Update app.py

app.py

@@ -93,8 +93,8 @@ with col_about:
             """, unsafe_allow_html=True
         )
 
-# Main query input
-var = st.text_input("Enter Question")
+# Main query input (with a key so we can reset it)
+var = st.text_input("Enter Question", key="query")
 
 ###########################################
 # Create or load the embeddings collection
@@ -134,7 +134,8 @@ def reset_filters():
     st.session_state["crs_filter"] = "All/Not allocated"
     st.session_state["min_budget"] = min_budget_val
     st.session_state["client_filter"] = "All/Not allocated"
-    # Optionally reset page number
+    st.session_state["query"] = ""
+    st.session_state["show_exact_matches"] = False
     st.session_state["page"] = 1
 
 ###########################################
@@ -144,7 +145,6 @@ col1, col2, col3, col4, col5 = st.columns([1, 1, 1, 1, 1])
 
 with col1:
     region_filter = st.selectbox("Region", ["All/Not allocated"] + sorted(unique_sub_regions), key="region_filter")
-
 if region_filter == "All/Not allocated":
     filtered_country_names = unique_country_names
 else:
@@ -180,7 +180,6 @@ with col5:
         key="min_budget"
     )
 
-
 ###########################################
 # Filter Controls - Row 2 (Additional Filters)
 ###########################################
@@ -196,10 +195,29 @@ with col3_2:
 with col4_2:
     st.empty()
 with col5_2:
-    st.button("Reset Filters", on_click=reset_filters)
+    # Plain reset button (will be moved to row 3 as well)
+    st.button("Reset Filters", on_click=reset_filters, key="reset_button_row2")
 
-
-show_exact_matches = st.checkbox("Show only exact matches", value=False)
+###########################################
+# Filter Controls - Row 3 (Remaining Filter)
+###########################################
+col1_3, col2_3, col3_3, col4_3, col5_3 = st.columns(5)
+with col1_3:
+    # Place the "Show only exact matches" checkbox here
+    show_exact_matches = st.checkbox("Show only exact matches", key="show_exact_matches")
+with col2_3:
+    st.empty()
+with col3_3:
+    st.empty()
+with col4_3:
+    st.empty()
+with col5_3:
+    # Right-align a more prominent reset button
+    with st.container():
+        st.markdown("<div style='text-align: right;'>", unsafe_allow_html=True)
+        if st.button("**Reset Filters**", key="reset_button_row3"):
+            reset_filters()
+        st.markdown("</div>", unsafe_allow_html=True)
 
 ###########################################
 # Main Search / Results
@@ -259,7 +277,11 @@ else:
         except (ValueError, TypeError):
             return value
 
+    # --- Reprint Query (Right Aligned with "Query:") ---
+    st.markdown(f"<div style='text-align: right; font-size:2.1em; font-style: italic; font-weight: bold;'>Query: {var}</div>", unsafe_allow_html=True)
+
     # 3) Display results
+    # Lexical Search Results Branch
     if show_exact_matches:
         st.write("Showing **Top Lexical Search results**")
         query_substring = var.strip().lower()
@@ -272,13 +294,31 @@ else:
             st.write('No exact matches, consider unchecking "Show only exact matches"')
         else:
             top_results = filtered_lexical_no_dupe  # Show all matching lexical results
-            for res in top_results:
+
+            # --- Pagination (Above Lexical Results) ---
+            page_size = 15
+            total_results = len(top_results)
+            total_pages = (total_results - 1) // page_size + 1
+            if "page" not in st.session_state:
+                st.session_state.page = 1
+            current_page = st.session_state.page
+            # Top pagination widget (right aligned, 1/7 width)
+            col_pag_top = st.columns([6, 1])[1]
+            new_page_top = col_pag_top.selectbox("Select Page", list(range(1, total_pages + 1)), index=current_page - 1, key="page_top")
+            st.session_state.page = new_page_top
+
+            start_index = (st.session_state.page - 1) * page_size
+            end_index = start_index + page_size
+            paged_results = top_results[start_index:end_index]
+
+            for i, res in enumerate(paged_results, start=start_index+1):
                 metadata = res.payload.get('metadata', {})
                 if "title" not in metadata:
                     metadata["title"] = compute_title(metadata)
                 title_html = highlight_query(metadata["title"], var) if var.strip() else metadata["title"]
                 title_clean = re.sub(r'<a.*?>|</a>', '', title_html)
-                st.markdown(f"#### {title_clean}", unsafe_allow_html=True)
+                # Prepend the result number
+                st.markdown(f"#### {i}. **{title_clean}**", unsafe_allow_html=True)
 
