Update app.py

app.py

@@ -9,13 +9,13 @@ from sklearn.preprocessing import StandardScaler
 from sklearn.impute import SimpleImputer
 from kneed import KneeLocator
 from openai import OpenAI
-import os # Make sure os is imported
+import os
 import warnings
 import base64
 import traceback
 import time
 
-# Add this in your code to check if the secret is loaded
+# Function to check the status of the API key from environment variables
 def check_api_key_status():
     api_key = os.environ.get("NEBIUS_API_KEY")
     if api_key:
@@ -23,9 +23,9 @@ def check_api_key_status():
     else:
         return "❌ API key not found in environment variables"
 
-# You can call this function to test
 print(check_api_key_status())
 
+# Function to encode the logo image to base64 for embedding in HTML
 def get_logo_base64():
     try:
         with open("logo.png", "rb") as f:
@@ -35,13 +35,12 @@ def get_logo_base64():
         print("Logo file not found")
         return None
 
-
-
+# Ignore specific warnings for a cleaner output
 warnings.filterwarnings("ignore", category=UserWarning, module='kneed')
 warnings.filterwarnings("ignore", category=FutureWarning)
 
 # ============================================================================
-# DATA PROCESSING FUNCTIONS (Unchanged)
+# DATA PROCESSING FUNCTIONS
 # ===========================================================================
 def read_csv_headers(file):
     if not file: return []
@@ -124,7 +123,7 @@ def perform_clustering_auto_k(df, features, max_k=10):
     return df_with_clusters, final_n_clusters, scree_fig
 
 # =============================================================================
-# PLOTTING & OTHER FUNCTIONS (Unchanged)
+# PLOTTING & OTHER FUNCTIONS
 # =============================================================================
 def get_basic_stats(df, primary_element):
     if df.empty or primary_element not in df.columns: return pd.DataFrame()
@@ -148,7 +147,8 @@ def create_multi_offset_3d_plot(df, primary_element):
     for hole_id in df['HOLE_ID'].unique():
         hole_data = df[df['HOLE_ID'] == hole_id].sort_values('FROM')
         for _, row in hole_data.iterrows():
-            az_rad, dip_rad = np.radians(90 - row['AZIMUTH']), np.radians(-row['DIP'])
+            # Calculate radians for azimuth and dip, using dip directly
+            az_rad, dip_rad = np.radians(90 - row['AZIMUTH']), np.radians(row['DIP'])
             start_x, start_y, start_z = row['EASTING'] + (row['FROM'] * np.cos(dip_rad) * np.cos(az_rad)), row['NORTHING'] + (row['FROM'] * np.cos(dip_rad) * np.sin(az_rad)), row['ELEVATION'] + (row['FROM'] * np.sin(dip_rad))
             end_x, end_y, end_z = row['EASTING'] + (row['TO'] * np.cos(dip_rad) * np.cos(az_rad)), row['NORTHING'] + (row['TO'] * np.cos(dip_rad) * np.sin(az_rad)), row['ELEVATION'] + (row['TO'] * np.sin(dip_rad))
             fig.add_trace(go.Scatter3d(x=[start_x + offsets['lithology'], end_x + offsets['lithology']], y=[start_y, end_y], z=[start_z, end_z], mode='lines', line=dict(width=20, color=litho_color_map.get(row['LITHO'], 'grey')), name=row['LITHO'], legendgroup=row['LITHO'], showlegend=row['LITHO'] not in [t.name for t in fig.data if hasattr(t, 'legendgroup') and t.legendgroup == row['LITHO']]))
@@ -238,7 +238,7 @@ def create_cross_section_plot_gr(df, collar_df, section_holes, width, primary_el
 
 def generate_summary_prompt(df, primary_element, element_cols, optimal_k, litho_dict, user_context=""):
     """
-    Generates a detailed, context-rich prompt for the LLM, now with a focus on significant intercepts
+    Generates a detailed, context-rich prompt for the LLM, focusing on significant intercepts
     and dynamic cluster signature analysis.
     """
     if df.empty:
@@ -251,7 +251,7 @@ def generate_summary_prompt(df, primary_element, element_cols, optimal_k, litho_
     num_samples = len(df)
     prompt += f"- Drillholes analysed: {num_holes}, Total samples: {num_samples}.\n"
 
-    # --- Primary Element Summary ---
+    # Primary Element Summary
     if primary_element and primary_element in df.columns:
         mean_val = df[primary_element].mean()
         median_val = df[primary_element].median()
@@ -260,7 +260,7 @@ def generate_summary_prompt(df, primary_element, element_cols, optimal_k, litho_
         prompt += f"- Primary Element of Interest: {primary_element}.\n"
         prompt += f"  - Overall Stats: Mean = {mean_val:.2f}, Median = {median_val:.2f}, Std Dev = {std_val:.2f}, Max = {max_val:.2f}.\n"
 
