Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -3,7 +3,7 @@ from PyPDF2 import PdfReader
|
|
| 3 |
import langchain
|
| 4 |
from textwrap import dedent
|
| 5 |
import pandas as pd
|
| 6 |
-
|
| 7 |
from langchain_community.callbacks import StreamlitCallbackHandler
|
| 8 |
from langchain_openai import ChatOpenAI
|
| 9 |
from langchain_community.chat_models import ChatGooglePalm
|
|
@@ -30,15 +30,57 @@ from langchain.prompts import (
|
|
| 30 |
|
| 31 |
|
| 32 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
#api_key2 = st.secrets["OPENAI_API_KEY"]
|
| 34 |
-
os.environ["GOOGLE_API_KEY"] =
|
| 35 |
|
| 36 |
-
st.set_page_config(page_title='Personal Chatbot', page_icon='books')
|
| 37 |
-
st.header('Knowledge Query Assistant')
|
| 38 |
-
st.write("Upload your file to begin a chat, or ask any general questions you have")
|
| 39 |
-
st.sidebar.title('Options')
|
| 40 |
|
| 41 |
|
|
|
|
| 42 |
st.sidebar.subheader("Please Choose the AI Engine")
|
| 43 |
use_google = st.sidebar.checkbox("Use Free AI", value =True)
|
| 44 |
use_openai = st.sidebar.checkbox("Use OpenAI with your API Key")
|
|
@@ -163,7 +205,6 @@ st.sidebar.markdown("[Website](https://tenlancer.com/)")
|
|
| 163 |
########--Save PDF--########
|
| 164 |
|
| 165 |
|
| 166 |
-
|
| 167 |
def text_to_audio(response, lang):
|
| 168 |
audio_buffer = BytesIO()
|
| 169 |
audio_file = gTTS(text=response, lang=lang, slow=False)
|
|
@@ -173,7 +214,7 @@ def text_to_audio(response, lang):
|
|
| 173 |
|
| 174 |
|
| 175 |
def main():
|
| 176 |
-
|
| 177 |
if (use_openai and openai_api_key) or use_google:
|
| 178 |
if uploaded_file:
|
| 179 |
db = processing_csv_pdf_docx(uploaded_file)
|
|
@@ -181,17 +222,20 @@ def main():
|
|
| 181 |
st.success(f'Your File: {file.name} is Embedded', icon="✅")
|
| 182 |
|
| 183 |
for msg in st.session_state.messages:
|
| 184 |
-
|
| 185 |
-
|
|
|
|
|
|
|
|
|
|
| 186 |
|
| 187 |
-
|
| 188 |
#st.audio(audio_msg, format='audio/mp3').audio(audio_msg)
|
| 189 |
|
| 190 |
|
| 191 |
if prompt := st.chat_input(placeholder="Type your question!"):
|
| 192 |
st.session_state.messages.append({"role": "user", "content": prompt})
|
| 193 |
-
st.chat_message("user").write(prompt)
|
| 194 |
-
memory = ConversationBufferMemory(memory_key="chat_history", input_key="question", human_prefix= "", ai_prefix= "")
|
| 195 |
user_message = {"role": "user", "content": prompt}
|
| 196 |
|
| 197 |
|
|
@@ -200,15 +244,15 @@ def main():
|
|
| 200 |
user_prompt = st.session_state.messages[i]
|
| 201 |
ai_res = st.session_state.messages[i + 1]
|
| 202 |
|
| 203 |
-
|
| 204 |
current_content = user_prompt["content"]
|
| 205 |
|
| 206 |
-
|
| 207 |
next_content = ai_res["content"]
|
| 208 |
|
| 209 |
# Concatenate role and content for context and output
|
| 210 |
-
user = f"
|
| 211 |
-
ai = f"
|
| 212 |
|
| 213 |
memory.save_context({"question": user}, {"output": ai})
|
| 214 |
|
|
@@ -224,12 +268,15 @@ def main():
|
|
| 224 |
compliment = ['thank you', 'thanks', 'thanks a lot', 'thanks a bunch', 'great', 'ok', 'ok thanks', 'okay', 'great', 'awesome', 'nice']
|
| 225 |
|
| 226 |
prompt_template =dedent(r"""
|
| 227 |
-
You are
|
| 228 |
-
|
| 229 |
-
|
| 230 |
-
|
|
|
|
|
|
|
|
|
|
| 231 |
|
| 232 |
-
this is the context
|
| 233 |
---------
|
| 234 |
{context}
|
| 235 |
---------
|
|
@@ -260,7 +307,7 @@ def main():
|
|
| 260 |
|
| 261 |
#chain = load_qa_chain(ChatOpenAI(temperature=0.9, model="gpt-3.5-turbo-0613", streaming=True) , verbose= True, prompt = PROMPT, memory=memory,chain_type="stuff")
|
| 262 |
|
| 263 |
-
with st.chat_message("Assistant"):
|
| 264 |
st_cb = StreamlitCallbackHandler(st.container())
|
| 265 |
if prompt.lower() in greetings:
|
| 266 |
response = 'Hi, how are you? I am here to help you get information from your file. How can I assist you?'
