Update app.py
Browse files
app.py
CHANGED
|
@@ -11,40 +11,56 @@ import psutil
|
|
| 11 |
from pathlib import Path
|
| 12 |
import torch
|
| 13 |
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
page_icon="🛩️",
|
| 17 |
-
layout="wide",
|
| 18 |
-
initial_sidebar_state="expanded"
|
| 19 |
-
)
|
| 20 |
|
| 21 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
def setup_environment():
|
| 23 |
-
cache_dir = Path("/
|
| 24 |
-
|
|
|
|
|
|
|
|
|
|
| 25 |
return cache_dir
|
| 26 |
|
| 27 |
cache_dir = setup_environment()
|
| 28 |
|
| 29 |
-
# Load models with caching
|
| 30 |
@st.cache_resource(ttl=3600)
|
| 31 |
def load_models():
|
| 32 |
try:
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
)
|
| 45 |
-
}
|
| 46 |
except Exception as e:
|
| 47 |
-
st.error(f"
|
| 48 |
st.stop()
|
| 49 |
|
| 50 |
models = load_models()
|
|
@@ -60,7 +76,7 @@ def extract_text(file):
|
|
| 60 |
doc = docx.Document(file)
|
| 61 |
return "\n".join(para.text for para in doc.paragraphs if para.text)
|
| 62 |
except Exception as e:
|
| 63 |
-
st.error(f"
|
| 64 |
return ""
|
| 65 |
|
| 66 |
def generate_summary(text, max_length=150):
|
|
@@ -74,7 +90,7 @@ def generate_summary(text, max_length=150):
|
|
| 74 |
for chunk in chunks:
|
| 75 |
result = models['summarizer'](
|
| 76 |
chunk,
|
| 77 |
-
max_length=max(max_length
|
| 78 |
min_length=30,
|
| 79 |
do_sample=False
|
| 80 |
)
|
|
@@ -82,18 +98,13 @@ def generate_summary(text, max_length=150):
|
|
| 82 |
return " ".join(summaries)
|
| 83 |
return models['summarizer'](text, max_length=max_length)[0]['summary_text']
|
| 84 |
except Exception as e:
|
| 85 |
-
st.error(f"
|
| 86 |
return ""
|
| 87 |
|
| 88 |
-
# UI
|
| 89 |
-
st.title("
|
| 90 |
-
st.markdown(
|
| 91 |
-
"LexPilot™ ingests text, PDF, and Word files to instantly analyze contracts. "
|
| 92 |
-
"It delivers concise summaries and lets you ask targeted questions—giving fast, precise insights to speed up legal and procurement reviews."
|
| 93 |
-
)
|
| 94 |
|
| 95 |
-
with st.
|
| 96 |
-
st.header("Upload Document")
|
| 97 |
uploaded_file = st.file_uploader("Choose PDF/DOCX", type=["pdf", "docx"])
|
| 98 |
manual_text = st.text_area("Or paste text here:", height=150)
|
| 99 |
context = extract_text(uploaded_file) if uploaded_file else manual_text
|
|
@@ -111,14 +122,12 @@ with tab1:
|
|
| 111 |
question=question,
|
| 112 |
context=context[:100000]
|
| 113 |
)
|
| 114 |
-
st.success(f"Answered in {time.time()
|
| 115 |
st.markdown(f"**Answer:** {result['answer']}")
|
| 116 |
st.progress(result['score'])
|
| 117 |
st.caption(f"Confidence: {result['score']:.0%}")
|
| 118 |
except Exception as e:
|
| 119 |
-
st.error(f"
|
| 120 |
-
else:
|
| 121 |
-
st.info("Upload a document or paste text to enable question answering.")
|
| 122 |
|
| 123 |
with tab2:
|
| 124 |
if context and len(context.strip()) > 0:
|
|
@@ -129,20 +138,10 @@ with tab2:
|
|
| 129 |
start_time = time.time()
|
| 130 |
summary = generate_summary(context, length)
|
| 131 |
if summary:
|
| 132 |
-
st.success(f"Generated in {time.time()
|
| 133 |
st.markdown(f"**Summary:**\n\n{summary}")
|
| 134 |
-
else:
|
| 135 |
-
st.info("Upload a document or paste text to enable summarization.")
|
| 136 |
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
| 141 |
-
device_status = 'CPU (torch not configured)'
|
| 142 |
-
st.code(f"""
|
| 143 |
-
Models loaded: {', '.join(models.keys())}
|
| 144 |
-
Device: {device_status}
|
| 145 |
-
Memory usage: {psutil.virtual_memory().percent}%
|
| 146 |
-
CPU usage: {psutil.cpu_percent()}%
|
| 147 |
-
Cache location: {cache_dir}
|
| 148 |
-
""")
|
|
|
|
| 11 |
from pathlib import Path
|
| 12 |
import torch
|
| 13 |
|
| 14 |
+
# Page config with wide layout
|
| 15 |
+
st.set_page_config(page_title="LexPilot", layout="wide")
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
|
| 17 |
+
# Sidebar with project info
|
| 18 |
+
with st.sidebar:
|
| 19 |
+
st.title("LexPilot™")
|
| 20 |
+
st.markdown(
|
| 21 |
+
"""
|
| 22 |
+
LexPilot™ ingests text, PDF, and Word files to instantly analyze contracts.