-    # --- Significant Intercepts Analysis ---
+    # Significant Intercepts Analysis
     if primary_element and primary_element in df.columns:
         prompt += "\n- Top 5 Significant Intercepts (by grade):\n"
         top_intercepts = df.nlargest(5, primary_element)
@@ -279,7 +279,7 @@ def generate_summary_prompt(df, primary_element, element_cols, optimal_k, litho_
                     intercept_info += f"    - Geochemical Cluster: {int(row['Cluster'])}\n"
                 prompt += intercept_info
 
-    # --- Spatial Trends (Swath Plot Summary) ---
+    # Spatial Trends (Swath Plot Summary)
     try:
         east_stats = create_swath_data(df, 'x', primary_element, num_bins=5)
         if not east_stats.empty:
@@ -289,7 +289,7 @@ def generate_summary_prompt(df, primary_element, element_cols, optimal_k, litho_
     except Exception:
         prompt += "- (Swath plot statistics calculation failed)\n"
 
-    # --- Lithology Analysis ---
+    # Lithology Analysis
     if 'LITHO' in df.columns and df['LITHO'].nunique() > 1:
         prompt += "\n- Lithology Analysis:\n"
         litho_counts = df['LITHO'].value_counts()
@@ -311,7 +311,7 @@ def generate_summary_prompt(df, primary_element, element_cols, optimal_k, litho_
                     prompt += f"    * Highest Median Grades are associated with: {', '.join(hg_list)}\n"
             except Exception: pass
 
-    # --- Geochemical Cluster Analysis ---
+    # Geochemical Cluster Analysis
     if 'Cluster' in df.columns and df['Cluster'].nunique() > 1:
         prompt += f"\n- Geochemical Cluster Analysis ({optimal_k} Clusters Identified):\n"
         # Calculate global medians for enrichment factor calculation
@@ -347,7 +347,6 @@ def generate_summary_prompt(df, primary_element, element_cols, optimal_k, litho_
                     med_list.append(f"{el} median={median_val:.3f} ({enrichment_val:.1f}x vs global)")
             
             prompt += f"    - Key Geochemical Signature: {'; '.join(med_list)}\n"
-            # --- MODIFICATION END ---
 
             if 'LITHO' in cluster_df.columns and cluster_df['LITHO'].nunique() > 0:
                 top_litho_code = cluster_df['LITHO'].mode()[0]
@@ -355,7 +354,7 @@ def generate_summary_prompt(df, primary_element, element_cols, optimal_k, litho_
                 litho_desc = f" ({litho_dict.get(str(top_litho_code), 'No Description')})" if litho_dict else ""
                 prompt += f"    - Dominant Lithology: {top_litho_code}{litho_desc} ({percentage:.1f}% of cluster)\n"
 
-    # --- Final Instructions ---
+    # Final Instructions for the LLM
     prompt += f"\n--- USER CONTEXT ---\n{user_context if user_context else 'No additional context provided.'}\n"
     prompt += "\n--- INSTRUCTIONS FOR LLM ---\n"
     prompt += "Based on the detailed data summary provided above, please provide a concise yet detailed geological interpretation. Focus on:\n"
@@ -369,7 +368,7 @@ def generate_summary_prompt(df, primary_element, element_cols, optimal_k, litho_
     
     return prompt
 
-def get_nebius_llm_response(prompt, history, model="Qwen/Qwen3-30B-A3B"):
+def get_nebius_llm_response(prompt, history, model="deepseek-ai/DeepSeek-V3-0324-fast"):
     """
     Gets a response from the Nebius LLM API using an environment variable for the key.
     """
@@ -408,7 +407,6 @@ def get_nebius_llm_response(prompt, history, model="Qwen/Qwen3-30B-A3B"):
         print(f"Nebius API Error: {e}")  # For server logs
         return error_msg
 
-        
 def update_plan_view_with_selection(state, selected_holes):
     if not state: return go.Figure().update_layout(title_text="Please run analysis first")
     collar_df = pd.read_json(state["collar_df"])
@@ -432,7 +430,6 @@ def check_runnable_state(*args):
 
 def chat_response(message, history, state):
     if not state: return "", history + [(message, "Please run an analysis first.")]
-    # The API key is now handled inside get_nebius_llm_response
     response = get_nebius_llm_response(f"Follow-up question: {message}", history)
     history.append((message, response)); return "", history
 
@@ -490,7 +487,6 @@ def run_analysis_pipeline(collar_file, assay_file, litho_file, litho_dict_file,
                           lm_hole, lm_from, lm_to, lm_lith, 
                           ldm_code, ldm_desc, primary_element, user_context):
     try:
-        # Adjusted total steps to better reflect the process
         total_steps = 7
         initial_outputs = (
             gr.Tabs(), gr.Column(visible=False), gr.Column(visible=True), pd.DataFrame(), go.Figure(), 
@@ -510,7 +506,7 @@ def run_analysis_pipeline(collar_file, assay_file, litho_file, litho_dict_file,
         merged_df = pd.merge(assay_df, collar_df, on='HOLE_ID', how='inner')
         if merged_df.empty: raise gr.Error("Data processing resulted in an empty dataset after merging.")
         
