File size: 4,957 Bytes
9105d7e
981469a
7279d6d
981469a
7279d6d
 
 
 
 
 
 
 
 
6bc96c2
7279d6d
 
 
 
9105d7e
7279d6d
 
 
 
 
 
 
8f7a4d5
 
7279d6d
9105d7e
 
7279d6d
 
 
9105d7e
7279d6d
8f7a4d5
7279d6d
 
 
8f7a4d5
7279d6d
8f7a4d5
 
 
 
 
 
7279d6d
9105d7e
7279d6d
 
8f7a4d5
7279d6d
8f7a4d5
981469a
8f7a4d5
7279d6d
 
 
981469a
8f7a4d5
981469a
7279d6d
 
 
 
 
 
 
9105d7e
 
 
 
 
 
7279d6d
 
9105d7e
7279d6d
9105d7e
7279d6d
9105d7e
7279d6d
 
 
 
9105d7e
7279d6d
 
 
 
 
 
 
6bc96c2
 
 
7279d6d
 
 
 
10d05a8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9105d7e
981469a
7279d6d
 
 
 
 
 
10d05a8
 
 
 
 
 
 
 
 
 
 
 
7279d6d
 
 
10d05a8
7279d6d
 
9105d7e
7279d6d
981469a
6bc96c2
 
 
 
 
 
7279d6d
 
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
from typing import Any, Tuple
import gradio as gr
from langchain_openai import OpenAIEmbeddings
from langchain_community.vectorstores import Chroma
from langchain.chains import ConversationalRetrievalChain
from langchain_openai import ChatOpenAI
from langchain_community.document_loaders import PyMuPDFLoader
import fitz
from PIL import Image
import os
import re
import openai

openai.api_key = "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"

def add_text(history, text: str):
    if not text:
        raise gr.Error("Enter text")
    history.append((text, ""))
    return history

class MyApp:
    def __init__(self) -> None:
        self.OPENAI_API_KEY: str = openai.api_key
        self.chain = None
        self.chat_history: list = []
        self.documents = None
        self.file_name = None

    def __call__(self, file: str) -> ConversationalRetrievalChain:
        if self.chain is None:
            self.chain = self.build_chain(file)
        return self.chain

    def process_file(self, file: str) -> Image:
        loader = PyMuPDFLoader(file.name)
        self.documents = loader.load()
        pattern = r"/([^/]+)$"
        match = re.search(pattern, file.name)
        try:
            self.file_name = match.group(1)
        except:
            self.file_name = os.path.basename(file)
        doc = fitz.open(file.name)
        page = doc[0]
        pix = page.get_pixmap(dpi=150)
        image = Image.frombytes("RGB", [pix.width, pix.height], pix.samples)
        return image

    def build_chain(self, file: str) -> str:
        embeddings = OpenAIEmbeddings(openai_api_key=self.OPENAI_API_KEY)
        pdfsearch = Chroma.from_documents(
            self.documents,
            embeddings,
            collection_name=self.file_name,
        )
        self.chain = ConversationalRetrievalChain.from_llm(
            ChatOpenAI(temperature=0.0, openai_api_key=self.OPENAI_API_KEY),
            retriever=pdfsearch.as_retriever(search_kwargs={"k": 1}),
            return_source_documents=True,
        )
        return "Vector database built successfully!"

def get_response(history, query, file):
    if not file:
        raise gr.Error(message="Upload a PDF")
    chain = app(file)
    result = chain(
        {"question": query, "chat_history": app.chat_history}, return_only_outputs=True
    )
    app.chat_history.append((query, result["answer"]))
    source_docs = result["source_documents"]
    source_texts = []
    for doc in source_docs:
        source_texts.append(f"Page {doc.metadata['page'] + 1}: {doc.page_content}")
    source_texts_str = "\n\n".join(source_texts)
    for char in result["answer"]:
        history[-1][-1] += char
        yield history, "", source_texts_str

def render_file(file) -> Image:
    doc = fitz.open(file.name)
    page = doc[0]
    pix = page.get_pixmap(dpi=150)
    image = Image.frombytes("RGB", [pix.width, pix.height], pix.samples)
    return image

def purge_chat_and_render_first(file) -> Tuple[Image, list]:
    app.chat_history = []
    doc = fitz.open(file.name)
    page = doc[0]
    pix = page.get_pixmap(dpi=150)
    image = Image.frombytes("RGB", [pix.width, pix.height], pix.samples)
    return image, []

def refresh_chat():
    app.chat_history = []
    return []

app = MyApp()

with gr.Blocks() as demo:
    with gr.Tab("Step 1: Upload PDF"):
        btn = gr.UploadButton("📁 Upload a PDF", file_types=[".pdf"])
        show_img = gr.Image(label="Uploaded PDF")

    with gr.Tab("Step 2: Process File"):
        process_btn = gr.Button("Process PDF")
        show_img_processed = gr.Image(label="Processed PDF")

    with gr.Tab("Step 3: Build Vector Database"):
        build_vector_btn = gr.Button("Build Vector Database")
        status_text = gr.Textbox(label="Status", value="", interactive=False)

    with gr.Tab("Step 4: Ask Questions"):
        chatbot = gr.Chatbot(value=[], elem_id="chatbot")
        txt = gr.Textbox(
            show_label=False,
            placeholder="Enter text and press submit",
            scale=2
        )
        submit_btn = gr.Button("Submit", scale=1)
        refresh_btn = gr.Button("Refresh Chat", scale=1)
        source_texts_output = gr.Textbox(label="Source Texts", interactive=False)

    btn.upload(
        fn=purge_chat_and_render_first,
        inputs=[btn],
        outputs=[show_img, chatbot],
    )

    process_btn.click(
        fn=app.process_file,
        inputs=[btn],
        outputs=[show_img_processed],
    )

    build_vector_btn.click(
        fn=app.build_chain,
        inputs=[btn],
        outputs=[status_text],
    )

    submit_btn.click(
        fn=add_text,
        inputs=[chatbot, txt],
        outputs=[chatbot],
        queue=False,
    ).success(
        fn=get_response, inputs=[chatbot, txt, btn], outputs=[chatbot, txt, source_texts_output]
    )

    refresh_btn.click(
        fn=refresh_chat,
        inputs=[],
        outputs=[chatbot],
    )

demo.queue()
demo.launch()