Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -39,7 +39,6 @@ from appStore.crs_utils import lookup_crs_value
|
|
39 |
###########################################
|
40 |
# Model Config
|
41 |
###########################################
|
42 |
-
|
43 |
config = configparser.ConfigParser()
|
44 |
config.read('model_params.cfg')
|
45 |
|
@@ -123,13 +122,28 @@ country_name_mapping, iso_code_to_sub_region = get_country_name_and_region_mappi
|
|
123 |
)
|
124 |
unique_country_names = sorted(country_name_mapping.keys())
|
125 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
126 |
###########################################
|
127 |
# Filter Controls - Row 1
|
128 |
###########################################
|
129 |
col1, col2, col3, col4, col5 = st.columns([1, 1, 1, 1, 1])
|
130 |
|
131 |
with col1:
|
132 |
-
region_filter = st.selectbox("Region", ["All/Not allocated"] + sorted(unique_sub_regions))
|
133 |
|
134 |
if region_filter == "All/Not allocated":
|
135 |
filtered_country_names = unique_country_names
|
@@ -140,7 +154,7 @@ else:
|
|
140 |
]
|
141 |
|
142 |
with col2:
|
143 |
-
country_filter = st.selectbox("Country", ["All/Not allocated"] + filtered_country_names)
|
144 |
|
145 |
with col3:
|
146 |
current_year = datetime.now().year
|
@@ -149,33 +163,32 @@ with col3:
|
|
149 |
"Project End Year",
|
150 |
min_value=2010,
|
151 |
max_value=max_end_year,
|
152 |
-
value=(default_start_year, max_end_year)
|
|
|
153 |
)
|
154 |
|
155 |
with col4:
|
156 |
crs_options = ["All/Not allocated"] + get_crs_options(client, collection_name)
|
157 |
-
crs_filter = st.selectbox("CRS", crs_options)
|
158 |
|
159 |
with col5:
|
160 |
min_budget = st.slider(
|
161 |
"Minimum Project Budget (Million €)",
|
162 |
min_value=min_budget_val,
|
163 |
max_value=max_budget_val,
|
164 |
-
value=min_budget_val
|
|
|
165 |
)
|
166 |
|
167 |
|
168 |
-
|
169 |
###########################################
|
170 |
# Filter Controls - Row 2 (Additional Filters)
|
171 |
###########################################
|
172 |
col1_2, col2_2, col3_2, col4_2, col5_2 = st.columns(5)
|
173 |
|
174 |
with col1_2:
|
175 |
-
# Get unique clients from project_data; adjust column name as needed.
|
176 |
client_options = sorted(project_data["client"].dropna().unique().tolist())
|
177 |
-
client_filter = st.selectbox("Client", ["All/Not allocated"] + client_options)
|
178 |
-
# Columns 2-4 reserved (empty)
|
179 |
with col2_2:
|
180 |
st.empty()
|
181 |
with col3_2:
|
@@ -183,10 +196,11 @@ with col3_2:
|
|
183 |
with col4_2:
|
184 |
st.empty()
|
185 |
with col5_2:
|
186 |
-
|
187 |
-
|
188 |
|
189 |
show_exact_matches = st.checkbox("Show only exact matches", value=False)
|
|
|
190 |
###########################################
|
191 |
# Main Search / Results
|
192 |
###########################################
|
@@ -230,7 +244,7 @@ else:
|
|
230 |
get_country_name
|
231 |
)
|
232 |
|
233 |
-
# Additional filter by client
|
234 |
if client_filter != "All/Not allocated":
|
235 |
filtered_semantic = [r for r in filtered_semantic if r.payload.get("metadata", {}).get("client", "Unknown Client") == client_filter]
|
236 |
filtered_lexical = [r for r in filtered_lexical if r.payload.get("metadata", {}).get("client", "Unknown Client") == client_filter]
|
@@ -240,9 +254,6 @@ else:
|
|
240 |
filtered_lexical_no_dupe = remove_duplicates(filtered_lexical)
|
241 |
|
242 |
def format_currency(value):
|
243 |
-
"""
|
244 |
-
Format a numeric or string value as currency (EUR) with commas.
