_message("assistant"):
        message_placeholder = st.empty()
        full_response = ""
        assistant_response = result_r
        # Simulate stream of response with milliseconds delay
        for chunk in assistant_response.split():
            full_response += chunk + " "
            time.sleep(0.05)
            # Add a blinking cursor to simulate typing
            message_placeholder.write(full_response + "▌")
        message_placeholder.write(result_r)
    # Add assistant response to chat history
    st.session_state.messages.append({"role": "assistant", "content": result_r})
        
    #st.session_state.messages.append({"role": "assistant", "content": result["result"]})
    #st.chat_message("assistant").write(result_r)
    #display_typing_effect(st.chat_message("assistant"), result_r)
    #st.markdown(result['source_documents'][0])
    #st.markdown(result['source_documents'][1])
    #st.markdown(result['source_documents'][2])
    #st.markdown(result['source_documents'][3])
    #st.markdown(result['source_documents'][4])


    #st.image("https://cataas.com/cat/says/"+highlighted_text)