                 objective = metadata.get("objective", "None")
                 desc_en = metadata.get("description.en", "").strip()
@@ -297,7 +337,6 @@ else:
                     if remainder_text:
                         with st.expander("Show more"):
                             st.markdown(highlight_query(remainder_text, var), unsafe_allow_html=True)
-
                 with col_right:
                     start_year_str = extract_year(metadata.get('start_year', None)) or "Unknown"
                     end_year_str = extract_year(metadata.get('end_year', None)) or "Unknown"
@@ -311,21 +350,37 @@ else:
                     new_crs_value = lookup_crs_value(crs_key_clean)
                     new_crs_value_clean = re.sub(r'\.0$', '', str(new_crs_value))
                     crs_combined = f"{crs_key_clean}: {new_crs_value_clean}" if crs_key_clean else "Unknown"
-
+                    
+                    # Insert Predecessor/Successor line if available
+                    predecessor = metadata.get("predecessor_id", "").strip()
+                    successor = metadata.get("successor_id", "").strip()
+                    extra_line = ""
+                    if predecessor:
+                        extra_line += f"<br>**Predecessor Project:** {predecessor}"
+                    if successor:
+                        extra_line += f"<br>**Successor Project:** {successor}"
+                    
                     additional_text = (
                         f"**Objective:** {highlight_query(objective, var)}<br>"
                         f"**Commissioned by:** {metadata.get('client', 'Unknown Client')}<br>"
                         f"**Projekt duration:** {start_year_str}-{end_year_str}<br>"
-                        f"**Budget:** Project: {formatted_project_budget}, Total volume: {formatted_total_volume}<br>"
-                        f"**Country:** {country_raw}<br>"
+                        f"**Budget:** Project: <b>{formatted_project_budget}</b>, Total volume: <b>{formatted_total_volume}</b><br>"
+                        + extra_line +
+                        f"<br>**Country:** {country_raw}<br>"
                         f"**Sector:** {crs_combined}"
                     )
                     contact = metadata.get("contact", "").strip()
                     if contact and contact.lower() != "[email protected]":
                         additional_text += f"<br>**Contact:** [email protected]"
                     st.markdown(additional_text, unsafe_allow_html=True)
-
                 st.divider()
+
+            # Bottom pagination widget
+            col_pag_bot = st.columns([6, 1])[1]
+            new_page_bot = col_pag_bot.selectbox("Select Page", list(range(1, total_pages + 1)), index=st.session_state.page - 1, key="page_bot")
+            st.session_state.page = new_page_bot
+
+    # Semantic Search Results Branch
     else:
         if not filtered_semantic_no_dupe:
             st.write("No relevant results found.")
@@ -334,30 +389,43 @@ else:
             total_results = len(filtered_semantic_no_dupe)
             total_pages = (total_results - 1) // page_size + 1
 
-            # Use session_state for page selection; default to 1 if not set.
             if "page" not in st.session_state:
                 st.session_state.page = 1
             current_page = st.session_state.page
 
-            start_index = (current_page - 1) * page_size
+            # Top pagination widget (right aligned, 1/7 width)
+            col_pag_top = st.columns([6, 1])[1]
+            new_page_top = col_pag_top.selectbox("Select Page", list(range(1, total_pages + 1)), index=current_page - 1, key="page_top_sem")
+            st.session_state.page = new_page_top
+
+            start_index = (st.session_state.page - 1) * page_size
             end_index = start_index + page_size
             top_results = filtered_semantic_no_dupe[start_index:end_index]
 
-            st.write(f"Showing **{len(top_results)}** Semantic Search results (Page {current_page} of {total_pages})")
+            # Prominent page info with bold numbers and green highlight if current page is not 1
+            page_num = f"<b style='color: green;'>{st.session_state.page}</b>" if st.session_state.page != 1 else f"<b>{st.session_state.page}</b>"
+            total_pages_str = f"<b>{total_pages}</b>"
+            st.markdown(f"Showing **{len(top_results)}** Semantic Search results (Page {page_num} of {total_pages_str})", unsafe_allow_html=True)
 