|
245 |
-
"""
|
246 |
try:
|
247 |
return f"€{int(float(value)):,}"
|
248 |
except (ValueError, TypeError):
|
@@ -250,8 +261,7 @@ else:
|
|
250 |
|
251 |
# 3) Display results
|
252 |
if show_exact_matches:
|
253 |
-
|
254 |
-
st.write("Showing **Top Lexical Search results**") # MOD: Removed RAG answer for lexical search
|
255 |
query_substring = var.strip().lower()
|
256 |
lexical_substring_filtered = [
|
257 |
r for r in filtered_lexical
|
@@ -261,18 +271,15 @@ else:
|
|
261 |
if not filtered_lexical_no_dupe:
|
262 |
st.write('No exact matches, consider unchecking "Show only exact matches"')
|
263 |
else:
|
264 |
-
#
|
265 |
-
top_results = filtered_lexical_no_dupe # or use [:15] if you want to limit
|
266 |
for res in top_results:
|
267 |
metadata = res.payload.get('metadata', {})
|
268 |
if "title" not in metadata:
|
269 |
metadata["title"] = compute_title(metadata)
|
270 |
-
# MOD: Remove hyperlink from title by stripping <a> tags.
|
271 |
title_html = highlight_query(metadata["title"], var) if var.strip() else metadata["title"]
|
272 |
title_clean = re.sub(r'<a.*?>|</a>', '', title_html)
|
273 |
st.markdown(f"#### {title_clean}", unsafe_allow_html=True)
|
274 |
|
275 |
-
# Description snippet
|
276 |
objective = metadata.get("objective", "None")
|
277 |
desc_en = metadata.get("description.en", "").strip()
|
278 |
desc_de = metadata.get("description.de", "").strip()
|
@@ -316,30 +323,28 @@ else:
|
|
316 |
contact = metadata.get("contact", "").strip()
|
317 |
if contact and contact.lower() != "transparenz@giz.de":
|
318 |
additional_text += f"<br>**Contact:** xxx@giz.de"
|
319 |
-
|
320 |
st.markdown(additional_text, unsafe_allow_html=True)
|
321 |
|
322 |
st.divider()
|
323 |
-
|
324 |
else:
|
325 |
-
# Semantic results
|
326 |
if not filtered_semantic_no_dupe:
|
327 |
st.write("No relevant results found.")
|
328 |
else:
|
329 |
-
# MOD: Pagination for semantic search results
|
330 |
page_size = 15
|
331 |
total_results = len(filtered_semantic_no_dupe)
|
332 |
total_pages = (total_results - 1) // page_size + 1
|
333 |
-
|
334 |
-
|
335 |
-
|
336 |
-
page = 1
|
337 |
-
|
|
|
|
|
338 |
end_index = start_index + page_size
|
339 |
top_results = filtered_semantic_no_dupe[start_index:end_index]
|
340 |
-
|
341 |
-
|
342 |
-
|
343 |
rag_answer = get_rag_answer(var, top_results, DEDICATED_ENDPOINT, WRITE_ACCESS_TOKEN)
|
344 |
st.markdown(
|
345 |
f"<div style='background-color: #f0f0f0; color: #333; padding: 10px; border-radius: 5px; font-size:1.2em; text-align:center;'>{rag_answer}</div>",
|
@@ -351,7 +356,6 @@ else:
|
|
351 |
metadata = res.payload.get('metadata', {})
|
352 |
if "title" not in metadata:
|
353 |
metadata["title"] = compute_title(metadata)
|
354 |
-
# MOD: Remove hyperlink from project title
|
355 |
title_clean = re.sub(r'<a.*?>|</a>', '', metadata["title"])
|
356 |
st.markdown(f"#### {title_clean}")
|
357 |
|
@@ -398,7 +402,10 @@ else:
|
|
398 |
contact = metadata.get("contact", "").strip()
|
399 |
if contact and contact.lower() != "transparenz@giz.de":
|
400 |
additional_text += f"<br>**Contact:** xxx@giz.de"
|
401 |
-
|
402 |
st.markdown(additional_text, unsafe_allow_html=True)
|
403 |
|
404 |
st.divider()
|
|
|
|
|
|
|
|
|
|
39 |
###########################################
|
40 |
# Model Config
|
41 |
###########################################
|
|
|
42 |
config = configparser.ConfigParser()
|
43 |
config.read('model_params.cfg')
|
44 |
|
|
|
122 |
)
|
123 |
