Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,64 +1,438 @@
|
|
1 |
-
import
|
2 |
-
from huggingface_hub import InferenceClient
|
3 |
-
|
4 |
-
"""
|
5 |
-
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
|
6 |
-
"""
|
7 |
-
client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
|
8 |
-
|
9 |
-
|
10 |
-
def respond(
|
11 |
-
message,
|
12 |
-
history: list[tuple[str, str]],
|
13 |
-
system_message,
|
14 |
-
max_tokens,
|
15 |
-
temperature,
|
16 |
-
top_p,
|
17 |
-
):
|
18 |
-
messages = [{"role": "system", "content": system_message}]
|
19 |
-
|
20 |
-
for val in history:
|
21 |
-
if val[0]:
|
22 |
-
messages.append({"role": "user", "content": val[0]})
|
23 |
-
if val[1]:
|
24 |
-
messages.append({"role": "assistant", "content": val[1]})
|
25 |
-
|
26 |
-
messages.append({"role": "user", "content": message})
|
27 |
-
|
28 |
-
response = ""
|
29 |
-
|
30 |
-
for message in client.chat_completion(
|
31 |
-
messages,
|
32 |
-
max_tokens=max_tokens,
|
33 |
-
stream=True,
|
34 |
-
temperature=temperature,
|
35 |
-
top_p=top_p,
|
36 |
-
):
|
37 |
-
token = message.choices[0].delta.content
|
38 |
-
|
39 |
-
response += token
|
40 |
-
yield response
|
41 |
-
|
42 |
-
|
43 |
-
"""
|
44 |
-
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
|
45 |
-
"""
|
46 |
-
demo = gr.ChatInterface(
|
47 |
-
respond,
|
48 |
-
additional_inputs=[
|
49 |
-
gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
|
50 |
-
gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
|
51 |
-
gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
|
52 |
-
gr.Slider(
|
53 |
-
minimum=0.1,
|
54 |
-
maximum=1.0,
|
55 |
-
value=0.95,
|
56 |
-
step=0.05,
|
57 |
-
label="Top-p (nucleus sampling)",
|
58 |
-
),
|
59 |
-
],
|
60 |
-
)
|
61 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
62 |
|
63 |
if __name__ == "__main__":
|
64 |
-
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
+
import boto3
|
4 |
+
|
5 |
+
import json
|
6 |
+
|
7 |
+
from qdrant_client import QdrantClient
|
8 |
+
|
9 |
+
from qdrant_client.http import models
|
10 |
+
|
11 |
+
import PyPDF2
|
12 |
+
|
13 |
+
import io
|
14 |
+
|
15 |
+
import uuid
|
16 |
+
|
17 |
+
# Simple function to connect to AWS Bedrock
|
18 |
+
|
19 |
+
def connect_to_bedrock():
|
20 |
+
|
21 |
+
client = boto3.client('bedrock-runtime', region_name='us-east-1')
|
22 |
+
|
23 |
+
return client
|
24 |
+
|
25 |
+
# Simple function to connect to QDrant Cloud
|
26 |
+
|
27 |
+
def connect_to_qdrant(api_key, url):
|
28 |
+
|
29 |
+
client = QdrantClient(url=url, api_key=api_key)
|
30 |
+
|
31 |
+
return client
|
32 |
+
|
33 |
+
# Extract text from PDF file
|
34 |
+
|
35 |
+
def extract_text_from_pdf(pdf_file):
|
36 |
+
|
37 |
+
pdf_reader = PyPDF2.PdfReader(pdf_file)
|
38 |
+
|
39 |
+
text = ""
|
40 |
+
|
41 |
+
for page in pdf_reader.pages:
|
42 |
+
|
43 |
+
text += page.extract_text() + "\n"
|
44 |
+
|
45 |
+
return text
|
46 |
+
|
47 |
+
# Split text into smaller chunks (simple way)
|
48 |
+
|
49 |
+
def split_text_into_chunks(text, chunk_size=1000):
|
50 |
+
|
51 |
+
words = text.split()
|
52 |
+
|
53 |
+
chunks = []
|
54 |
+
|
55 |
+
current_chunk = []
|
56 |
+
|
57 |
+
current_size = 0
|
58 |
+
|
59 |
+
for word in words:
|
60 |
+
|
61 |
+
current_chunk.append(word)
|
62 |
+
|
63 |
+
current_size += len(word) + 1 # +1 for space
