Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
import streamlit as st
|
2 |
import pandas as pd
|
3 |
import os
|
4 |
-
import
|
5 |
|
6 |
# Import evaluation modules
|
7 |
from phoenix_code import phoenix_eval
|
@@ -16,106 +16,89 @@ st.set_page_config(
|
|
16 |
initial_sidebar_state="expanded"
|
17 |
)
|
18 |
|
19 |
-
#
|
20 |
-
def
|
21 |
-
|
22 |
-
|
23 |
-
"""
|
24 |
-
templates = {
|
25 |
-
"Phoenix": pd.DataFrame({
|
26 |
-
'question': ['What is machine learning?', 'Explain Python programming'],
|
27 |
-
'answer': ['Machine learning is...', 'Python is a programming language...'],
|
28 |
-
'cleaned_context': ['Context about machine learning', 'Context about Python programming']
|
29 |
-
}),
|
30 |
-
"RAGAS": pd.DataFrame({
|
31 |
-
'question': ['What is AI?', 'Describe data science'],
|
32 |
-
'answer': ['Artificial Intelligence is...', 'Data science involves...'],
|
33 |
-
'contexts': ['Detailed context about AI', 'Comprehensive context on data science'],
|
34 |
-
'ground_truth': ['Verified definition of AI', 'Verified explanation of data science']
|
35 |
-
}),
|
36 |
-
"Traditional Metrics": pd.DataFrame({
|
37 |
-
'question': ['What is deep learning?', 'Explain neural networks'],
|
38 |
-
'answer': ['Deep learning is...', 'Neural networks are...'],
|
39 |
-
'contexts': ['Context about deep learning', 'Context about neural networks']
|
40 |
-
})
|
41 |
-
}
|
42 |
-
|
43 |
-
return templates.get(evaluation_type, pd.DataFrame())
|
44 |
|
45 |
-
# Function to create a
|
46 |
-
def
|
47 |
-
"""
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
|
|
|
|
|
|
|
|
52 |
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
workbook = writer.book
|
60 |
-
worksheet = workbook.add_worksheet('README')
|
61 |
-
|
62 |
-
# Write column descriptions
|
63 |
-
readme_text = [
|
64 |
-
f"Sample Template for {evaluation_type} Evaluation",
|
65 |
-
"",
|
66 |
-
"Column Descriptions:",
|
67 |
-
]
|
68 |
-
|
69 |
-
if evaluation_type == "Phoenix":
|
70 |
-
readme_text.extend([
|
71 |
-
"- 'question': The input query or prompt",
|
72 |
-
"- 'answer': The generated response to the question",
|
73 |
-
"- 'cleaned_context': The context used to generate the answer"
|
74 |
-
])
|
75 |
-
elif evaluation_type == "RAGAS":
|
76 |
-
readme_text.extend([
|
77 |
-
"- 'question': The input query or prompt",
|
78 |
-
"- 'answer': The generated response to the question",
|
79 |
-
"- 'contexts': The context used to generate the answer",
|
80 |
-
"- 'ground_truth': The verified or gold standard answer"
|
81 |
-
])
|
82 |
-
else: # Traditional Metrics
|
83 |
-
readme_text.extend([
|
84 |
-
"- 'question': The input query or prompt",
|
85 |
-
"- 'answer': The generated response to the question",
|
86 |
-
"- 'contexts': The context used to generate the answer"
|
87 |
-
])
|
88 |
-
|
89 |
-
# Write README text
|
90 |
-
for i, line in enumerate(readme_text):
|
91 |
-
worksheet.write(i, 0, line)
|
92 |
-
|
93 |
-
output.seek(0)
|
94 |
-
return output
|
95 |
|
|
|
96 |
def main():
|
97 |
-
# Custom CSS
|
98 |
st.markdown("""
|
99 |
<style>
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
border-radius: 10px;
|
104 |
-
padding: 20px;
|
105 |
-
margin-bottom: 20px;
|
106 |
-
border: 1px solid #E2E8F0;
|
107 |
}
|
108 |
-
.
|
109 |
color: #2C3E50;
|
110 |
-
|
|
|
|
|
|
|
|
|
111 |
}
|
112 |
-
.