unique_country_names = sorted(country_name_mapping.keys())
|
124 |
|
125 |
+
###########################################
|
126 |
+
# Define reset_filters function using session_state
|
127 |
+
###########################################
|
128 |
+
def reset_filters():
|
129 |
+
st.session_state["region_filter"] = "All/Not allocated"
|
130 |
+
st.session_state["country_filter"] = "All/Not allocated"
|
131 |
+
current_year = datetime.now().year
|
132 |
+
default_start_year = current_year - 4
|
133 |
+
st.session_state["end_year_range"] = (default_start_year, max_end_year)
|
134 |
+
st.session_state["crs_filter"] = "All/Not allocated"
|
135 |
+
st.session_state["min_budget"] = min_budget_val
|
136 |
+
st.session_state["client_filter"] = "All/Not allocated"
|
137 |
+
# Optionally reset page number
|
138 |
+
st.session_state["page"] = 1
|
139 |
+
|
140 |
###########################################
|
141 |
# Filter Controls - Row 1
|
142 |
###########################################
|
143 |
col1, col2, col3, col4, col5 = st.columns([1, 1, 1, 1, 1])
|
144 |
|
145 |
with col1:
|
146 |
+
region_filter = st.selectbox("Region", ["All/Not allocated"] + sorted(unique_sub_regions), key="region_filter")
|
147 |
|
148 |
if region_filter == "All/Not allocated":
|
149 |
filtered_country_names = unique_country_names
|
|
|
154 |
]
|
155 |
|
156 |
with col2:
|
157 |
+
country_filter = st.selectbox("Country", ["All/Not allocated"] + filtered_country_names, key="country_filter")
|
158 |
|
159 |
with col3:
|
160 |
current_year = datetime.now().year
|
|
|
163 |
"Project End Year",
|
164 |
min_value=2010,
|
165 |
max_value=max_end_year,
|
166 |
+
value=(default_start_year, max_end_year),
|
167 |
+
key="end_year_range"
|
168 |
)
|
169 |
|
170 |
with col4:
|
171 |
crs_options = ["All/Not allocated"] + get_crs_options(client, collection_name)
|
172 |
+
crs_filter = st.selectbox("CRS", crs_options, key="crs_filter")
|
173 |
|
174 |
with col5:
|
175 |
min_budget = st.slider(
|
176 |
"Minimum Project Budget (Million €)",
|
177 |
min_value=min_budget_val,
|
178 |
max_value=max_budget_val,
|
179 |
+
value=min_budget_val,
|
180 |
+
key="min_budget"
|
181 |
)
|
182 |
|
183 |
|
|
|
184 |
###########################################
|
185 |
# Filter Controls - Row 2 (Additional Filters)
|
186 |
###########################################
|
187 |
col1_2, col2_2, col3_2, col4_2, col5_2 = st.columns(5)
|
188 |
|
189 |
with col1_2:
|
|
|
190 |
client_options = sorted(project_data["client"].dropna().unique().tolist())
|
191 |
+
client_filter = st.selectbox("Client", ["All/Not allocated"] + client_options, key="client_filter")
|
|
|
192 |
with col2_2:
|
193 |
st.empty()
|
194 |
with col3_2:
|
|
|
196 |
with col4_2:
|
197 |
st.empty()
|
198 |
with col5_2:
|
199 |
+
st.button("Reset Filters", on_click=reset_filters)
|
200 |
+
|
201 |
|
202 |
show_exact_matches = st.checkbox("Show only exact matches", value=False)
|
203 |
+
|
204 |
###########################################
|
205 |
# Main Search / Results
|
206 |
###########################################
|
|
|
244 |
get_country_name
|
245 |
)
|
246 |
|
247 |
+
# Additional filter by client
|
248 |
if client_filter != "All/Not allocated":
|
249 |
filtered_semantic = [r for r in filtered_semantic if r.payload.get("metadata", {}).get("client", "Unknown Client") == client_filter]
|
250 |
filtered_lexical = [r for r in filtered_lexical if r.payload.get("metadata", {}).get("client", "Unknown Client") == client_filter]
