Update app/ui.py

app/ui.py

@@ -83,17 +83,37 @@ with st.sidebar.expander("Select Dataset & Topic Model", expanded=True):
     selected_dataset = st.selectbox("Dataset", datasets, help="Choose an available dataset.")
     selected_model = st.selectbox("Model", AVAILABLE_MODELS, help="Select topic model architecture.")
 
+# Check if the dataset has changed and reset session state if it has.
+if 'current_dataset' not in st.session_state or st.session_state.current_dataset != selected_dataset:
+    st.session_state.current_dataset = selected_dataset
+    # List all session state keys that depend on the dataset
+    keys_to_clear = [
+        "selected_words",
+        "interesting_words",
+        "word_counts_multiselect",
+        "collected_deduplicated_docs",
+        "summary",
+        "context_for_followup",
+        "followup_history"
+    ]
+    for key in keys_to_clear:
+        if key in st.session_state:
+            del st.session_state[key]
+    # Rerun the script to apply the clean state
+    st.rerun()
+
 # Resolve paths
 dataset_path = os.path.join(DATA_DIR, selected_dataset)
 model_path = os.path.join(dataset_path, selected_model)
-docs_path = os.path.join(dataset_path, "docs.jsonl")
 vocab_path = os.path.join(dataset_path, "processed/vocab.txt")
 time2id_path = os.path.join(dataset_path, "processed/time2id.txt")
+length_stats_path = os.path.join(dataset_path, "processed/length_stats.json")
+lemma_map_path = os.path.join(dataset_path, "processed/lemma_to_forms.json")
+docs_path = os.path.join(dataset_path, "docs.jsonl")
 index_path = os.path.join(dataset_path, "inverted_index.json")
 beta_path = os.path.join(model_path, "beta.npy")
 label_cache_path = os.path.join(model_path, "topic_label_cache.json")
-length_stats_path = os.path.join(dataset_path, "processed/length_stats.json")
-lemma_map_path = os.path.join(dataset_path, "processed/lemma_to_forms.json")
+
 
 with st.sidebar.expander("LLM Settings", expanded=True):
     provider = st.selectbox("LLM Provider", options=list(ENV_VAR_MAP.keys()), help="Choose the LLM backend.")
@@ -256,15 +276,26 @@ if selected_words:
     for i, word in enumerate(selected_words):
         trend = get_word_trend(beta, vocab, word, topic_id=selected_topic)
         color = color_cycle[i % len(color_cycle)]
+        # --- START: Modify this line ---
         fig.add_trace(go.Scatter(
             x=time_labels,
             y=trend,
             name=word,
+            mode='lines+markers',  # Explicitly add markers to the lines
             line=dict(color=color),
             legendgroup=word,
             showlegend=True
         ))
-    fig.update_layout(title="", xaxis_title="Year", yaxis_title="Importance", legend=dict(font=dict(size=16)))
+    fig.update_layout(
+        title="", 
+        xaxis_title="Year", 
+        yaxis_title="Importance",
+        legend=dict(
+            font=dict(
+                size=16
+            )
+        )
+    )
     _, chart_col, _ = st.columns([0.2, 0.6, 0.2])
     with chart_col:
         st.plotly_chart(fig, use_container_width=True)