Spaces:
Sleeping
Sleeping
CB commited on
Update streamlit_app.py
Browse files- streamlit_app.py +126 -455
streamlit_app.py
CHANGED
|
@@ -1,9 +1,11 @@
|
|
| 1 |
# streamlit_app.py
|
|
|
|
| 2 |
import os
|
| 3 |
import time
|
| 4 |
import string
|
| 5 |
import hashlib
|
| 6 |
import traceback
|
|
|
|
| 7 |
from glob import glob
|
| 8 |
from pathlib import Path
|
| 9 |
|
|
@@ -28,13 +30,9 @@ except Exception:
|
|
| 28 |
# google.generativeai SDK (guarded)
|
| 29 |
try:
|
| 30 |
import google.generativeai as genai
|
| 31 |
-
# older SDK exposed upload_file/get_file at top-level; try to bind them if present
|
| 32 |
-
upload_file = getattr(genai, "upload_file", None)
|
| 33 |
-
get_file = getattr(genai, "get_file", None)
|
| 34 |
HAS_GENAI = True
|
| 35 |
except Exception:
|
| 36 |
genai = None
|
| 37 |
-
upload_file = get_file = None
|
| 38 |
HAS_GENAI = False
|
| 39 |
|
| 40 |
st.set_page_config(page_title="Generate the story of videos", layout="wide")
|
|
@@ -147,84 +145,109 @@ def download_video_ytdlp(url: str, save_dir: str, video_password: str = None) ->
|
|
| 147 |
final = convert_video_to_mp4(chosen)
|
| 148 |
return final
|
| 149 |
|
| 150 |
-
def
|
| 151 |
-
if
|
| 152 |
-
return
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
|
| 156 |
-
for attr in ("name", "id", "fileId", "file_id"):
|
| 157 |
-
v = getattr(file_obj, attr, None)
|
| 158 |
-
if v:
|
| 159 |
-
return v
|
| 160 |
-
# sometimes SDK returns {'file': {...}}
|
| 161 |
try:
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
if isinstance(d, dict):
|
| 165 |
-
return d.get("name") or d.get("id") or d.get("fileId")
|
| 166 |
except Exception:
|
| 167 |
-
|
| 168 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 169 |
|
| 170 |
-
|
| 171 |
-
|
|
|
|
| 172 |
|
| 173 |
-
def configure_genai_if_needed():
|
| 174 |
-
key = get_effective_api_key()
|
| 175 |
-
if not key:
|
| 176 |
-
return False
|
| 177 |
try:
|
| 178 |
-
|
|
|
|
|
|
|
|
|
|
| 179 |
except Exception:
|
| 180 |
-
|
| 181 |
-
|
| 182 |
-
|
| 183 |
-
# ----
|
| 184 |
-
|
| 185 |
-
|
| 186 |
-
|
| 187 |
-
|
| 188 |
-
|
| 189 |
-
|
| 190 |
-
|
| 191 |
-
|
| 192 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 193 |
try:
|
| 194 |
-
genai.