|
113 |
-
background-color: #
|
114 |
color: white;
|
115 |
border: none;
|
116 |
border-radius: 6px;
|
117 |
padding: 10px 20px;
|
118 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
119 |
}
|
120 |
</style>
|
121 |
""", unsafe_allow_html=True)
|
@@ -123,49 +106,10 @@ def main():
|
|
123 |
# App Title
|
124 |
st.markdown("<h1 class='stTitle'>π RAG Evaluation Toolkit</h1>", unsafe_allow_html=True)
|
125 |
|
126 |
-
#
|
127 |
-
st.markdown("<div class='template-section'>", unsafe_allow_html=True)
|
128 |
-
st.markdown("<h2 class='template-header'>π Data Template Guidelines</h2>", unsafe_allow_html=True)
|
129 |
-
|
130 |
-
# Expandable sections for each evaluation type
|
131 |
-
with st.expander("π Phoenix Evaluation Template"):
|
132 |
-
st.write("Required Columns: 'question', 'answer', 'cleaned_context'")
|
133 |
-
if st.button("Download Phoenix Template", key="phoenix_template"):
|
134 |
-
phoenix_template = create_downloadable_excel("Phoenix")
|
135 |
-
st.download_button(
|
136 |
-
label="Save Phoenix Template",
|
137 |
-
data=phoenix_template,
|
138 |
-
file_name="phoenix_evaluation_template.xlsx",
|
139 |
-
mime="application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
|
140 |
-
)
|
141 |
-
|
142 |
-
with st.expander("π RAGAS Evaluation Template"):
|
143 |
-
st.write("Required Columns: 'question', 'answer', 'contexts', 'ground_truth'")
|
144 |
-
if st.button("Download RAGAS Template", key="ragas_template"):
|
145 |
-
ragas_template = create_downloadable_excel("RAGAS")
|
146 |
-
st.download_button(
|
147 |
-
label="Save RAGAS Template",
|
148 |
-
data=ragas_template,
|
149 |
-
file_name="ragas_evaluation_template.xlsx",
|
150 |
-
mime="application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
|
151 |
-
)
|
152 |
-
|
153 |
-
with st.expander("π Traditional Metrics Template"):
|
154 |
-
st.write("Required Columns: 'question', 'answer', 'contexts'")
|
155 |
-
if st.button("Download Traditional Metrics Template", key="traditional_template"):
|
156 |
-
traditional_template = create_downloadable_excel("Traditional Metrics")
|
157 |
-
st.download_button(
|
158 |
-
label="Save Traditional Metrics Template",
|
159 |
-
data=traditional_template,
|
160 |
-
file_name="traditional_metrics_template.xlsx",
|
161 |
-
mime="application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
|
162 |
-
)
|
163 |
-
st.markdown("</div>", unsafe_allow_html=True)
|
164 |
-
|
165 |
-
# Sidebar for Configuration (keep previous sidebar code)
|
166 |
st.sidebar.header("π Evaluation Configuration")
|
167 |
|
168 |
-
# API Key Input
|
169 |
st.sidebar.subheader("OpenAI API Key")
|
170 |
openai_api_key = st.sidebar.text_input(
|
171 |
"Enter your OpenAI API Key",
|
@@ -175,13 +119,7 @@ def main():
|
|
175 |
|
176 |
# File Upload Section
|
177 |
st.markdown("### π Upload Your Dataset")
|
178 |
-
uploaded_file =
|
179 |
-
"Upload Dataset",
|
180 |
-
type=["csv", "xls", "xlsx"]
|
181 |
-
)