+            # --- RAG Answer (Right aligned, bullet points, bold numbers) ---
             rag_answer = get_rag_answer(var, top_results, DEDICATED_ENDPOINT, WRITE_ACCESS_TOKEN)
-            st.markdown(
-                f"<div style='background-color: #f0f0f0; color: #333; padding: 10px; border-radius: 5px; font-size:1.2em; text-align:center;'>{rag_answer}</div>", 
-                unsafe_allow_html=True
-            )
+            bullet_lines = []
+            for line in rag_answer.splitlines():
+                if line.strip():
+                    # Bold any numbers in the line
+                    line_bold = re.sub(r'(\d+)', r'<b>\1</b>', line)
+                    bullet_lines.append(f"<li>{line_bold}</li>")
+            formatted_rag_answer = "<ul style='text-align: right; list-style-position: inside;'>" + "".join(bullet_lines) + "</ul>"
+            st.markdown(formatted_rag_answer, unsafe_allow_html=True)
             st.divider()
 
-            for res in top_results:
+            for i, res in enumerate(top_results, start=start_index+1):
                 metadata = res.payload.get('metadata', {})
                 if "title" not in metadata:
                     metadata["title"] = compute_title(metadata)
                 title_clean = re.sub(r'<a.*?>|</a>', '', metadata["title"])
-                st.markdown(f"#### {title_clean}")
+                # Prepend result number and make title bold
+                st.markdown(f"#### {i}. **{title_clean}**", unsafe_allow_html=True)
 
                 desc_en = metadata.get("description.en", "").strip()
                 desc_de = metadata.get("description.de", "").strip()
@@ -376,7 +444,6 @@ else:
                     if remainder_text:
                         with st.expander("Show more"):
                             st.markdown(highlight_query(remainder_text, var), unsafe_allow_html=True)
-
                 with col_right:
                     start_year_str = extract_year(metadata.get('start_year', None)) or "Unknown"
                     end_year_str = extract_year(metadata.get('end_year', None)) or "Unknown"
@@ -390,22 +457,31 @@ else:
                     new_crs_value = lookup_crs_value(crs_key_clean)
                     new_crs_value_clean = re.sub(r'\.0$', '', str(new_crs_value))
                     crs_combined = f"{crs_key_clean}: {new_crs_value_clean}" if crs_key_clean else "Unknown"
-
+                    
+                    predecessor = metadata.get("predecessor_id", "").strip()
+                    successor = metadata.get("successor_id", "").strip()
+                    extra_line = ""
+                    if predecessor:
+                        extra_line += f"<br>**Predecessor Project:** {predecessor}"
+                    if successor:
+                        extra_line += f"<br>**Successor Project:** {successor}"
+                    
                     additional_text = (
                         f"**Objective:** {metadata.get('objective', '')}<br>"
                         f"**Commissioned by:** {metadata.get('client', 'Unknown Client')}<br>"
                         f"**Projekt duration:** {start_year_str}-{end_year_str}<br>"
-                        f"**Budget:** Project: {formatted_project_budget}, Total volume: {formatted_total_volume}<br>"
-                        f"**Country:** {country_raw}<br>"
+                        f"**Budget:** Project: <b>{formatted_project_budget}</b>, Total volume: <b>{formatted_total_volume}</b><br>"
+                        + extra_line +
+                        f"<br>**Country:** {country_raw}<br>"
                         f"**Sector:** {crs_combined}"
                     )
                     contact = metadata.get("contact", "").strip()
                     if contact and contact.lower() != "[email protected]":
                         additional_text += f"<br>**Contact:** [email protected]"
                     st.markdown(additional_text, unsafe_allow_html=True)
-
                 st.divider()
 
-            # Pagination widget moved to the very end of the page
-            new_page = st.selectbox("Select Page", list(range(1, total_pages + 1)), index=current_page - 1, key="page")
-            # The selected page value automatically updates st.session_state["page"]
+            # Bottom pagination widget (right aligned, 1/7 width)
+            col_pag_bot = st.columns([6, 1])[1]
+            new_page_bot = col_pag_bot.selectbox("Select Page", list(range(1, total_pages + 1)), index=st.session_state.page - 1, key="page_bot_sem")
+            st.session_state.page = new_page_bot