|
| 195 |
-
|
| 196 |
-
|
|
|
|
|
|
|
|
|
|
| 197 |
except Exception:
|
| 198 |
-
|
| 199 |
-
|
| 200 |
-
|
| 201 |
-
|
| 202 |
-
|
| 203 |
-
|
| 204 |
-
|
| 205 |
-
st.session_state["last_local_path"] = ""
|
| 206 |
-
st.session_state["analysis_out"] = ""
|
| 207 |
-
st.session_state["last_error"] = ""
|
| 208 |
-
st.session_state["file_hash"] = None
|
| 209 |
-
for f in glob(str(DATA_DIR / "*")):
|
| 210 |
-
try:
|
| 211 |
-
os.remove(f)
|
| 212 |
-
except Exception:
|
| 213 |
-
pass
|
| 214 |
|
| 215 |
# Reset when URL changes (compare against last_url only)
|
| 216 |
current_url = st.session_state.get("url", "")
|
| 217 |
if current_url != st.session_state.get("last_url"):
|
| 218 |
if st.session_state.get("last_url"):
|
| 219 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 220 |
st.session_state["last_url"] = current_url
|
| 221 |
|
| 222 |
-
#
|
| 223 |
st.sidebar.header("Video Input")
|
| 224 |
st.sidebar.text_input("Video URL", key="url", placeholder="https://")
|
| 225 |
|
| 226 |
settings_exp = st.sidebar.expander("Settings", expanded=False)
|
| 227 |
-
model_choice = settings_exp.selectbox("Select model", options=MODEL_OPTIONS,
|
|
|
|
| 228 |
if model_choice == "custom":
|
| 229 |
model_input = settings_exp.text_input("Custom model id", value=DEFAULT_MODEL, key="model_input")
|
| 230 |
model_selected = model_input.strip() or DEFAULT_MODEL
|
|
@@ -254,191 +277,13 @@ settings_exp.number_input(
|
|
| 254 |
|
| 255 |
key_source = "session" if st.session_state.get("api_key") else ".env" if os.getenv("GOOGLE_API_KEY") else "none"
|
| 256 |
settings_exp.caption(f"Using API key from: {key_source}")
|
| 257 |
-
if not
|
| 258 |
-
settings_exp.warning("No Google API key provided;
|
| 259 |
-
|
| 260 |
-
safety_settings = [
|
| 261 |
-
{"category": "HARM_CATEGORY_HARASSMENT", "threshold": "OFF"},
|
| 262 |
-
{"category": "HARM_CATEGORY_HATE_SPEECH", "threshold": "OFF"},
|
| 263 |
-
{"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT", "threshold": "OFF"},
|
| 264 |
-
{"category": "HARM_CATEGORY_DANGEROUS_CONTENT", "threshold": "OFF"},
|
| 265 |
-
]
|
| 266 |
-
|
| 267 |
-
# ---- Upload & processing helpers ----
|
| 268 |
-
def _upload_with_kwargs(fn, filepath):
|
| 269 |
-
"""
|
| 270 |
-
Call an upload function `fn` with the required ragStoreName argument.
|
| 271 |
-
Most recent google.generativeai SDK versions expect:
|
| 272 |
-
upload_file(path, ragStoreName="default")
|
| 273 |
-
Older versions accept just the path, so we try both signatures.
|
| 274 |
-
"""
|
| 275 |
-
try:
|
| 276 |
-
# Preferred modern signature
|
| 277 |
-
return fn(filepath, ragStoreName="default")
|
| 278 |
-
except TypeError:
|
| 279 |
-
# Fallback to older signature (no ragStoreName)
|
| 280 |
-
return fn(filepath)
|
| 281 |
-
except Exception as e:
|
| 282 |
-
# Propagate any other error for higher‑level handling
|
| 283 |
-
raise e
|
| 284 |
-
|
| 285 |
-
|
| 286 |
-
def upload_video_sdk(filepath: str):
|
| 287 |
-
"""Upload a video file using whichever upload function the SDK provides."""