-        litho_dict = {} # Define litho_dict with a default value
+        litho_dict = {}
         if litho_file:
             litho_map = {'HOLE_ID': lm_hole, 'FROM': lm_from, 'TO': lm_to, 'LITHO': lm_lith}
             litho_df = pd.read_csv(litho_file.name); litho_df = litho_df.rename(columns={v: k for k, v in litho_map.items() if v is not None})
@@ -525,14 +521,14 @@ def run_analysis_pipeline(collar_file, assay_file, litho_file, litho_dict_file,
         # Step 3/7: Calculating Coordinates
         yield (gr.Row(visible=True), create_progress_html(3, total_steps, "Calculating 3D Coordinates..."), *initial_outputs)
         merged_df['MIDPOINT'] = (merged_df['FROM'] + merged_df['TO']) / 2
-        azimuth_rad, dip_rad = np.radians(90 - merged_df['AZIMUTH']), np.radians(-merged_df['DIP'])
+        # Convert angles to radians, using dip directly as provided (e.g., -90 for vertical down)
+        azimuth_rad, dip_rad = np.radians(90 - merged_df['AZIMUTH']), np.radians(merged_df['DIP'])
         merged_df['x'] = merged_df['EASTING'] + (merged_df['MIDPOINT'] * np.cos(dip_rad) * np.cos(azimuth_rad))
         merged_df['y'] = merged_df['NORTHING'] + (merged_df['MIDPOINT'] * np.cos(dip_rad) * np.sin(azimuth_rad))
         merged_df['z'] = merged_df['ELEVATION'] + (merged_df['MIDPOINT'] * np.sin(dip_rad))
         merged_df.dropna(subset=['x', 'y', 'z', primary_element], inplace=True)
         if merged_df.empty: raise gr.Error("Data processing resulted in an empty dataset after coordinate calculation.")
 
-        
         # Step 4/7: Clustering
         yield (gr.Row(visible=True), create_progress_html(4, total_steps, "Running K-Means Clustering..."), *initial_outputs)
         cluster_features = element_cols
@@ -550,7 +546,6 @@ def run_analysis_pipeline(collar_file, assay_file, litho_file, litho_dict_file,
         # Step 6/7: AI Analysis
         yield (gr.Row(visible=True), create_progress_html(6, total_steps, "Generating AI Summary... (This may take a moment)"), *initial_outputs)
         
-        # Call the prompt function 
         summary_prompt = generate_summary_prompt(
             merged_df, 
             primary_element, 
@@ -565,7 +560,7 @@ def run_analysis_pipeline(collar_file, assay_file, litho_file, litho_dict_file,
         
         # Step 7/7: Finalizing
         yield (gr.Row(visible=True), create_progress_html(7, total_steps, "Finalising and Loading Dashboard..."), *initial_outputs)
-        time.sleep(0.5) # A brief pause to ensure the user sees the 100% state
+        time.sleep(0.5)
 
         # Final yield to show the results
         yield (
@@ -596,19 +591,15 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Agentic Geo", css="""
         <img src="{logo_b64 if logo_b64 else ''}" alt="GeoInsights Agent Logo" style="max-height: 600px; width: auto; margin-bottom: 15px; display: {'block' if logo_b64 else 'none'};">
         <h1 style="color: #2c3e50; font-weight: 600; margin: 0 0 10px 0; font-size: 2.0em;">{subtitle_text}</h1>
         
-        <!-- START: ADDED VIDEO LINK -->
         <a href="https://drive.google.com/file/d/15A7LNl2ON4YimiAMcc_ihxkxTGfNIFR1/view?usp=sharing" target="_blank" style="font-size: 1.2em; color: #007BFF; text-decoration: none; margin-top: 15px; margin-bottom: 20px; display: inline-block;">
             📹 View Video Demonstration
         </a>
-        <!-- END: ADDED VIDEO LINK -->
 
         <h2 style="color: #333; font-weight: 500; margin: 10px 0; border-top: 1px solid #ddd; padding-top: 20px; width: 80%;">Step 1: Upload Data & Map Columns → Step 2: Run Analysis</h2>
     </div>
     """
-    # If the logo wasn't found, we add a fallback emoji icon instead of the image tag.
     if not logo_b64:
         fallback_icon = '<div style="font-size: 60px; margin-bottom: 15px;">🪨</div>'
-        # Insert the icon right before the main title
         header_html = header_html.replace('<h1', f'{fallback_icon}<h1')
 
     gr.HTML(header_html)