|
|
@@ -272,8 +319,7 @@ def main():
|
|
| 272 |
|
| 273 |
audio_buffer = text_to_audio(response, lang)
|
| 274 |
#st.audio(audio_buffer, format='audio/mp3')
|
| 275 |
-
st.session_state.messages.append({"role": "Assistant", "content": response})
|
| 276 |
-
#st.session_state.messages.append({"role": "Assistant", "content": response, "audio_content": audio_buffer})
|
| 277 |
|
| 278 |
elif prompt.lower() in compliment:
|
| 279 |
response = 'My pleasure! If you have any more questions, feel free to ask.'
|
|
@@ -285,12 +331,11 @@ def main():
|
|
| 285 |
|
| 286 |
audio_buffer = text_to_audio(response, lang)
|
| 287 |
#st.audio(audio_buffer, format='audio/mp3')
|
| 288 |
-
st.session_state.messages.append({"role": "Assistant", "content": response})
|
| 289 |
-
#st.session_state.messages.append({"role": "Assistant", "content": response, "audio_content": audio_buffer})
|
| 290 |
|
| 291 |
elif uploaded_file:
|
| 292 |
with st.spinner('Bot is typing ...'):
|
| 293 |
-
docs = db.similarity_search(prompt, k=
|
| 294 |
response = chain.run(input_documents=docs, question=prompt)
|
| 295 |
|
| 296 |
|
|
@@ -301,8 +346,8 @@ def main():
|
|
| 301 |
audio_buffer = text_to_audio(response, lang)
|
| 302 |
# st.audio(audio_buffer, format='audio/mp3')
|
| 303 |
#st.session_state.audio.append({"role": "Assistant", "audio": audio_buffer})
|
| 304 |
-
|
| 305 |
-
|
| 306 |
assistant_message = {"role": "assistant", "content": response}
|
| 307 |
else:
|
| 308 |
with st.spinner('Bot is typing ...'):
|
|
@@ -315,27 +360,25 @@ def main():
|
|
| 315 |
|
| 316 |
|
| 317 |
|
| 318 |
-
|
| 319 |
#st.audio(audio_buffer, format='audio/mp3')
|
| 320 |
#st.session_state.audio.append({"role": "Assistant", "audio": audio_buffer})
|
| 321 |
-
|
| 322 |
-
|
| 323 |
assistant_message = {"role": "assistant", "content": response}
|
| 324 |
|
| 325 |
st.write(response)
|
| 326 |
-
|
| 327 |
|
| 328 |
|
| 329 |
-
except Exception as e:
|
| 330 |
|
| 331 |
-
|
| 332 |
-
|
| 333 |
-
|
| 334 |
-
elif use_openai:
|
| 335 |
-
|
| 336 |
|
| 337 |
-
|
| 338 |
-
|
| 339 |
|
| 340 |
|
| 341 |
hide_streamlit_style = """
|
|
|
|
| 3 |
import langchain
|
| 4 |
from textwrap import dedent
|
| 5 |
import pandas as pd
|
| 6 |
+
from langchain_google_genai import ChatGoogleGenerativeAI
|
| 7 |
from langchain_community.callbacks import StreamlitCallbackHandler
|
| 8 |
from langchain_openai import ChatOpenAI
|
| 9 |
from langchain_community.chat_models import ChatGooglePalm
|
|
|
|
| 30 |
|
| 31 |
|
| 32 |
|
| 33 |
+
st.set_page_config(page_title='Personal Chatbot', page_icon='books')
|
| 34 |
+
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
|
| 38 |
+
st.markdown(
|
| 39 |
+
"""
|
| 40 |
+
<style>
|
| 41 |
+
[data-testid=stImage]{
|
| 42 |
+
text-align: center;
|
| 43 |
+
display: block;
|
| 44 |
+
margin-left: 10%;
|
| 45 |
+
margin-right:10%;
|
| 46 |
+
width: 100%;
|
| 47 |
+
}
|
| 48 |
+
img {
|
| 49 |
+
border-radius: 50%;
|
| 50 |
+
align: center;
|
| 51 |
+
}
|
| 52 |
+
</style>
|
| 53 |
+
""", unsafe_allow_html=True
|
| 54 |
+
)
|
| 55 |
+
|
| 56 |
+
|
| 57 |
+
|
| 58 |
+
