Upload 3 files

Updated app.py files by adding OCR Recognition Toggle bar

DEPLOYMENT.md

@@ -4,7 +4,7 @@ This guide covers deploying DocuChat_AI as a Streamlit app on Hugging Face Space
 
 ## 1. Recommended Hugging Face Setup
 
-For the standard app:
+For the standard app without scanned PDF OCR:
 
 ```text
 SDK: Streamlit
@@ -48,7 +48,7 @@ Pillow
 Tesseract OCR binary
 ```
 
-Python packages are installed from `requirements.txt`, but the Tesseract system binary may not be available in a normal Streamlit Space.
+Python packages are installed from `requirements.txt`, but the Tesseract system binary is usually not available in a normal Streamlit SDK Space.
 
 For best OCR support, deploy using the included `Dockerfile`, which installs:
 
@@ -66,6 +66,12 @@ Hardware: CPU Upgrade if processing larger scanned PDFs
 
 If OCR dependencies are missing, the app fails gracefully and still works for normal text-based PDFs, DOCX, and TXT files.
 
+Important:
+
+```text
+If your scanned PDF returns "No readable text found", switch the Space SDK to Docker or create a new Docker Space using this repository.
+```
+
 ## 3. Evaluation Dashboard
 
 The Evaluation tab supports labeled RAG testing with CSV files.
@@ -117,4 +123,3 @@ The correctness and faithfulness scores are LLM-judged and should be treated as
 - Add `APP_PASSWORD` for portfolio demos shared with recruiters.
 - Use the Evaluation tab with a small labeled test set before demos.
 - Use CPU Upgrade for larger PDFs or frequent OCR use.
-

README.md

@@ -267,7 +267,7 @@ GROQ_API_KEY = your_key_here
 5. Optional: add `APP_PASSWORD` in **Settings -> Secrets** to protect the app.
 6. Restart the Space after dependency changes.
 
-For scanned PDF OCR, use the included Dockerfile because OCR needs system packages such as Tesseract.
+For scanned PDF OCR, use the included Dockerfile because OCR needs the Tesseract system binary. Normal Streamlit SDK Spaces may not include it.
 
 See [DEPLOYMENT.md](DEPLOYMENT.md) for the full deployment guide.
 

app.py

@@ -390,6 +390,9 @@ def ocr_pdf_pages(pdf_path: str, source_name: str, max_pages: int = 8) -> list:
             )
     return docs
 
+def has_readable_text(docs: list) -> bool:
+    return any(doc.page_content and doc.page_content.strip() for doc in docs)
+
 def normalize_source_metadata(docs: list, source_name: str, file_type: str, extraction: str = "text") -> list:
     for doc in docs:
         doc.metadata["source"] = source_name
@@ -421,6 +424,7 @@ def load_documents(files, use_ocr: bool = False, ocr_page_limit: int = 8) -> lis
                 if extracted_chars < 250:
                     st.write(f"🔍 Running OCR for scanned PDF: {file.name}")
                     loaded = ocr_pdf_pages(temp_path, file.name, max_pages=ocr_page_limit)
+            loaded = [doc for doc in loaded if doc.page_content and doc.page_content.strip()]
             docs.extend(loaded[:MAX_PAGES]) 
         except Exception as e:
             st.error(f"⚠️ Error loading `{file.name}`: {e}")
@@ -989,8 +993,6 @@ with st.sidebar:
     with st.expander("⚙️ Advanced Settings"):
         temperature = st.slider("Temperature", 0.0, 1.0, 0.3, 0.05)
         top_k = st.slider("Retrieved Chunks (Top-K)", 2, 10, 4)
-        use_ocr = st.checkbox("Enable OCR fallback for scanned PDFs", value=False)
-        ocr_page_limit = st.slider("OCR Page Limit", 1, 25, 8)
 
     st.divider()
 
@@ -1001,6 +1003,11 @@ with st.sidebar:
         accept_multiple_files=True,
     )
 
+    st.subheader("🔍 Scanned PDF OCR")
+    use_ocr = st.toggle("Enable OCR for scanned PDFs", value=True, help="Use this for image-based PDFs that have no selectable text.")
+    ocr_page_limit = st.slider("OCR page limit", 1, 25, 8, help="Higher values are slower on CPU Spaces.")
+    st.caption("OCR needs Tesseract. Use Docker Space for the most reliable scanned PDF support.")
+
     col1, col2 = st.columns(2)
     process_btn   = col1.button("🔄 Process", type="primary", use_container_width=True)
     summarize_btn = col2.button("📜 Summary", use_container_width=True)
@@ -1025,8 +1032,11 @@ with st.sidebar:
                     st.write("📥 Loading files into memory...")
                     raw_docs = load_documents(uploaded_files, use_ocr=use_ocr, ocr_page_limit=ocr_page_limit)
                     
-                    if not raw_docs:
+                    if not raw_docs or not has_readable_text(raw_docs):
                         status.update(label="No content found!", state="error")
+                        st.error(
+                            "No readable text was found. If this is a scanned PDF, keep OCR enabled and deploy with Docker so Tesseract OCR is installed."
+                        )
                         st.stop()
 
                     # BUG FIX: Save the full raw text into session state for the summary function
@@ -1040,6 +1050,13 @@ with st.sidebar:
                     )
                     chunks = splitter.split_documents(raw_docs)
 
+                    if not chunks:
+                        status.update(label="No searchable chunks created!", state="error")
+                        st.error(
+                            "The document loaded, but no searchable text chunks were created. For scanned PDFs, enable OCR and use Docker deployment with Tesseract."
+                        )
+                        st.stop()
+
                     st.write("🧠 Generating embeddings...")
                     embeddings = get_embeddings()