File size: 6,770 Bytes
d5edf96
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6f7e600
d5edf96
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6f7e600
d5edf96
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
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
"""Main entrypoint for the app."""
import os
import time
from queue import Queue
from timeit import default_timer as timer

import gradio as gr
from anyio.from_thread import start_blocking_portal

from app_modules.init import app_init
from app_modules.llm_chat_chain import ChatChain
from app_modules.utils import print_llm_response, remove_extra_spaces

llm_loader, qa_chain = app_init()

show_param_settings = os.environ.get("SHOW_PARAM_SETTINGS") == "true"
share_gradio_app = os.environ.get("SHARE_GRADIO_APP") == "true"
using_openai = os.environ.get("LLM_MODEL_TYPE") == "openai"
chat_with_llama_2 = (
    not using_openai and os.environ.get("USE_LLAMA_2_PROMPT_TEMPLATE") == "true"
)
chat_history_enabled = (
    not chat_with_llama_2 and os.environ.get("CHAT_HISTORY_ENABLED") == "true"
)

model = (
    "OpenAI GPT-3.5"
    if using_openai
    else os.environ.get("HUGGINGFACE_MODEL_NAME_OR_PATH")
)
href = (
    "https://platform.openai.com/docs/models/gpt-3-5"
    if using_openai
    else f"https://huggingface.co/{model}"
)

if chat_with_llama_2:
    qa_chain = ChatChain(llm_loader)
    name = "Llama-2"
else:
    name = "SMU Library Chatbot"

title = f"""<h1 align="left" style="min-width:200px; margin-top:0;"> Chat with {name} </h1>"""

description_top = f"""\
<div align="left">
<p> Currently Running: <a href="{href}">{model}</a></p>
</div>
"""

description = """\
<div align="center" style="margin:16px 0">
The demo is built on <a href="https://github.com/hwchase17/langchain">LangChain</a>.
</div>
"""

CONCURRENT_COUNT = 1


def qa(chatbot):
    user_msg = chatbot[-1][0]
    q = Queue()
    result = Queue()
    job_done = object()

    def task(question, chat_history):
        start = timer()
        inputs = {"question": question}
        if not chat_with_llama_2:
            inputs["chat_history"] = chat_history
        ret = qa_chain.call_chain(inputs, None, q)
        end = timer()

        print(f"Completed in {end - start:.3f}s")
        print_llm_response(ret)

        q.put(job_done)
        result.put(ret)

    with start_blocking_portal() as portal:
        chat_history = []
        if chat_history_enabled:
            for i in range(len(chatbot) - 1):
                element = chatbot[i]
                item = (element[0] or "", element[1] or "")
                chat_history.append(item)

        portal.start_task_soon(task, user_msg, chat_history)

        content = ""
        count = 2 if len(chat_history) > 0 else 1

        while count > 0:
            while q.empty():
                print("nothing generated yet - retry in 0.5s")
                time.sleep(0.5)

            for next_token in llm_loader.streamer:
                if next_token is job_done:
                    break
                content += next_token or ""
                chatbot[-1][1] = remove_extra_spaces(content)

                if count == 1:
                    yield chatbot

            count -= 1

        if not chat_with_llama_2:
            chatbot[-1][1] += "\n\nSources:\n"
            ret = result.get()
            titles = []
            for doc in ret["source_documents"]:
                url = doc.metadata["url"]
                if "page" in doc.metadata:
                    page = doc.metadata["page"] + 1
                    url = f"{url}#page={page}"
                title = url
                if title not in titles:
                    titles.append(title)
                    chatbot[-1][1] += f"1. [{title}]({url})\n"

        yield chatbot


with open("assets/custom.css", "r", encoding="utf-8") as f:
    customCSS = f.read()

with gr.Blocks(css=customCSS) as demo:
    user_question = gr.State("")
    with gr.Row():
        gr.HTML(title)
    gr.Markdown(description_top)
    with gr.Row().style(equal_height=True):
        with gr.Column(scale=5):
            with gr.Row():
                chatbot = gr.Chatbot(elem_id="inflaton_chatbot").style(height="100%")
            with gr.Row():
                with gr.Column(scale=2):
                    user_input = gr.Textbox(
                        show_label=False, placeholder="Enter your question here"
                    ).style(container=False)
                with gr.Column(
                    min_width=70,
                ):
                    submitBtn = gr.Button("Send")
                with gr.Column(
                    min_width=70,
                ):
                    clearBtn = gr.Button("Clear")
        if show_param_settings:
            with gr.Column():
                with gr.Column(
                    min_width=50,
                ):
                    with gr.Tab(label="Parameter Setting"):
                        gr.Markdown("# Parameters")
                        top_p = gr.Slider(
                            minimum=-0,
                            maximum=1.0,
                            value=0.95,
                            step=0.05,
                            # interactive=True,
                            label="Top-p",
                        )
                        temperature = gr.Slider(
                            minimum=0.1,
                            maximum=2.0,
                            value=0,
                            step=0.1,
                            # interactive=True,
                            label="Temperature",
                        )
                        max_new_tokens = gr.Slider(
                            minimum=0,
                            maximum=2048,
                            value=2048,
                            step=8,
                            # interactive=True,
                            label="Max Generation Tokens",
                        )
                        max_context_length_tokens = gr.Slider(
                            minimum=0,
                            maximum=4096,
                            value=4096,
                            step=128,
                            # interactive=True,
                            label="Max Context Tokens",
                        )
    gr.Markdown(description)

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

    user_input.submit(
        chat, [user_input, chatbot], [user_input, chatbot], queue=True
    ).then(qa, chatbot, chatbot)

    submitBtn.click(
        chat, [user_input, chatbot], [user_input, chatbot], queue=True, api_name="chat"
    ).then(qa, chatbot, chatbot)

    def reset():
        return "", []

    clearBtn.click(
        reset,
        outputs=[user_input, chatbot],
        show_progress=True,
        api_name="reset",
    )

demo.title = "Chat with SMU Library Chatbot" if chat_with_llama_2 else "Chat with Llama-2"
demo.queue(concurrency_count=CONCURRENT_COUNT).launch(share=share_gradio_app)