st.image("tenlancer.png", width=80)
|
| 59 |
+
|
| 60 |
+
st.markdown("<h3 style='text-align: center; color: white;'> Knowledge Query Assistant </h3>", unsafe_allow_html=True)
|
| 61 |
+
|
| 62 |
+
|
| 63 |
+
|
| 64 |
+
|
| 65 |
+
st.markdown(
|
| 66 |
+
"""
|
| 67 |
+
<style>
|
| 68 |
+
[data-testid="stChatMessageContent"] p{
|
| 69 |
+
font-size: 1.2rem;
|
| 70 |
+
color: #404040
|
| 71 |
+
}
|
| 72 |
+
</style>
|
| 73 |
+
""", unsafe_allow_html=True
|
| 74 |
+
)
|
| 75 |
+
|
| 76 |
+
|
| 77 |
+
google_api_key = st.secrets["GOOGLE_API_KEY"]
|
| 78 |
#api_key2 = st.secrets["OPENAI_API_KEY"]
|
| 79 |
+
os.environ["GOOGLE_API_KEY"] = google_api_key
|
| 80 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 81 |
|
| 82 |
|
| 83 |
+
st.sidebar.header("options")
|
| 84 |
st.sidebar.subheader("Please Choose the AI Engine")
|
| 85 |
use_google = st.sidebar.checkbox("Use Free AI", value =True)
|
| 86 |
use_openai = st.sidebar.checkbox("Use OpenAI with your API Key")
|
|
|
|
| 205 |
########--Save PDF--########
|
| 206 |
|
| 207 |
|
|
|
|
| 208 |
def text_to_audio(response, lang):
|
| 209 |
audio_buffer = BytesIO()
|
| 210 |
audio_file = gTTS(text=response, lang=lang, slow=False)
|
|
|
|
| 214 |
|
| 215 |
|
| 216 |
def main():
|
| 217 |
+
# try:
|
| 218 |
if (use_openai and openai_api_key) or use_google:
|
| 219 |
if uploaded_file:
|
| 220 |
db = processing_csv_pdf_docx(uploaded_file)
|
|
|
|
| 222 |
st.success(f'Your File: {file.name} is Embedded', icon="✅")
|
| 223 |
|
| 224 |
for msg in st.session_state.messages:
|
| 225 |
+
if msg["role"] == "user":
|
| 226 |
+
st.chat_message("user", avatar="user.png").write(msg["content"])
|
| 227 |
+
|
| 228 |
+
if msg["role"] == "Assistant":
|
| 229 |
+
st.chat_message("Assistant", avatar="logo.png").write(msg["content"])
|
| 230 |
|
| 231 |
+
st.audio(msg["audio_content"], format='audio/wav')
|
| 232 |
#st.audio(audio_msg, format='audio/mp3').audio(audio_msg)
|
| 233 |
|
| 234 |
|
| 235 |
if prompt := st.chat_input(placeholder="Type your question!"):
|
| 236 |
st.session_state.messages.append({"role": "user", "content": prompt})
|
| 237 |
+
st.chat_message("user", avatar="user.png").write(prompt)
|
| 238 |
+
memory = ConversationBufferMemory(memory_key="chat_history", input_key="question", human_prefix= "User", ai_prefix= "Assistant")
|
| 239 |
user_message = {"role": "user", "content": prompt}
|
| 240 |
|
| 241 |
|
|
|
|
| 244 |
user_prompt = st.session_state.messages[i]
|
| 245 |
ai_res = st.session_state.messages[i + 1]
|
| 246 |
|
| 247 |
+
|
| 248 |
current_content = user_prompt["content"]
|
| 249 |
|
| 250 |
+
|
| 251 |
next_content = ai_res["content"]
|
| 252 |
|
| 253 |
# Concatenate role and content for context and output
|
| 254 |
+
user = f" {current_content}"
|
| 255 |
+
ai = f" {next_content}"
|
| 256 |
|
| 257 |
memory.save_context({"question": user}, {"output": ai})
|
| 258 |
|
|
|
|
| 268 |
compliment = ['thank you', 'thanks', 'thanks a lot', 'thanks a bunch', 'great', 'ok', 'ok thanks', 'okay', 'great', 'awesome', 'nice']
|
| 269 |
|
| 270 |
prompt_template =dedent(r"""
|
| 271 |
+
You are a helpful assistant.
