Harden analysis result rendering

doc_intelligence/src/app/streamlit_app.py

@@ -485,6 +485,7 @@ with tab_extract:
                             "then restart the Space and try again."
                         )
                     pipeline = get_end_to_end_pipeline(model_choice, t1_thresh, t2_thresh, ocr_model_choice)
+                    print(f"[ANALYSIS] {doc_id}: request accepted", flush=True)
                     with st.spinner("Processing document layout, extracting tokens with OCR.space, and running classification..."):
                         computed_result = pipeline.process_document(doc_path, doc_id=doc_id)
                         st.session_state["analysis_results"][doc_id] = computed_result
@@ -558,7 +559,7 @@ with tab_extract:
                         {"Category": k.replace("_", " ").title(), "Probability": v}
                         for k, v in result.get("class_probabilities", {}).items()
                     ])
-                    st.bar_chart(probs_df.set_index("Category"), height=110)
+                    st.dataframe(probs_df, width="stretch", hide_index=True)
 
                 # Routing Decision Panel
                 if result["needs_review"]:
@@ -598,11 +599,10 @@ with tab_extract:
                     })
 
                 df_extracted = pd.DataFrame(fields_data)
-                edited_df = st.data_editor(
+                st.dataframe(
                     df_extracted[["Field Name", "Extracted Value", "Confidence"]],
-                        width="stretch",
-                    num_rows="dynamic",
-                    key="editor_fields"
+                    width="stretch",
+                    hide_index=True
                 )
 
                 # Export & Download Actions

doc_intelligence/src/pipeline/end_to_end.py

@@ -37,7 +37,9 @@ class DocumentIntelligencePipeline:
         t0 = time.time()
 
         # 1. OCR (with optional preprocessing)
+        print(f"[ANALYSIS] {doc_id}: OCR started", flush=True)
         ocr_words, _ = self.ocr_engine.extract_with_cache(image_or_pdf_path, doc_id=doc_id)
+        print(f"[ANALYSIS] {doc_id}: OCR completed with {len(ocr_words)} words", flush=True)
 
         # Build DocumentRecord for inference
         doc_rec = DocumentRecord(
@@ -55,6 +57,7 @@ class DocumentIntelligencePipeline:
         predicted_class = self.classifier.classes_[class_idx]
         class_confidence = float(class_probs[class_idx])
         class_probabilities = {c: round(float(p), 4) for c, p in zip(self.classifier.classes_, class_probs)}
+        print(f"[ANALYSIS] {doc_id}: classified as {predicted_class}", flush=True)
 
         # 3. Stage 2: Class-specific Field Extraction
         extractor = self.extractors.get(predicted_class)
@@ -67,6 +70,7 @@ class DocumentIntelligencePipeline:
                 raw_fields = extractor.extract(doc_rec)
         else:
             raw_fields = {}
+        print(f"[ANALYSIS] {doc_id}: field extraction completed", flush=True)
 
         # 4. Post-processing & Business Validation
         normalized_fields, validation_res = postprocess_extracted_fields(predicted_class, raw_fields)