File size: 5,893 Bytes
507d13c
 
 
 
 
 
 
 
 
 
 
 
 
bcf7c58
507d13c
bcf7c58
507d13c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bcf7c58
99e964c
507d13c
 
 
 
 
d118bdb
507d13c
 
 
 
bcf7c58
507d13c
 
 
 
 
 
 
 
 
 
2c0084d
507d13c
bcf7c58
 
d118bdb
 
 
 
 
507d13c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d118bdb
 
 
 
 
507d13c
 
 
 
 
 
99e964c
2c0084d
 
bcf7c58
507d13c
 
 
 
 
 
 
 
 
bcf7c58
 
 
507d13c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bcf7c58
507d13c
 
 
 
 
 
 
 
 
 
 
4fa708b
 
 
 
 
 
 
 
 
 
 
 
 
d118bdb
 
 
 
 
 
 
4fa708b
 
 
 
2c0084d
 
 
4fa708b
 
 
 
 
 
 
2c0084d
 
4fa708b
2c0084d
 
 
bcf7c58
2c0084d
 
 
 
 
 
4fa708b
d118bdb
 
 
 
 
4fa708b
 
 
 
507d13c
4fa708b
 
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
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
import cohere
import os
import pinecone
import uuid

from typing import List, Dict
from dotenv import load_dotenv


load_dotenv()

co = cohere.Client(os.environ["COHERE_API_KEY"])

pc = pinecone.Pinecone(api_key=os.environ["PINECONE_API_KEY"])

index = pc.Index("td-sec-embeddings")


def retrieve(index: pinecone.Index, query: str) -> List[Dict[str, str]]:
    """
    Retrieves documents based on the given query.

    Parameters:
    query (str): The query to retrieve documents for.

    Returns:
    List[Dict[str, str]]: A list of dictionaries representing the retrieved  documents, with 'title', 'snippet', and 'url' keys.
    """
    docs_retrieved = []

    query_emb = co.embed(
        texts=[query], model="embed-english-v3.0", input_type="search_query"
    ).embeddings

    res = index.query(vector=query_emb, top_k=100, include_metadata=True)

    docs_to_rerank = [match["metadata"] for match in res["matches"]]

    rerank_results = co.rerank(
        query=query,
        documents=docs_to_rerank,
        top_n=5,
        model="rerank-english-v2.0",
    )

    docs_retrieved = []
    for hit in rerank_results.results:
        docs_retrieved.append(docs_to_rerank[hit.index])

    return docs_retrieved


class Chatbot:
    def __init__(self, co: cohere.Client, index: pinecone.Index):
        self.index = index
        self.conversation_id = str(uuid.uuid4())
        self.co = co
        self.docs = None

    def send_initial_instructions(self):
        response = self.co.chat_stream(
            message="""You are an expert in TD Bank's annual reports and have access to the 2023 and 2022 annual report. Respond with a polite welcome message.""",
            conversation_id=self.conversation_id,
        )
        return response

    def generate_response(self, message: str):
        """
        Generates a response to the user's message.

        Parameters:
        message (str): The user's message.

        Yields:
        Event: A response event generated by the chatbot.

        Returns:
        List[Dict[str, str]]: A list of dictionaries representing the retrieved documents.

        """

        # Generate search queries (if any)
        response = self.co.chat(
            message=message,
            search_queries_only=True,
            conversation_id=self.conversation_id,
        )

        # If there are search queries, retrieve documents and respond
        if response.search_queries:
            print("Retrieving information...")

            documents = self.retrieve_docs(response)

            self.docs = {f"doc_{i}": document for i, document in enumerate(documents)}

            response = self.co.chat_stream(
                message=message,
                documents=documents,
                conversation_id=self.conversation_id,
            )
            for event in response:
                yield event

        # If there is no search query, directly respond
        else:
            response = self.co.chat_stream(
                message=message,
                conversation_id=self.conversation_id,
            )
            for event in response:
                yield event

    def retrieve_docs(self, response) -> List[Dict[str, str]]:
        """
        Retrieves documents based on the search queries in the response.

        Parameters:
        response: The response object containing search queries.

        Returns:
        List[Dict[str, str]]: A list of dictionaries representing the retrieved documents.

        """
        # Get the query(s)
        queries = []
        for search_query in response.search_queries:
            queries.append(search_query.text)

        # Retrieve documents for each query
        retrieved_docs = []
        for query in queries:
            retrieved_docs.extend(retrieve(self.index, query))

        return retrieved_docs


import gradio as gr

with gr.Blocks() as demo:
    chatbot = gr.Chatbot()
    msg = gr.Textbox()
    clear = gr.Button("Clear")

    cohere_chatbot_var = gr.State()

    def user(user_message, history):
        return "", history + [[user_message, None]]

    def chat_function(history, cohere_chatbot):
        if cohere_chatbot is None:
            cohere_chatbot = Chatbot(co, index)
            response = cohere_chatbot.send_initial_instructions()
            history = [[None, ""]]
            for event in response:
                if event.event_type == "text-generation":
                    history[0][1] += str(event.text)
                    yield history, cohere_chatbot
            return

        message = history[-1][0]
        history[-1][1] = ""

        documents_used = set()
        flag = True

        for event in cohere_chatbot.generate_response(message):
            if event.event_type == "text-generation":
                history[-1][1] += str(event.text)
                yield history, cohere_chatbot

            # Citations
            if event.event_type == "citation-generation":
                if flag:
                    history[-1][1] += "\n\n**DOCUMENTS CONSULTED:**\n\n"
                    yield history, cohere_chatbot
                    flag = False

                for citation in event.citations:
                    documents_used.update(citation.document_ids)

        urls_used = set(cohere_chatbot.docs[doc_id]["url"] for doc_id in documents_used)

        for url in sorted(urls_used):
            history[-1][1] += f"* {url}\n"
            yield history, cohere_chatbot

    # Make sure we run the thing once to initialize!
    demo.load(
        chat_function, [chatbot, cohere_chatbot_var], [chatbot, cohere_chatbot_var]
    )

    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
        chat_function, [chatbot, cohere_chatbot_var], [chatbot, cohere_chatbot_var]
    )
    clear.click(lambda: None, None, chatbot, queue=False)

demo.queue()
demo.launch()