Deploy CodeMode via Agent

app.py

@@ -1,4 +1,3 @@
-# CodeMode Comparison Demo - Last updated: 2026-02-11 23:59
 import gradio as gr
 import torch
 import torch.nn.functional as F
@@ -115,6 +114,125 @@ def list_finetuned_files():
     except Exception as e:
         return [[f"Error: {str(e)}", "-", "-"]]
 
+# --- Chunk Inspector Functions ---
+def get_files_list_baseline():
+    """Get list of unique files in baseline collection"""
+    try:
+        data = baseline_collection.get(include=["metadatas"])
+        if not data['metadatas']:
+            return []
+        files = list(set([m.get("file_name", "unknown") for m in data['metadatas']]))
+        return sorted(files)
+    except Exception as e:
+        return []
+
+def get_files_list_finetuned():
+    """Get list of unique files in fine-tuned collection"""
+    try:
+        data = finetuned_collection.get(include=["metadatas"])
+        if not data['metadatas']:
+            return []
+        files = list(set([m.get("file_name", "unknown") for m in data['metadatas']]))
+        return sorted(files)
+    except Exception as e:
+        return []
+
+def get_chunks_for_file_baseline(file_name):
+    """Get all chunks for a specific file from baseline collection"""
+    if not file_name:
+        return {"error": "No file selected"}
+    
+    try:
+        data = baseline_collection.get(
+            where={"file_name": file_name},
+            include=["documents", "metadatas", "embeddings"]
+        )
+        
+        if not data['documents']:
+            return {"error": "No chunks found for this file"}
+        
+        chunks = []
+        for i, (doc, meta) in enumerate(zip(data['documents'], data['metadatas'])):
+            chunks.append({
+                "chunk_id": i + 1,
+                "content": doc[:500] + "..." if len(doc) > 500 else doc,
+                "full_length": len(doc),
+                "metadata": meta,
+                "embedding_dim": len(data['embeddings'][i]) if data['embeddings'] else 0
+            })
+        
+        return {
+            "file_name": file_name,
+            "total_chunks": len(chunks),
+            "chunks": chunks
+        }
+    except Exception as e:
+        return {"error": str(e)}
+
+def get_chunks_for_file_finetuned(file_name):
+    """Get all chunks for a specific file from fine-tuned collection"""
+    if not file_name:
+        return {"error": "No file selected"}
+    
+    try:
+        data = finetuned_collection.get(
+            where={"file_name": file_name},
+            include=["documents", "metadatas", "embeddings"]
+        )
+        
+        if not data['documents']:
+            return {"error": "No chunks found for this file"}
+        
+        chunks = []
+        for i, (doc, meta) in enumerate(zip(data['documents'], data['metadatas'])):
+            chunks.append({
+                "chunk_id": i + 1,
+                "content": doc[:500] + "..." if len(doc) > 500 else doc,
+                "full_length": len(doc),
+                "metadata": meta,
+                "embedding_dim": len(data['embeddings'][i]) if data['embeddings'] else 0
+            })
+        
+        return {
+            "file_name": file_name,
+            "total_chunks": len(chunks),
+            "chunks": chunks
+        }
+    except Exception as e:
+        return {"error": str(e)}
+
+def download_chunks_baseline(file_name):
+    """Export chunks to JSON file for baseline"""
+    if not file_name:
+        return None
+    
+    import json
+    import tempfile
+    
+    chunks_data = get_chunks_for_file_baseline(file_name)
+    
+    temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.json')
+    json.dump(chunks_data, temp_file, indent=2)
+    temp_file.close()
+    
+    return temp_file.name
+
+def download_chunks_finetuned(file_name):
+    """Export chunks to JSON file for fine-tuned"""
+    if not file_name:
+        return None
+    
+    import json
+    import tempfile
+    
+    chunks_data = get_chunks_for_file_finetuned(file_name)
+    
+    temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.json')
+    json.dump(chunks_data, temp_file, indent=2)
+    temp_file.close()
+    
+    return temp_file.name
+
 # --- Search Functions ---
 def search_baseline(query, top_k=5):
     if baseline_collection.count() == 0: return []
@@ -600,6 +718,66 @@ with gr.Blocks(theme=theme, css=css, title="CodeMode - Baseline vs Fine-tuned")
                         value=[["No data yet", "-", "-"]]
                     )
                     inspect_finetuned_btn.click(list_finetuned_files, inputs=[], outputs=finetuned_files_df)
+            
+            gr.Markdown("---")
+            gr.Markdown("### Chunk Inspector")
+            gr.Markdown("View detailed chunk information for indexed files (content, metadata, schema)")
+            
+            with gr.Row():
+                with gr.Column():
+                    gr.Markdown("#### Baseline Collection")
+                    baseline_file_dropdown = gr.Dropdown(
+                        label="Select File to Inspect",
+                        choices=[],
+                        interactive=True
+                    )
+                    baseline_refresh_files = gr.Button("Refresh File List", variant="secondary")
+                    baseline_chunks_display = gr.JSON(label="Chunk Details")
+                    baseline_download_btn = gr.Button("Download Chunks as JSON", variant="primary")
+                    baseline_download_output = gr.File(label="Download")
+                
+                with gr.Column():
+                    gr.Markdown("#### Fine-tuned Collection")
+                    finetuned_file_dropdown = gr.Dropdown(
+                        label="Select File to Inspect",
+                        choices=[],
+                        interactive=True
+                    )
+                    finetuned_refresh_files = gr.Button("Refresh File List", variant="secondary")
+                    finetuned_chunks_display = gr.JSON(label="Chunk Details")
+                    finetuned_download_btn = gr.Button("Download Chunks as JSON", variant="primary")
+                    finetuned_download_output = gr.File(label="Download")
+            
+            # Wire up Chunk Inspector events
+            baseline_refresh_files.click(
+                lambda: gr.Dropdown(choices=get_files_list_baseline()),
+                outputs=baseline_file_dropdown
+            )
+            baseline_file_dropdown.change(
+                get_chunks_for_file_baseline,
+                inputs=baseline_file_dropdown,
+                outputs=baseline_chunks_display
+            )
+            baseline_download_btn.click(
+                download_chunks_baseline,
+                inputs=baseline_file_dropdown,
+                outputs=baseline_download_output
+            )
+            
+            finetuned_refresh_files.click(
+                lambda: gr.Dropdown(choices=get_files_list_finetuned()),
+                outputs=finetuned_file_dropdown
+            )
+            finetuned_file_dropdown.change(
+                get_chunks_for_file_finetuned,
+                inputs=finetuned_file_dropdown,
+                outputs=finetuned_chunks_display
+            )
+            finetuned_download_btn.click(
+                download_chunks_finetuned,
+                inputs=finetuned_file_dropdown,
+                outputs=finetuned_download_output
+            )
         
         # TAB 2: COMPARISON SEARCH
         with gr.Tab("2. Comparison Search (Note: Semantic search is sensitive to query phrasing)"):