|
| 272 |
+
talk humbly. Answer the question from the provided context. Do not answer from your own training data.
|
| 273 |
+
Use the following pieces of context to answer the question at the end.
|
| 274 |
+
If you don't know the answer, just say that you don't know. Do not makeup any answer.
|
| 275 |
+
Do not answer hypothetically. Do not answer in more than 100 words.
|
| 276 |
+
Please Do Not say: "Based on the provided context"
|
| 277 |
+
|
| 278 |
|
| 279 |
+
this is the context:
|
| 280 |
---------
|
| 281 |
{context}
|
| 282 |
---------
|
|
|
|
| 307 |
|
| 308 |
#chain = load_qa_chain(ChatOpenAI(temperature=0.9, model="gpt-3.5-turbo-0613", streaming=True) , verbose= True, prompt = PROMPT, memory=memory,chain_type="stuff")
|
| 309 |
|
| 310 |
+
with st.chat_message("Assistant", avatar="logo.png"):
|
| 311 |
st_cb = StreamlitCallbackHandler(st.container())
|
| 312 |
if prompt.lower() in greetings:
|
| 313 |
response = 'Hi, how are you? I am here to help you get information from your file. How can I assist you?'
|
|
|
|
| 319 |
|
| 320 |
audio_buffer = text_to_audio(response, lang)
|
| 321 |
#st.audio(audio_buffer, format='audio/mp3')
|
| 322 |
+
st.session_state.messages.append({"role": "Assistant", "content": response, "audio_content": audio_buffer})
|
|
|
|
| 323 |
|
| 324 |
elif prompt.lower() in compliment:
|
| 325 |
response = 'My pleasure! If you have any more questions, feel free to ask.'
|
|
|
|
| 331 |
|
| 332 |
audio_buffer = text_to_audio(response, lang)
|
| 333 |
#st.audio(audio_buffer, format='audio/mp3')
|
| 334 |
+
st.session_state.messages.append({"role": "Assistant", "content": response, "audio_content": audio_buffer})
|
|
|
|
| 335 |
|
| 336 |
elif uploaded_file:
|
| 337 |
with st.spinner('Bot is typing ...'):
|
| 338 |
+
docs = db.similarity_search(prompt, k=5, fetch_k=10)
|
| 339 |
response = chain.run(input_documents=docs, question=prompt)
|
| 340 |
|
| 341 |
|
|
|
|
| 346 |
audio_buffer = text_to_audio(response, lang)
|
| 347 |
# st.audio(audio_buffer, format='audio/mp3')
|
| 348 |
#st.session_state.audio.append({"role": "Assistant", "audio": audio_buffer})
|
| 349 |
+
st.session_state.messages.append({"role": "Assistant", "content": response, "audio_content": audio_buffer})
|
| 350 |
+
|
| 351 |
assistant_message = {"role": "assistant", "content": response}
|
| 352 |
else:
|
| 353 |
with st.spinner('Bot is typing ...'):
|
|
|
|
| 360 |
|
| 361 |
|
| 362 |
|
| 363 |
+
audio_buffer = text_to_audio(response, lang)
|
| 364 |
#st.audio(audio_buffer, format='audio/mp3')
|
| 365 |
#st.session_state.audio.append({"role": "Assistant", "audio": audio_buffer})
|
| 366 |
+
st.session_state.messages.append({"role": "Assistant", "content": response, "audio_content": audio_buffer})
|
| 367 |
+
|
| 368 |
assistant_message = {"role": "assistant", "content": response}
|
| 369 |
|
| 370 |
st.write(response)
|
| 371 |
+
st.audio(audio_buffer, format='audio/wav')
|
| 372 |
|
| 373 |
|
| 374 |
+
#except Exception as e:
|
| 375 |
|
| 376 |
+
# "Sorry, there was a problem. A corrupted file or;"
|
| 377 |
+
# if use_google:
|
| 378 |
+
# "Google PaLM AI only take English Data and Questions. Or the AI could not find the answer in your provided document."
|
| 379 |
+
#elif use_openai:
|
| 380 |
+
# "Please check your OpenAI API key"
|
| 381 |
|
|
|
|
|
|
|
| 382 |
|
| 383 |
|
| 384 |
hide_streamlit_style = """
|