Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -13,11 +13,7 @@ logging.basicConfig(level=logging.INFO)
|
|
| 13 |
logger = logging.getLogger(__name__)
|
| 14 |
|
| 15 |
# Page configuration
|
| 16 |
-
st.set_page_config(
|
| 17 |
-
page_title="DeepSeek Chatbot - ruslanmv.com",
|
| 18 |
-
page_icon="🤖",
|
| 19 |
-
layout="centered"
|
| 20 |
-
)
|
| 21 |
|
| 22 |
# Initialize session state for chat history
|
| 23 |
if "messages" not in st.session_state:
|
|
@@ -29,31 +25,13 @@ with st.sidebar:
|
|
| 29 |
st.markdown("[Get HuggingFace Token](https://huggingface.co/settings/tokens)")
|
| 30 |
|
| 31 |
# Dropdown to select model
|
| 32 |
-
model_options = [
|
| 33 |
-
"deepseek-ai/DeepSeek-R1-Distill-Qwen-32B",
|
| 34 |
-
]
|
| 35 |
selected_model = st.selectbox("Select Model", model_options, index=0)
|
| 36 |
|
| 37 |
-
system_message = st.text_area(
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
)
|
| 42 |
-
|
| 43 |
-
max_tokens = st.slider(
|
| 44 |
-
"Max Tokens",
|
| 45 |
-
10, 4000, 100
|
| 46 |
-
)
|
| 47 |
-
|
| 48 |
-
temperature = st.slider(
|
| 49 |
-
"Temperature",
|
| 50 |
-
0.1, 4.0, 0.3
|
| 51 |
-
)
|
| 52 |
-
|
| 53 |
-
top_p = st.slider(
|
| 54 |
-
"Top-p",
|
| 55 |
-
0.1, 1.0, 0.6
|
| 56 |
-
)
|
| 57 |
|
| 58 |
# Function to query the Hugging Face API
|
| 59 |
def query(payload, api_url):
|
|
@@ -79,11 +57,7 @@ def process_pdf(uploaded_file):
|
|
| 79 |
documents = loader.load()
|
| 80 |
|
| 81 |
# Split the documents into chunks
|
| 82 |
-
text_splitter = RecursiveCharacterTextSplitter(
|
| 83 |
-
chunk_size=1000,
|
| 84 |
-
chunk_overlap=200,
|
| 85 |
-
add_start_index=True
|
| 86 |
-
)
|
| 87 |
return text_splitter.split_documents(documents)
|
| 88 |
|
| 89 |
# Function to generate response using LangChain
|
|
@@ -120,8 +94,13 @@ if uploaded_file:
|
|
| 120 |
documents = process_pdf(uploaded_file)
|
| 121 |
context = "\n\n".join([doc.page_content for doc in documents])
|
| 122 |
|
| 123 |
-
#
|
| 124 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 125 |
st.session_state.messages.append({"role": "user", "content": prompt})
|
| 126 |
|
| 127 |
with st.chat_message("user"):
|
|
@@ -129,60 +108,14 @@ if uploaded_file:
|
|
| 129 |
|
| 130 |
try:
|
| 131 |
with st.spinner("Generating response..."):
|
| 132 |
-
|
| 133 |
-
|
| 134 |
-
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
"top_p": top_p,
|
| 140 |
-
"return_full_text": False
|
| 141 |
-
}
|
| 142 |
-
}
|
| 143 |
-
|
| 144 |
-
# Dynamically construct the API URL based on the selected model
|
| 145 |
-
api_url = f"https://api-inference.huggingface.co/models/{selected_model}"
|
| 146 |
-
logger.info(f"Selected model: {selected_model}, API URL: {api_url}")
|
| 147 |
-
|
| 148 |
-
# Query the Hugging Face API using the selected model
|
| 149 |
-
output = query(payload, api_url)
|
| 150 |
-
|
| 151 |
-
# Handle API response
|
| 152 |
-
if output is not None and isinstance(output, list) and len(output) > 0:
|
| 153 |
-
if 'generated_text' in output[0]:
|
| 154 |
-
assistant_response = output[0]['generated_text'].strip()
|
| 155 |
-
|
| 156 |
-
# Check for and remove duplicate responses
|
| 157 |
-
responses = assistant_response.split("\n</think>\n")
|
| 158 |
-
unique_response = responses[0].strip()
|
| 159 |
-
|
| 160 |
-
logger.info(f"Generated response: {unique_response}")
|
| 161 |
-
|
| 162 |
-
# Append response to chat only once
|
| 163 |
-
with st.chat_message("assistant"):
|
| 164 |
-
st.markdown(unique_response)
|
| 165 |
-
|
| 166 |
-
st.session_state.messages.append({"role": "assistant", "content": unique_response})
|
| 167 |
-
else:
|
| 168 |
-
logger.error(f"Unexpected API response structure: {output}")
|
| 169 |
-
st.error("Error: Unexpected response from the model. Please try again.")