|
|
|
254 |
filtered_lexical_no_dupe = remove_duplicates(filtered_lexical)
|
255 |
|
256 |
def format_currency(value):
|
|
|
|
|
|
|
257 |
try:
|
258 |
return f"€{int(float(value)):,}"
|
259 |
except (ValueError, TypeError):
|
|
|
261 |
|
262 |
# 3) Display results
|
263 |
if show_exact_matches:
|
264 |
+
st.write("Showing **Top Lexical Search results**")
|
|
|
265 |
query_substring = var.strip().lower()
|
266 |
lexical_substring_filtered = [
|
267 |
r for r in filtered_lexical
|
|
|
271 |
if not filtered_lexical_no_dupe:
|
272 |
st.write('No exact matches, consider unchecking "Show only exact matches"')
|
273 |
else:
|
274 |
+
top_results = filtered_lexical_no_dupe # Show all matching lexical results
|
|
|
275 |
for res in top_results:
|
276 |
metadata = res.payload.get('metadata', {})
|
277 |
if "title" not in metadata:
|
278 |
metadata["title"] = compute_title(metadata)
|
|
|
279 |
title_html = highlight_query(metadata["title"], var) if var.strip() else metadata["title"]
|
280 |
title_clean = re.sub(r'<a.*?>|</a>', '', title_html)
|
281 |
st.markdown(f"#### {title_clean}", unsafe_allow_html=True)
|
282 |
|
|
|
283 |
objective = metadata.get("objective", "None")
|
284 |
desc_en = metadata.get("description.en", "").strip()
|
285 |
desc_de = metadata.get("description.de", "").strip()
|
|
|
323 |
contact = metadata.get("contact", "").strip()
|
324 |
if contact and contact.lower() != "transparenz@giz.de":
|
325 |
additional_text += f"<br>**Contact:** xxx@giz.de"
|
|
|
326 |
st.markdown(additional_text, unsafe_allow_html=True)
|
327 |
|
328 |
st.divider()
|
|
|
329 |
else:
|
|
|
330 |
if not filtered_semantic_no_dupe:
|
331 |
st.write("No relevant results found.")
|
332 |
else:
|
|
|
333 |
page_size = 15
|
334 |
total_results = len(filtered_semantic_no_dupe)
|
335 |
total_pages = (total_results - 1) // page_size + 1
|
336 |
+
|
337 |
+
# Use session_state for page selection; default to 1 if not set.
|
338 |
+
if "page" not in st.session_state:
|
339 |
+
st.session_state.page = 1
|
340 |
+
current_page = st.session_state.page
|
341 |
+
|
342 |
+
start_index = (current_page - 1) * page_size
|
343 |
end_index = start_index + page_size
|
344 |
top_results = filtered_semantic_no_dupe[start_index:end_index]
|
345 |
+
|
346 |
+
st.write(f"Showing **{len(top_results)}** Semantic Search results (Page {current_page} of {total_pages})")
|
347 |
+
|
348 |
rag_answer = get_rag_answer(var, top_results, DEDICATED_ENDPOINT, WRITE_ACCESS_TOKEN)
|
349 |
st.markdown(
|
350 |
f"<div style='background-color: #f0f0f0; color: #333; padding: 10px; border-radius: 5px; font-size:1.2em; text-align:center;'>{rag_answer}</div>",
|
|
|
356 |
metadata = res.payload.get('metadata', {})
|
357 |
if "title" not in metadata:
|
358 |
metadata["title"] = compute_title(metadata)
|
|
|
359 |
title_clean = re.sub(r'<a.*?>|</a>', '', metadata["title"])
|
360 |
st.markdown(f"#### {title_clean}")
|
361 |
|
|
|
402 |
contact = metadata.get("contact", "").strip()
|
403 |
if contact and contact.lower() != "transparenz@giz.de":
|
404 |
additional_text += f"<br>**Contact:** xxx@giz.de"
|
|
|
405 |
st.markdown(additional_text, unsafe_allow_html=True)
|
406 |
|
407 |
st.divider()
|
408 |
+
|
409 |
+
# Pagination widget moved to the very end of the page
|
410 |
+
new_page = st.selectbox("Select Page", list(range(1, total_pages + 1)), index=current_page - 1, key="page")
|
411 |
+
# The selected page value automatically updates st.session_state["page"]
|