|
182 |
-
|
183 |
-
# Rest of the previous implementation follows...
|
184 |
-
# (Keep the rest of the previous main() function code)
|
185 |
|
186 |
# Evaluation Type Selection
|
187 |
st.sidebar.subheader("π Evaluation Methods")
|
@@ -220,42 +158,72 @@ def main():
|
|
220 |
metrics
|
221 |
)
|
222 |
|
223 |
-
# Validation function for DataFrame columns
|
224 |
-
def validate_dataframe(df, evaluation_type):
|
225 |
-
"""
|
226 |
-
Validate DataFrame columns based on the evaluation type
|
227 |
-
"""
|
228 |
-
required_columns = {
|
229 |
-
"Phoenix": ['question', 'answer', 'cleaned_context'],
|
230 |
-
"RAGAS": ['question', 'answer', 'contexts', 'ground_truth'],
|
231 |
-
"Traditional Metrics": ['question', 'answer', 'contexts']
|
232 |
-
}
|
233 |
-
|
234 |
-
# Check columns for the selected evaluation methods
|
235 |
-
for method in selected_metrics.keys():
|
236 |
-
missing_columns = [col for col in required_columns.get(method, []) if col not in df.columns]
|
237 |
-
|
238 |
-
if missing_columns:
|
239 |
-
st.error(f"Missing required columns for {method}: {', '.join(missing_columns)}")
|
240 |
-
return False
|
241 |
-
|
242 |
-
return True
|
243 |
-
|
244 |
# Evaluation Button
|
245 |
if uploaded_file and openai_api_key and selected_metrics:
|
246 |
-
|
247 |
-
|
248 |
-
|
249 |
-
|
250 |
-
|
251 |
-
|
252 |
-
|
253 |
-
|
254 |
-
|
255 |
-
|
256 |
-
|
257 |
-
|
258 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
259 |
|
260 |
# Run the app
|
261 |
if __name__ == "__main__":
|
|
|
1 |
import streamlit as st
|
2 |
import pandas as pd
|
3 |
import os
|
4 |
+
import base64
|
5 |
|
6 |
# Import evaluation modules
|
7 |
from phoenix_code import phoenix_eval
|
|
|
16 |
initial_sidebar_state="expanded"
|
17 |
)
|
18 |
|
19 |
+
# Custom CSS for improved styling
|
20 |
+
def local_css(file_name):
|
21 |
+
with open(file_name) as f:
|
22 |
+
st.markdown(f'<style>{f.read()}</style>', unsafe_allow_html=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
|
24 |
+
# Function to create a more visually appealing file uploader
|
25 |
+
def custom_file_uploader():
|
26 |
+
st.markdown("""
|
27 |
+
<div class="file-upload-container">
|
28 |
+
<div class="file-upload-icon">π</div>
|
29 |
+
<div class="file-upload-text">
|
30 |
+
Drag and Drop or <span class="file-upload-browse">Browse Files</span>
|
31 |
+
</div>
|
32 |
+
<small>Supports CSV, XLS, XLSX</small>
|
33 |
+
</div>
|
34 |
+
""", unsafe_allow_html=True)
|
35 |
|
36 |
+
uploaded_file = st.file_uploader(
|
37 |
+
"Upload Dataset",
|
38 |
+
type=["csv", "xls", "xlsx"],
|
39 |
+
label_visibility="collapsed"
|
40 |
+
)
|
41 |
+
return uploaded_file
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
42 |
|
43 |
+
# Main Streamlit App
|
44 |
def main():
|
45 |
+
# Custom CSS for enhanced styling
|
46 |
st.markdown("""
|
47 |
<style>
|
48 |
+
.stApp {
|
49 |
+
background-color: #f0f2f6;
|
50 |
+
font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
|
|
|
|
|
|
|
|
|
51 |
}
|
52 |
+
.stTitle {
|
53 |
color: #2C3E50;
|
54 |
+
text-align: center;
|
55 |
+
margin-bottom: 30px;
|
56 |
+
}
|
57 |
+
.stMarkdown {
|
58 |
+
color: #34495E;
|
59 |
}
|
60 |
+
.stButton>button {
|
61 |
+
background-color: #3498DB;
|
62 |
color: white;
|
63 |
border: none;
|
64 |
border-radius: 6px;
|
65 |
padding: 10px 20px;
|
66 |
+
transition: all 0.3s ease;
|
67 |
+
}
|
68 |
+
.stButton>button:hover {
|
69 |
+
background-color: #2980B9;
|
70 |
+
transform: scale(1.05);
|
71 |
+
}
|
72 |
+
.sidebar .sidebar-content {
|
73 |
+
background-color: #FFFFFF;
|
74 |
+
border-radius: 10px;
|
75 |
+
padding: 20px;
|
76 |
+
box-shadow: 0 4px 6px rgba(0,0,0,0.1);
|
77 |
+
}
|
78 |
+
.file-upload-container {
|
79 |
+
border: 2px dashed #3498DB;
|
80 |
+
border-radius: 10px;
|
81 |
+
padding: 30px;
|
82 |
+
text-align: center;
|
83 |
+
background-color: #FFFFFF;
|
84 |
+
transition: all 0.3s ease;
|
85 |
+
}
|
86 |
+