|
64 |
+
|
65 |
+
if current_size >= chunk_size:
|
66 |
+
|
67 |
+
chunks.append(" ".join(current_chunk))
|
68 |
+
|
69 |
+
current_chunk = []
|
70 |
+
|
71 |
+
current_size = 0
|
72 |
+
|
73 |
+
if current_chunk: # Add last chunk if not empty
|
74 |
+
|
75 |
+
chunks.append(" ".join(current_chunk))
|
76 |
+
|
77 |
+
return chunks
|
78 |
+
|
79 |
+
# Get embeddings (vector numbers) from AI
|
80 |
+
|
81 |
+
def get_embeddings(bedrock_client, text):
|
82 |
+
|
83 |
+
body = json.dumps({
|
84 |
+
|
85 |
+
"inputText": text
|
86 |
+
|
87 |
+
})
|
88 |
+
|
89 |
+
response = bedrock_client.invoke_model(
|
90 |
+
|
91 |
+
modelId="amazon.titan-embed-text-v1",
|
92 |
+
|
93 |
+
body=body
|
94 |
+
|
95 |
+
)
|
96 |
+
|
97 |
+
result = json.loads(response['body'].read())
|
98 |
+
|
99 |
+
return result['embedding']
|
100 |
+
|
101 |
+
# Store PDF chunks in QDrant vector database
|
102 |
+
|
103 |
+
def store_pdf_in_qdrant(qdrant_client, bedrock_client, pdf_chunks, collection_name):
|
104 |
+
|
105 |
+
# Create collection if it doesn't exist
|
106 |
+
|
107 |
+
try:
|
108 |
+
|
109 |
+
qdrant_client.create_collection(
|
110 |
+
|
111 |
+
collection_name=collection_name,
|
112 |
+
|
113 |
+
vectors_config=models.VectorParams(size=1536, distance=models.Distance.COSINE)
|
114 |
+
|
115 |
+
)
|
116 |
+
|
117 |
+
except:
|
118 |
+
|
119 |
+
pass # Collection might already exist
|
120 |
+
|
121 |
+
# Store each chunk
|
122 |
+
|
123 |
+
points = []
|
124 |
+
|
125 |
+
for i, chunk in enumerate(pdf_chunks):
|
126 |
+
|
127 |
+
# Get vector representation of text
|
128 |
+
|
129 |
+
embedding = get_embeddings(bedrock_client, chunk)
|
130 |
+
|
131 |
+
# Create a point for QDrant
|
132 |
+
|
133 |
+
point = models.PointStruct(
|
134 |
+
|
135 |
+
id=str(uuid.uuid4()),
|
136 |
+
|
137 |
+
vector=embedding,
|
138 |
+
|
139 |
+
payload={"text": chunk, "chunk_id": i}
|
140 |
+
|
141 |
+
)
|
142 |
+
|
143 |
+
points.append(point)
|
144 |
+
|
145 |
+
# Upload to QDrant
|
146 |
+
|
147 |
+
qdrant_client.upsert(
|
148 |
+
|
149 |
+
collection_name=collection_name,
|
150 |
+
|
151 |
+
points=points
|
152 |
+
|
153 |
+
)
|
154 |
+
|
155 |
+
return len(points)
|
156 |
+
|
157 |
+
# Search for relevant text in QDrant
|
158 |
+
|
159 |
+
def search_in_qdrant(qdrant_client, bedrock_client, question, collection_name, top_k=3):
|
160 |
+
|
161 |
+
# Get vector for question
|
162 |
+
|
163 |
+
question_embedding = get_embeddings(bedrock_client, question)
|
164 |
+
|
165 |
+
# Search in QDrant
|
166 |
+
|
167 |
+
results = qdrant_client.search(
|
168 |
+
|
169 |
+
collection_name=collection_name,
|
170 |
+
|
171 |
+
query_vector=question_embedding,
|
172 |
+
|
173 |
+
limit=top_k
|
174 |
+
|
175 |
+
)
|
176 |
+
|
177 |
+
# Extract relevant text
|
178 |
+
|
179 |
+
relevant_texts = []
|
180 |
+
|
181 |
+
for result in results:
|
182 |
+
|
183 |
+
relevant_texts.append(result.payload["text"])
|
184 |
+
|
185 |
+
return relevant_texts
|
186 |
+
|
187 |
+
# Ask AI to answer question based on PDF content
|
188 |
+
|
189 |
+
def ask_ai_with_context(bedrock_client, question, relevant_texts):
|
190 |
+
|
191 |
+
context = "\n\n".join(relevant_texts)
|
192 |
+
|
193 |
+
prompt = f"""
|
194 |
+
|
195 |
+
Based on the following information from a PDF document, please answer the question.
