Update src/streamlit_app.py
Browse files- src/streamlit_app.py +18 -35
src/streamlit_app.py
CHANGED
|
@@ -212,7 +212,7 @@ st.caption("Query SAP documentation and enterprise PDFs β powered by reasoning
|
|
| 212 |
doc_choice = st.radio("Select a document:", ["-- Select --", "Sample PDF", "Upload Custom PDF"], index=0)
|
| 213 |
|
| 214 |
# ==========================================================
|
| 215 |
-
# π DOCUMENT HANDLING β
|
| 216 |
# ==========================================================
|
| 217 |
if doc_choice == "-- Select --":
|
| 218 |
st.info("β¬
οΈ Select or upload a document to begin.")
|
|
@@ -234,7 +234,7 @@ else:
|
|
| 234 |
with open(temp_path, "wb") as f:
|
| 235 |
f.write(uploaded_file.getbuffer())
|
| 236 |
else:
|
| 237 |
-
st.stop()
|
| 238 |
|
| 239 |
# --- Real processing begins here ---
|
| 240 |
if temp_path:
|
|
@@ -243,23 +243,22 @@ else:
|
|
| 243 |
# Process only once per document
|
| 244 |
if "doc_ready" not in st.session_state or st.session_state.get("last_doc") != doc_name:
|
| 245 |
status = st.empty()
|
| 246 |
-
status.info("π€ Upload complete β
|
| 247 |
|
|
|
|
| 248 |
text, toc, toc_source = extract_text_from_pdf(temp_path)
|
| 249 |
status.info("π Parsing and chunking document...")
|
| 250 |
chunks = chunk_text(text, chunk_size=chunk_size, overlap=overlap)
|
| 251 |
|
|
|
|
| 252 |
status.info("π§ Building embeddings and search index...")
|
| 253 |
embeddings = cache_embeddings(doc_name, chunks, embed_chunks)
|
| 254 |
index = build_faiss_index(embeddings)
|
| 255 |
|
| 256 |
-
#
|
| 257 |
-
|
| 258 |
-
status.success("β
Document processed successfully.")
|
| 259 |
-
time.sleep(1.2)
|
| 260 |
-
status.empty()
|
| 261 |
|
| 262 |
-
#
|
| 263 |
st.session_state.update({
|
| 264 |
"text": text,
|
| 265 |
"toc": toc,
|
|
@@ -268,19 +267,21 @@ else:
|
|
| 268 |
"index": index,
|
| 269 |
"doc_ready": True,
|
| 270 |
"last_doc": doc_name,
|
| 271 |
-
"status_text":
|
| 272 |
})
|
| 273 |
|
| 274 |
-
# Build suggestions once
|
| 275 |
query_suggestions = generate_dynamic_suggestions_from_toc(toc, chunks, doc_name)
|
| 276 |
st.session_state["query_suggestions_fixed"] = query_suggestions
|
| 277 |
st.session_state["user_query_input"] = ""
|
| 278 |
st.session_state["selected_suggestion"] = None
|
| 279 |
st.session_state["show_more"] = False
|
|
|
|
|
|
|
| 280 |
st.rerun()
|
| 281 |
|
| 282 |
else:
|
| 283 |
-
# Reuse
|
| 284 |
text = st.session_state["text"]
|
| 285 |
toc = st.session_state["toc"]
|
| 286 |
chunks = st.session_state["chunks"]
|
|
@@ -288,28 +289,9 @@ else:
|
|
| 288 |
index = st.session_state["index"]
|
| 289 |
query_suggestions = st.session_state.get("query_suggestions_fixed", [])
|
| 290 |
|
| 291 |
-
|
| 292 |
-
|
| 293 |
-
|
| 294 |
-
animation: fadeIn 1.2s ease-in-out;
|
| 295 |
-
background: #072541;
|
| 296 |
-
border-left: 4px solid #10b981;
|
| 297 |
-
padding: 10px 14px;
|
| 298 |
-
border-radius: 6px;
|
| 299 |
-
color: #d1fae5;
|
| 300 |
-
margin-top: 8px;
|
| 301 |
-
font-size: 15px;">
|
| 302 |
-
β
{st.session_state.get('status_text', f"{doc_name} ready for queries.")}
|
| 303 |
-
</div>
|
| 304 |
-
<style>
|
| 305 |
-
@keyframes fadeIn {{
|
| 306 |
-
from {{opacity: 0; transform: translateY(-4px);}}
|
| 307 |
-
to {{opacity: 1; transform: translateY(0);}}
|
| 308 |
-
}}
|
| 309 |
-
</style>
|
| 310 |
-
""", unsafe_allow_html=True)
|
| 311 |
-
|
| 312 |
-
# --- Ask section ---
|
| 313 |
st.markdown("### π¬ Ask the Assistant")
|
| 314 |
if query_suggestions:
|
| 315 |
visible = query_suggestions if st.session_state["show_more"] else query_suggestions[:3]
|
|
@@ -323,6 +305,7 @@ else:
|
|
| 323 |
st.session_state["show_more"] = not st.session_state["show_more"]
|
| 324 |
st.rerun()
|
| 325 |
|
|
|
|
| 326 |
user_query = st.text_input("Type your question or click one above:", key="user_query_input")
|
| 327 |
|
| 328 |
if user_query.strip():
|
|
@@ -334,13 +317,13 @@ else:
|
|
| 334 |
|
| 335 |
st.markdown("### π€ Assistantβs Answer")
|
| 336 |
|
|
|
|
| 337 |
if not reasoning_mode and not answer.startswith("β οΈ"):
|
| 338 |
answer = re.sub(r"\*\*(.*?)\*\*", r"\1", answer)
|
| 339 |
answer = re.sub(r"(^|\n)-\s*", r"\1<br>β’ ", answer)
|
| 340 |
st.markdown(f"<div class='answer-box'>{answer}</div>", unsafe_allow_html=True)
|
| 341 |
|
| 342 |
|
| 343 |
-
|
| 344 |
# ==========================================================
|
| 345 |
# π¨ Optional Sidebar Scroll Styling (keeps it clean)
|
| 346 |
# ==========================================================
|
|
|
|
| 212 |
doc_choice = st.radio("Select a document:", ["-- Select --", "Sample PDF", "Upload Custom PDF"], index=0)
|
| 213 |
|
| 214 |
# ==========================================================
|
| 215 |
+
# π DOCUMENT HANDLING β POLISHED UI FLOW
|
| 216 |
# ==========================================================
|
| 217 |
if doc_choice == "-- Select --":
|
| 218 |
st.info("β¬
οΈ Select or upload a document to begin.")