.file-upload-container:hover {
|
87 |
+
border-color: #2980B9;
|
88 |
+
background-color: #F1F8FF;
|
89 |
+
}
|
90 |
+
.file-upload-icon {
|
91 |
+
font-size: 50px;
|
92 |
+
color: #3498DB;
|
93 |
+
margin-bottom: 15px;
|
94 |
+
}
|
95 |
+
.file-upload-text {
|
96 |
+
color: #2C3E50;
|
97 |
+
font-size: 18px;
|
98 |
+
}
|
99 |
+
.file-upload-browse {
|
100 |
+
color: #3498DB;
|
101 |
+
font-weight: bold;
|
102 |
}
|
103 |
</style>
|
104 |
""", unsafe_allow_html=True)
|
|
|
106 |
# App Title
|
107 |
st.markdown("<h1 class='stTitle'>π RAG Evaluation Toolkit</h1>", unsafe_allow_html=True)
|
108 |
|
109 |
+
# Sidebar for Configuration
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
110 |
st.sidebar.header("π Evaluation Configuration")
|
111 |
|
112 |
+
# API Key Input with improved styling
|
113 |
st.sidebar.subheader("OpenAI API Key")
|
114 |
openai_api_key = st.sidebar.text_input(
|
115 |
"Enter your OpenAI API Key",
|
|
|
119 |
|
120 |
# File Upload Section
|
121 |
st.markdown("### π Upload Your Dataset")
|
122 |
+
uploaded_file = custom_file_uploader()
|
|
|
|
|
|
|
|
|
|
|
|
|
123 |
|
124 |
# Evaluation Type Selection
|
125 |
st.sidebar.subheader("π Evaluation Methods")
|
|
|
158 |
metrics
|
159 |
)
|
160 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
161 |
# Evaluation Button
|
162 |
if uploaded_file and openai_api_key and selected_metrics:
|
163 |
+
if st.button("π Run Evaluation"):
|
164 |
+
# Load data
|
165 |
+
file_extension = os.path.splitext(uploaded_file.name)[1]
|
166 |
+
if file_extension.lower() == ".csv":
|
167 |
+
df = pd.read_csv(uploaded_file)
|
168 |
+
elif file_extension.lower() in [".xls", ".xlsx"]:
|
169 |
+
df = pd.read_excel(uploaded_file)
|
170 |
+
|
171 |
+
# Combine results
|
172 |
+
combined_results = pd.DataFrame()
|
173 |
+
|
174 |
+
# Progress bar
|
175 |
+
progress_bar = st.progress(0)
|
176 |
+
|
177 |
+
# Run evaluations
|
178 |
+
with st.spinner("Processing evaluations..."):
|
179 |
+
# Phoenix Evaluation
|
180 |
+
if "Phoenix Evaluation" in selected_metrics:
|
181 |
+
progress_bar.progress(33)
|
182 |
+
phoenix_results = phoenix_eval(
|
183 |
+
selected_metrics.get("Phoenix Evaluation", []),
|
184 |
+
openai_api_key,
|
185 |
+
df.copy()
|
186 |
+
)
|
187 |
+
combined_results = pd.concat([combined_results, phoenix_results], axis=1)
|
188 |
+
|
189 |
+
# RAGAS Evaluation
|
190 |
+
if "RAGAS Evaluation" in selected_metrics:
|
191 |
+
progress_bar.progress(66)
|
192 |
+
ragas_results = ragas_eval(
|
193 |
+
selected_metrics.get("RAGAS Evaluation", []),
|
194 |
+
openai_api_key,
|
195 |
+
df.copy()
|
196 |
+
)
|
197 |
+
combined_results = pd.concat([combined_results, ragas_results], axis=1)
|
198 |
+
|
199 |
+
# Traditional Metrics Evaluation
|
200 |
+
if "Traditional Metrics" in selected_metrics:
|
201 |
+
progress_bar.progress(100)
|
202 |
+
traditional_results = RAGEvaluator(
|
203 |
+
df=df.copy(),
|
204 |
+
selected_metrics=selected_metrics.get("Traditional Metrics", [])
|
205 |
+
)
|
206 |
+
combined_results = pd.concat([combined_results, traditional_results], axis=1)
|
207 |
+
|
208 |
+
# Save results
|
209 |
+
results_filename = "rag_evaluation_results.xlsx"
|
210 |
+
combined_results.to_excel(results_filename, index=False)
|
211 |
+
|
212 |
+
# Success message and download button
|
213 |
+
st.success("Evaluation Completed Successfully!")
|
214 |
+
|
215 |
+
# Create download button with improved styling
|
216 |
+
with open(results_filename, "rb") as file:
|
217 |
+
btn = st.download_button(
|
218 |
+
label="π₯ Download Evaluation Results",
|
219 |
+
data=file,
|
220 |
+
file_name=results_filename,
|
221 |
+
mime="application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
|
222 |
+
)
|
223 |
+
|
224 |
+
# Display results preview
|
225 |
+
st.markdown("### π Results Preview")
|
226 |
+
st.dataframe(combined_results)
|
227 |
|
228 |
# Run the app
|
229 |
if __name__ == "__main__":
|