|
196 |
+
|
197 |
+
PDF Content:
|
198 |
+
|
199 |
+
{context}
|
200 |
+
|
201 |
+
Question: {question}
|
202 |
+
|
203 |
+
Please provide a clear and helpful answer based only on the information provided above.
|
204 |
+
|
205 |
+
If the answer is not in the provided content, please say so.
|
206 |
+
|
207 |
+
"""
|
208 |
+
|
209 |
+
body = json.dumps({
|
210 |
+
|
211 |
+
"anthropic_version": "bedrock-2023-05-31",
|
212 |
+
|
213 |
+
"max_tokens": 500,
|
214 |
+
|
215 |
+
"messages": [{"role": "user", "content": prompt}]
|
216 |
+
|
217 |
+
})
|
218 |
+
|
219 |
+
response = bedrock_client.invoke_model(
|
220 |
+
|
221 |
+
modelId="anthropic.claude-3-haiku-20240307-v1:0",
|
222 |
+
|
223 |
+
body=body
|
224 |
+
|
225 |
+
)
|
226 |
+
|
227 |
+
result = json.loads(response['body'].read())
|
228 |
+
|
229 |
+
return result['content'][0]['text']
|
230 |
+
|
231 |
+
# Main app
|
232 |
+
|
233 |
+
def main():
|
234 |
+
|
235 |
+
st.title("π Simple PDF Chatbot")
|
236 |
+
|
237 |
+
st.write("Upload a PDF and ask questions about it!")
|
238 |
+
|
239 |
+
# Sidebar for settings
|
240 |
+
|
241 |
+
with st.sidebar:
|
242 |
+
|
243 |
+
st.subheader("π§ Setup")
|
244 |
+
|
245 |
+
st.write("You need these to use the app:")
|
246 |
+
|
247 |
+
# QDrant settings
|
248 |
+
|
249 |
+
st.write("**QDrant Cloud Settings:**")
|
250 |
+
|
251 |
+
qdrant_url = st.text_input("QDrant URL", placeholder="https://your-cluster.qdrant.io")
|
252 |
+
|
253 |
+
qdrant_api_key = st.text_input("QDrant API Key", type="password")
|
254 |
+
|
255 |
+
st.write("**Collection Name:**")
|
256 |
+
|
257 |
+
collection_name = st.text_input("Collection Name", value="pdf_documents")
|
258 |
+
|
259 |
+
st.markdown("---")
|
260 |
+
|
261 |
+
st.markdown("""
|
262 |
+
|
263 |
+
**How to get QDrant settings:**
|
264 |
+
|
265 |
+
1. Go to qdrant.io
|
266 |
+
|
267 |
+
2. Create free account
|
268 |
+
|
269 |
+
3. Create a cluster
|
270 |
+
|
271 |
+
4. Copy URL and API key
|
272 |
+
|
273 |
+
""")
|
274 |
+
|
275 |
+
# Main content
|
276 |
+
|
277 |
+
tab1, tab2 = st.tabs(["π€ Upload PDF", "π¬ Chat with PDF"])
|
278 |
+
|
279 |
+
with tab1:
|
280 |
+
|
281 |
+
st.subheader("Upload Your PDF")
|
282 |
+
|
283 |
+
uploaded_file = st.file_uploader("Choose a PDF file", type="pdf")
|
284 |
+
|
285 |
+
if uploaded_file and qdrant_url and qdrant_api_key:
|
286 |
+
|
287 |
+
if st.button("π Process PDF"):
|
288 |
+
|
289 |
+
try:
|
290 |
+
|
291 |
+
with st.spinner("Processing your PDF..."):
|
292 |
+
|
293 |
+
# Connect to services
|
294 |
+
|
295 |
+
bedrock_client = connect_to_bedrock()
|
296 |
+
|
297 |
+
qdrant_client = connect_to_qdrant(qdrant_api_key, qdrant_url)
|
298 |
+
|
299 |
+
# Extract text from PDF
|
300 |
+
|
301 |
+
st.write("π Extracting text from PDF...")
|
302 |
+
|
303 |
+
pdf_text = extract_text_from_pdf(uploaded_file)
|
304 |
+
|
305 |
+
# Split into chunks
|
306 |
+
|
307 |
+
st.write("βοΈ Breaking text into smaller pieces...")
|
308 |
+
|
309 |
+
chunks = split_text_into_chunks(pdf_text)
|
310 |
+
|
311 |
+
# Store in QDrant
|
312 |
+
|
313 |
+
st.write("πΎ Storing in vector database...")