|
| 170 |
-
else:
|
| 171 |
-
logger.error(f"Empty or invalid API response: {output}")
|
| 172 |
-
st.error("Error: Unable to generate a response. Please check the model and try again.")
|
| 173 |
|
| 174 |
except Exception as e:
|
| 175 |
logger.error(f"Application Error: {str(e)}", exc_info=True)
|
| 176 |
st.error(f"Application Error: {str(e)}")
|
| 177 |
-
|
| 178 |
-
# Allow user to ask a question based on extracted PDF content
|
| 179 |
-
if uploaded_file and documents: # Ensure documents exist before proceeding
|
| 180 |
-
if prompt := st.chat_input("Ask a question about the PDF content"):
|
| 181 |
-
context = "\n\n".join([doc.page_content for doc in documents]) # Get context from documents
|
| 182 |
-
answer = generate_response_with_langchain(prompt, context)
|
| 183 |
-
|
| 184 |
-
# Show the answer from LangChain model
|
| 185 |
-
with st.chat_message("assistant"):
|
| 186 |
-
st.markdown(answer)
|
| 187 |
-
|
| 188 |
-
st.session_state.messages.append({"role": "assistant", "content": answer})
|
|
|
|
| 13 |
logger = logging.getLogger(__name__)
|
| 14 |
|
| 15 |
# Page configuration
|
| 16 |
+
st.set_page_config(page_title="DeepSeek Chatbot - ruslanmv.com", page_icon="🤖", layout="centered")
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
|
| 18 |
# Initialize session state for chat history
|
| 19 |
if "messages" not in st.session_state:
|
|
|
|
| 25 |
st.markdown("[Get HuggingFace Token](https://huggingface.co/settings/tokens)")
|
| 26 |
|
| 27 |
# Dropdown to select model
|
| 28 |
+
model_options = ["deepseek-ai/DeepSeek-R1-Distill-Qwen-32B"]
|
|
|
|
|
|
|
| 29 |
selected_model = st.selectbox("Select Model", model_options, index=0)
|
| 30 |
|
| 31 |
+
system_message = st.text_area("System Message", value="You are a friendly chatbot. Provide clear, accurate, and brief answers.", height=100)
|
| 32 |
+
max_tokens = st.slider("Max Tokens", 10, 4000, 100)
|
| 33 |
+
temperature = st.slider("Temperature", 0.1, 4.0, 0.3)
|
| 34 |
+
top_p = st.slider("Top-p", 0.1, 1.0, 0.6)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 35 |
|
| 36 |
# Function to query the Hugging Face API
|
| 37 |
def query(payload, api_url):
|
|
|
|
| 57 |
documents = loader.load()
|
| 58 |
|
| 59 |
# Split the documents into chunks
|
| 60 |
+
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200, add_start_index=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
return text_splitter.split_documents(documents)
|
| 62 |
|
| 63 |
# Function to generate response using LangChain
|
|
|
|
| 94 |
documents = process_pdf(uploaded_file)
|
| 95 |
context = "\n\n".join([doc.page_content for doc in documents])
|
| 96 |
|
| 97 |
+
# Combine system message and user input into a single prompt
|
| 98 |
+
prompt_input = "Ask a question about the PDF content"
|
| 99 |
+
|
| 100 |
+
# Show the PDF-based question input if the PDF is uploaded
|
| 101 |
+
prompt = st.chat_input(prompt_input) if documents else None
|
| 102 |
+
|
| 103 |
+
if prompt:
|
| 104 |
st.session_state.messages.append({"role": "user", "content": prompt})
|
| 105 |
|
| 106 |
with st.chat_message("user"):
|
|
|
|
| 108 |
|
| 109 |
try:
|
| 110 |
with st.spinner("Generating response..."):
|
| 111 |
+
answer = generate_response_with_langchain(prompt, context)
|
| 112 |
+
|
| 113 |
+
# Show the answer from LangChain model
|
| 114 |
+
with st.chat_message("assistant"):
|
| 115 |
+
st.markdown(answer)
|
| 116 |
+
|
| 117 |
+
st.session_state.messages.append({"role": "assistant", "content": answer})
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 118 |
|
| 119 |
except Exception as e:
|
| 120 |
logger.error(f"Application Error: {str(e)}", exc_info=True)
|
| 121 |
st.error(f"Application Error: {str(e)}")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|