|
|
|
|
| 234 |
with open(temp_path, "wb") as f:
|
| 235 |
f.write(uploaded_file.getbuffer())
|
| 236 |
else:
|
| 237 |
+
st.stop() # Wait until file is uploaded
|
| 238 |
|
| 239 |
# --- Real processing begins here ---
|
| 240 |
if temp_path:
|
|
|
|
| 243 |
# Process only once per document
|
| 244 |
if "doc_ready" not in st.session_state or st.session_state.get("last_doc") != doc_name:
|
| 245 |
status = st.empty()
|
| 246 |
+
status.info("π€ Upload complete β hang tight while we process your document...")
|
| 247 |
|
| 248 |
+
# Step 1: Extract text
|
| 249 |
text, toc, toc_source = extract_text_from_pdf(temp_path)
|
| 250 |
status.info("π Parsing and chunking document...")
|
| 251 |
chunks = chunk_text(text, chunk_size=chunk_size, overlap=overlap)
|
| 252 |
|
| 253 |
+
# Step 2: Build embeddings + FAISS index
|
| 254 |
status.info("π§ Building embeddings and search index...")
|
| 255 |
embeddings = cache_embeddings(doc_name, chunks, embed_chunks)
|
| 256 |
index = build_faiss_index(embeddings)
|
| 257 |
|
| 258 |
+
# Step 3: Final ready state
|
| 259 |
+
status.success("β
All set β your AI assistant is ready to help.")
|
|
|
|
|
|
|
|
|
|
| 260 |
|
| 261 |
+
# Persist session data for reruns
|
| 262 |
st.session_state.update({
|
| 263 |
"text": text,
|
| 264 |
"toc": toc,
|
|
|
|
| 267 |
"index": index,
|
| 268 |
"doc_ready": True,
|
| 269 |
"last_doc": doc_name,
|
| 270 |
+
"status_text": "π Document is ready for queries."
|
| 271 |
})
|
| 272 |
|
| 273 |
+
# Build question suggestions (once per doc)
|
| 274 |
query_suggestions = generate_dynamic_suggestions_from_toc(toc, chunks, doc_name)
|
| 275 |
st.session_state["query_suggestions_fixed"] = query_suggestions
|
| 276 |
st.session_state["user_query_input"] = ""
|
| 277 |
st.session_state["selected_suggestion"] = None
|
| 278 |
st.session_state["show_more"] = False
|
| 279 |
+
|
| 280 |
+
# Refresh to cleanly show "ready" state
|
| 281 |
st.rerun()
|
| 282 |
|
| 283 |
else:
|
| 284 |
+
# --- Reuse existing state (rerun-safe) ---
|
| 285 |
text = st.session_state["text"]
|
| 286 |
toc = st.session_state["toc"]
|
| 287 |
chunks = st.session_state["chunks"]
|
|
|
|
| 289 |
index = st.session_state["index"]
|
| 290 |
query_suggestions = st.session_state.get("query_suggestions_fixed", [])
|
| 291 |
|
| 292 |
+
st.info(st.session_state.get("status_text", f"π {doc_name} is ready for queries."))
|
| 293 |
+
|
| 294 |
+
# --- Ask the Assistant section ---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 295 |
st.markdown("### π¬ Ask the Assistant")
|
| 296 |
if query_suggestions:
|
| 297 |
visible = query_suggestions if st.session_state["show_more"] else query_suggestions[:3]
|
|
|
|
| 305 |
st.session_state["show_more"] = not st.session_state["show_more"]
|
| 306 |
st.rerun()
|
| 307 |
|
| 308 |
+
# --- Query input box ---
|
| 309 |
user_query = st.text_input("Type your question or click one above:", key="user_query_input")
|
| 310 |
|
| 311 |
if user_query.strip():
|
|
|
|
| 317 |
|
| 318 |
st.markdown("### π€ Assistantβs Answer")
|
| 319 |
|
| 320 |
+
# Clean up answer format (bullets, bold)
|
| 321 |
if not reasoning_mode and not answer.startswith("β οΈ"):
|
| 322 |
answer = re.sub(r"\*\*(.*?)\*\*", r"\1", answer)
|
| 323 |
answer = re.sub(r"(^|\n)-\s*", r"\1<br>β’ ", answer)
|
| 324 |
st.markdown(f"<div class='answer-box'>{answer}</div>", unsafe_allow_html=True)
|
| 325 |
|
| 326 |
|
|
|
|
| 327 |
# ==========================================================
|
| 328 |
# π¨ Optional Sidebar Scroll Styling (keeps it clean)
|
| 329 |
# ==========================================================
|