|
314 |
+
|
315 |
+
num_chunks = store_pdf_in_qdrant(qdrant_client, bedrock_client, chunks, collection_name)
|
316 |
+
|
317 |
+
st.success(f"β
PDF processed successfully! Stored {num_chunks} text chunks.")
|
318 |
+
|
319 |
+
st.balloons()
|
320 |
+
|
321 |
+
except Exception as e:
|
322 |
+
|
323 |
+
st.error(f"β Error processing PDF: {str(e)}")
|
324 |
+
|
325 |
+
elif uploaded_file:
|
326 |
+
|
327 |
+
st.warning("β οΈ Please enter QDrant settings in the sidebar first!")
|
328 |
+
|
329 |
+
with tab2:
|
330 |
+
|
331 |
+
st.subheader("Ask Questions About Your PDF")
|
332 |
+
|
333 |
+
if qdrant_url and qdrant_api_key:
|
334 |
+
|
335 |
+
question = st.text_input("π What would you like to know about your PDF?")
|
336 |
+
|
337 |
+
if question:
|
338 |
+
|
339 |
+
if st.button("π Get Answer"):
|
340 |
+
|
341 |
+
try:
|
342 |
+
|
343 |
+
with st.spinner("Searching for answer..."):
|
344 |
+
|
345 |
+
# Connect to services
|
346 |
+
|
347 |
+
bedrock_client = connect_to_bedrock()
|
348 |
+
|
349 |
+
qdrant_client = connect_to_qdrant(qdrant_api_key, qdrant_url)
|
350 |
+
|
351 |
+
# Search for relevant content
|
352 |
+
|
353 |
+
st.write("π Searching relevant content...")
|
354 |
+
|
355 |
+
relevant_texts = search_in_qdrant(qdrant_client, bedrock_client, question, collection_name)
|
356 |
+
|
357 |
+
# Get AI answer
|
358 |
+
|
359 |
+
st.write("π€ Generating answer...")
|
360 |
+
|
361 |
+
answer = ask_ai_with_context(bedrock_client, question, relevant_texts)
|
362 |
+
|
363 |
+
# Show answer
|
364 |
+
|
365 |
+
st.subheader("π Answer:")
|
366 |
+
|
367 |
+
st.write(answer)
|
368 |
+
|
369 |
+
# Show sources (optional)
|
370 |
+
|
371 |
+
with st.expander("π Source content used"):
|
372 |
+
|
373 |
+
for i, text in enumerate(relevant_texts, 1):
|
374 |
+
|
375 |
+
st.write(f"**Source {i}:**")
|
376 |
+
|
377 |
+
st.write(text[:200] + "..." if len(text) > 200 else text)
|
378 |
+
|
379 |
+
st.write("---")
|
380 |
+
|
381 |
+
except Exception as e:
|
382 |
+
|
383 |
+
st.error(f"β Error: {str(e)}")
|
384 |
+
|
385 |
+
else:
|
386 |
+
|
387 |
+
st.warning("β οΈ Please enter QDrant settings in the sidebar first!")
|
388 |
+
|
389 |
+
# Quick setup guide
|
390 |
+
|
391 |
+
def show_setup_guide():
|
392 |
+
|
393 |
+
with st.expander("π Quick Setup Guide"):
|
394 |
+
|
395 |
+
st.markdown("""
|
396 |
+
|
397 |
+
**Step 1: Install Required Libraries**
|
398 |
+
|
399 |
+
```bash
|
400 |
+
|
401 |
+
pip install streamlit boto3 qdrant-client PyPDF2
|
402 |
+
|
403 |
+
```
|
404 |
+
|
405 |
+
**Step 2: Set up AWS**
|
406 |
+
|
407 |
+
- Create AWS account
|
408 |
+
|
409 |
+
- Run `aws configure` and enter your keys
|
410 |
+
|
411 |
+
**Step 3: Set up QDrant Cloud**
|
412 |
+
|
413 |
+
- Go to qdrant.io
|
414 |
+
|
415 |
+
- Create free account
|
416 |
+
|
417 |
+
- Create a cluster
|
418 |
+
|
419 |
+
- Copy URL and API key to sidebar
|
420 |
+
|
421 |
+
**Step 4: Run the App**
|
422 |
+
|
423 |
+
```bash
|
424 |
+
|
425 |
+
streamlit run pdf_chatbot.py
|
426 |
+
|
427 |
+
```
|
428 |
+
|
429 |
+
""")
|
430 |
+
|
431 |
+
# Run the app
|
432 |
|
433 |
if __name__ == "__main__":
|
434 |
+
|
435 |
+
show_setup_guide()
|
436 |
+
|
437 |
+
main()
|
438 |
+
|