|
| 288 |
-
key = get_effective_api_key()
|
| 289 |
-
if not key:
|
| 290 |
-
raise RuntimeError("No API key provided")
|
| 291 |
-
if not HAS_GENAI or genai is None:
|
| 292 |
-
raise RuntimeError("google.generativeai SDK not available; cannot upload")
|
| 293 |
-
|
| 294 |
-
genai.configure(api_key=key)
|
| 295 |
-
|
| 296 |
-
# Build a list of possible upload callables (newest first)
|
| 297 |
-
candidate_calls = []
|
| 298 |
-
|
| 299 |
-
# Newer SDK exposes upload_file directly
|
| 300 |
-
if upload_file:
|
| 301 |
-
candidate_calls.append(("upload_file", upload_file))
|
| 302 |
-
|
| 303 |
-
# Some installations expose it under ragstore or files namespaces
|
| 304 |
-
candidate_calls.append(
|
| 305 |
-
("genai.ragstore.upload_file",
|
| 306 |
-
getattr(genai, "ragstore", None) and getattr(genai.ragstore, "upload_file", None))
|
| 307 |
-
)
|
| 308 |
-
candidate_calls.append(
|
| 309 |
-
("genai.files.upload",
|
| 310 |
-
getattr(genai, "files", None) and getattr(genai.files, "upload", None))
|
| 311 |
-
)
|
| 312 |
-
candidate_calls.append(
|
| 313 |
-
("genai.upload",
|
| 314 |
-
getattr(genai, "upload", None))
|
| 315 |
-
)
|
| 316 |
-
|
| 317 |
-
# Filter out any None entries
|
| 318 |
-
candidate_calls = [(name, fn) for name, fn in candidate_calls if fn]
|
| 319 |
-
|
| 320 |
-
if not candidate_calls:
|
| 321 |
-
raise RuntimeError("No upload function available in google.generativeai SDK")
|
| 322 |
-
|
| 323 |
-
last_exc = None
|
| 324 |
-
for name, fn in candidate_calls:
|
| 325 |
-
try:
|
| 326 |
-
# Use the wrapper that injects ragStoreName when needed
|
| 327 |
-
res = _upload_with_kwargs(fn, filepath)
|
| 328 |
-
# Normalise the return value so the rest of the code works
|
| 329 |
-
return _normalize_uploaded_obj(res)
|
| 330 |
-
except Exception as e:
|
| 331 |
-
last_exc = e
|
| 332 |
-
# Log the attempted method for debugging (optional)
|
| 333 |
-
st.session_state["last_error"] = f"Upload attempt '{name}' failed: {e}"
|
| 334 |
-
continue
|
| 335 |
-
|
| 336 |
-
# If we get here every method failed
|
| 337 |
-
raise RuntimeError(f"All upload methods failed. Last error: {last_exc}")
|
| 338 |
-
|
| 339 |
-
def wait_for_processed(file_obj, timeout: int = None):
|
| 340 |
-
if timeout is None:
|
| 341 |
-
timeout = st.session_state.get("processing_timeout", 900)
|
| 342 |
-
if not HAS_GENAI or genai is None:
|
| 343 |
-
return file_obj
|
| 344 |
-
start = time.time()
|
| 345 |
-
name = file_name_or_id(file_obj)
|
| 346 |
-
# If no determinable name/id, just return file_obj
|
| 347 |
-
if not name:
|
| 348 |
-
return file_obj
|
| 349 |
-
genai.configure(api_key=get_effective_api_key())
|
| 350 |
-
backoff = 1.0
|
| 351 |
-
last_exc = None
|
| 352 |
-
# prefer get_file if present, otherwise try genai.ragstore.get or genai.files.get
|
| 353 |
-
candidate_getters = []
|
| 354 |
-
if get_file:
|
| 355 |
-
candidate_getters.append(("get_file", get_file))
|
| 356 |
-
candidate_getters.append(("genai.ragstore.get", getattr(genai, "ragstore", None) and getattr(genai.ragstore, "get", None)))
|
| 357 |
-
candidate_getters.append(("genai.files.get", getattr(genai, "files", None) and getattr(genai.files, "get", None)))
|
| 358 |
-
candidate_getters = [(n, fn) for n, fn in candidate_getters if fn]
|
| 359 |
-
# If none, return file_obj immediately
|
| 360 |
-
if not candidate_getters:
|
| 361 |
-
return file_obj
|
| 362 |
-
|
| 363 |
-
while True:
|
| 364 |