|
| 23 |
+
It delivers concise summaries and lets you ask targeted questions—
|
| 24 |
+
giving fast, precise insights to speed up legal and procurement reviews.
|
| 25 |
+
"""
|
| 26 |
+
)
|
| 27 |
+
st.markdown("---")
|
| 28 |
+
st.write("### System Status")
|
| 29 |
+
try:
|
| 30 |
+
device_status = 'GPU ✅' if torch.cuda.is_available() else 'CPU ⚠️'
|
| 31 |
+
except:
|
| 32 |
+
device_status = 'CPU (torch not configured)'
|
| 33 |
+
st.text(f"Device: {device_status}")
|
| 34 |
+
st.text(f"Memory: {psutil.virtual_memory().percent}% used")
|
| 35 |
+
st.text(f"CPU: {psutil.cpu_percent()}% used")
|
| 36 |
+
|
| 37 |
+
# Setup cache directory for models
|
| 38 |
def setup_environment():
|
| 39 |
+
cache_dir = Path(".cache/models")
|
| 40 |
+
try:
|
| 41 |
+
cache_dir.mkdir(exist_ok=True, parents=True)
|
| 42 |
+
except Exception as e:
|
| 43 |
+
st.error(f"Failed to create cache directory: {e}")
|
| 44 |
return cache_dir
|
| 45 |
|
| 46 |
cache_dir = setup_environment()
|
| 47 |
|
|
|
|
| 48 |
@st.cache_resource(ttl=3600)
|
| 49 |
def load_models():
|
| 50 |
try:
|
| 51 |
+
qa_model = pipeline(
|
| 52 |
+
"question-answering",
|
| 53 |
+
model="distilbert-base-cased-distilled-squad",
|
| 54 |
+
device=-1
|
| 55 |
+
)
|
| 56 |
+
summarizer_model = pipeline(
|
| 57 |
+
"summarization",
|
| 58 |
+
model="sshleifer/distilbart-cnn-6-6",
|
| 59 |
+
device=-1
|
| 60 |
+
)
|
| 61 |
+
return {'qa': qa_model, 'summarizer': summarizer_model}
|
|
|
|
|
|
|
| 62 |
except Exception as e:
|
| 63 |
+
st.error(f"Failed to load models: {e}")
|
| 64 |
st.stop()
|
| 65 |
|
| 66 |
models = load_models()
|
|
|
|
| 76 |
doc = docx.Document(file)
|
| 77 |
return "\n".join(para.text for para in doc.paragraphs if para.text)
|
| 78 |
except Exception as e:
|
| 79 |
+
st.error(f"Error processing document: {e}")
|
| 80 |
return ""
|
| 81 |
|
| 82 |
def generate_summary(text, max_length=150):
|
|
|
|
| 90 |
for chunk in chunks:
|
| 91 |
result = models['summarizer'](
|
| 92 |
chunk,
|
| 93 |
+
max_length=max(max_length//len(chunks), 30),
|
| 94 |
min_length=30,
|
| 95 |
do_sample=False
|
| 96 |
)
|
|
|
|
| 98 |
return " ".join(summaries)
|
| 99 |
return models['summarizer'](text, max_length=max_length)[0]['summary_text']
|
| 100 |
except Exception as e:
|
| 101 |
+
st.error(f"Summarization failed: {e}")
|
| 102 |
return ""
|
| 103 |
|
| 104 |
+
# Main UI title
|
| 105 |
+
st.title("📄 LexPilot")
|
|
|
|
|
|
|
|
|
|
|
|
|
| 106 |
|
| 107 |
+
with st.expander("📤 Upload Document", expanded=True):
|
|
|
|
| 108 |
uploaded_file = st.file_uploader("Choose PDF/DOCX", type=["pdf", "docx"])
|
| 109 |
manual_text = st.text_area("Or paste text here:", height=150)
|
| 110 |
context = extract_text(uploaded_file) if uploaded_file else manual_text
|
|
|
|
| 122 |
question=question,
|
| 123 |
context=context[:100000]
|
| 124 |
)
|
| 125 |
+
st.success(f"Answered in {time.time()-start_time:.1f}s")
|
| 126 |
st.markdown(f"**Answer:** {result['answer']}")
|
| 127 |
st.progress(result['score'])
|
| 128 |
st.caption(f"Confidence: {result['score']:.0%}")
|
| 129 |
except Exception as e:
|
| 130 |
+
st.error(f"Question answering failed: {e}")
|
|
|
|
|
|
|
| 131 |
|
| 132 |
with tab2:
|
| 133 |
if context and len(context.strip()) > 0:
|
|
|
|
| 138 |
start_time = time.time()
|
| 139 |
summary = generate_summary(context, length)
|
| 140 |
if summary:
|
| 141 |
+
st.success(f"Generated in {time.time()-start_time:.1f}s")
|
| 142 |
st.markdown(f"**Summary:**\n\n{summary}")
|
|
|
|
|
|
|
| 143 |
|
| 144 |
+
# Show cache dir path in sidebar (optional)
|
| 145 |
+
with st.sidebar:
|
| 146 |
+
st.markdown("---")
|
| 147 |
+
st.write(f"Cache directory: {cache_dir}")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|