-
for name_label, getter in candidate_getters:
|
| 365 |
-
try:
|
| 366 |
-
obj = getter(name)
|
| 367 |
-
# normalize possible SDK object/dict
|
| 368 |
-
# check for state attribute or dict field
|
| 369 |
-
state = None
|
| 370 |
-
if isinstance(obj, dict):
|
| 371 |
-
state = obj.get("state") or (obj.get("status") and {"name": obj.get("status")})
|
| 372 |
-
else:
|
| 373 |
-
state = getattr(obj, "state", None) or getattr(obj, "status", None)
|
| 374 |
-
if not state:
|
| 375 |
-
return obj
|
| 376 |
-
# state might be dict or object with .name
|
| 377 |
-
state_name = state.get("name") if isinstance(state, dict) else getattr(state, "name", None)
|
| 378 |
-
if state_name and state_name.upper() == "PROCESSING":
|
| 379 |
-
# still processing; continue polling
|
| 380 |
-
last_obj = obj
|
| 381 |
-
break
|
| 382 |
-
return obj
|
| 383 |
-
except Exception as e:
|
| 384 |
-
last_exc = e
|
| 385 |
-
# transient errors: backoff and retry until timeout
|
| 386 |
-
continue
|
| 387 |
-
if time.time() - start > timeout:
|
| 388 |
-
raise TimeoutError(f"File processing timed out. Last error: {last_exc}")
|
| 389 |
-
time.sleep(backoff)
|
| 390 |
-
backoff = min(backoff * 2, 8.0)
|
| 391 |
-
|
| 392 |
-
def remove_prompt_echo(prompt: str, text: str, check_len: int = 600, ratio_threshold: float = 0.68):
|
| 393 |
-
if not prompt or not text:
|
| 394 |
-
return text
|
| 395 |
-
a = " ".join(prompt.strip().lower().split())
|
| 396 |
-
b_full = text.strip()
|
| 397 |
-
b = " ".join(b_full[:check_len].lower().split())
|
| 398 |
-
try:
|
| 399 |
-
from difflib import SequenceMatcher
|
| 400 |
-
ratio = SequenceMatcher(None, a, b).ratio()
|
| 401 |
-
except Exception:
|
| 402 |
-
ratio = 0.0
|
| 403 |
-
if ratio >= ratio_threshold:
|
| 404 |
-
cut = min(len(b_full), max(int(len(prompt) * 0.9), len(a)))
|
| 405 |
-
new_text = b_full[cut:].lstrip(" \n:-")
|
| 406 |
-
if len(new_text) >= 3:
|
| 407 |
-
return new_text
|
| 408 |
-
placeholders = ["enter analysis", "enter your analysis", "enter analysis here", "please enter analysis"]
|
| 409 |
-
low = b_full.strip().lower()
|
| 410 |
-
for ph in placeholders:
|
| 411 |
-
if low.startswith(ph):
|
| 412 |
-
return b_full[len(ph):].lstrip(" \n:-")
|
| 413 |
-
return text
|
| 414 |
-
|
| 415 |
-
def compress_video_if_large(local_path: str, threshold_mb: int = 200):
|
| 416 |
-
try:
|
| 417 |
-
file_size_mb = os.path.getsize(local_path) / (1024 * 1024)
|
| 418 |
-
except Exception:
|
| 419 |
-
st.session_state["last_error"] = "Failed to stat file before compression"
|
| 420 |
-
return local_path, False
|
| 421 |
-
|
| 422 |
-
if file_size_mb <= threshold_mb:
|
| 423 |
-
return local_path, False
|
| 424 |
-
|
| 425 |
-
p = Path(local_path)
|
| 426 |
-
compressed_name = f"{p.stem}_compressed.mp4"
|
| 427 |
-
compressed_path = str(p.with_name(compressed_name))
|
| 428 |
-
|
| 429 |
-
try:
|
| 430 |
-
result = compress_video(local_path, compressed_path, crf=28, preset="fast")
|
| 431 |
-
if result and os.path.exists(result) and os.path.getsize(result) > 0:
|
| 432 |
-
return result, True
|
| 433 |
-
return local_path, False
|
| 434 |
-
except Exception:
|
| 435 |
-
st.session_state["last_error"] = "Video compression failed"
|
| 436 |
-
return local_path, False
|
| 437 |
|
| 438 |
-
#
|
| 439 |
col1, col2 = st.columns([1, 3])
|
| 440 |
with col1:
|
| 441 |
-
generate_now = st.button("Generate the story", type="primary"
|
| 442 |
with col2:
|
| 443 |
if not st.session_state.get("videos"):
|
| 444 |
st.info("Load a video first (sidebar) to enable generation.", icon="ℹ️")
|
|
@@ -453,10 +298,7 @@ if st.sidebar.button("Load Video", use_container_width=True):
|
|
| 453 |
st.session_state["last_local_path"] = path
|
| 454 |
st.session_state.pop("uploaded_file", None)
|
| 455 |
st.session_state.pop("processed_file", None)
|
| 456 |
-
|
| 457 |
-
st.session_state["file_hash"] = file_sha256(path)
|
| 458 |
-
except Exception:
|
| 459 |
-
st.session_state["file_hash"] = None
|
| 460 |
except Exception as e:
|
| 461 |
st.sidebar.error(f"Failed to load video: {e}")
|
| 462 |
|
|
@@ -475,11 +317,25 @@ if st.session_state["videos"]:
|
|
| 475 |
st.session_state["loop_video"] = loop_checkbox
|
| 476 |
|
| 477 |
if st.button("Clear Video(s)"):
|
| 478 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 479 |
|
| 480 |
try:
|
| 481 |
with open(st.session_state["videos"], "rb") as vf:
|
| 482 |
-
st.download_button("Download Video", data=vf,
|
|
|
|
|
|
|
| 483 |
except Exception:
|
| 484 |
st.sidebar.error("Failed to prepare download")
|
| 485 |
|
|
@@ -488,226 +344,41 @@ if st.session_state["videos"]:
|
|
| 488 |
file_size_mb = os.path.getsize(st.session_state["videos"]) / (1024 * 1024)
|
| 489 |
st.sidebar.caption(f"File size: {file_size_mb:.1f} MB")
|
| 490 |
if file_size_mb > st.session_state.get("compress_threshold_mb", 200):
|
| 491 |
-
st.sidebar.warning(
|
|
|
|
|
|
|
|
|
|
| 492 |
except Exception:
|
| 493 |
pass
|
| 494 |
|
| 495 |
-
#
|
| 496 |
if generate_now and not st.session_state.get("busy"):
|
| 497 |
if not st.session_state.get("videos"):
|
| 498 |
st.error("No video loaded. Use 'Load Video' in the sidebar.")
|
| 499 |
else:
|
| 500 |
-
|
| 501 |
-
if not
|
| 502 |
st.error("Google API key not set.")
|
| 503 |
else:
|
|
|
|
| 504 |
try:
|
| 505 |
-
|
| 506 |
-
try:
|
| 507 |
-
if HAS_GENAI and genai is not None:
|
| 508 |
-
genai.configure(api_key=key_to_use)
|
| 509 |
-
except Exception:
|
| 510 |
-
pass
|
| 511 |
-
|
| 512 |
model_id = (st.session_state.get("model_input") or model_selected or DEFAULT_MODEL).strip()
|
| 513 |
-
if st.session_state.get("last_model") != model_id:
|
| 514 |
-
st.session_state["last_model"] = ""
|
| 515 |
-
maybe_create_agent(model_id)
|
| 516 |
-
|
| 517 |
-
processed = st.session_state.get("processed_file")
|
| 518 |
-
current_path = st.session_state.get("videos")
|
| 519 |
-
try:
|
| 520 |
-
current_hash = file_sha256(current_path) if current_path and os.path.exists(current_path) else None
|
| 521 |
-
except Exception:
|
| 522 |
-
current_hash = None
|
| 523 |
-
|
| 524 |
-
reupload_needed = True
|
| 525 |
-
uploaded_file = st.session_state.get("uploaded_file")
|
| 526 |
-
uploaded_name = file_name_or_id(uploaded_file)
|
| 527 |
-
if processed and st.session_state.get("last_local_path") == current_path and st.session_state.get("file_hash") == current_hash and uploaded_name:
|
| 528 |
-
reupload_needed = False
|
| 529 |
-
|
| 530 |
-
if reupload_needed:
|
| 531 |
-
if not HAS_GENAI:
|
| 532 |
-
raise RuntimeError("google.generativeai SDK not available; install it.")
|
| 533 |
-
local_path = current_path
|
| 534 |
-
upload_path, compressed = compress_video_if_large(local_path, threshold_mb=st.session_state.get("compress_threshold_mb", 200))
|
| 535 |
-
|
| 536 |
-
with st.spinner(f"Uploading video{' (compressed)' if compressed else ''}..."):
|
| 537 |
-
try:
|
| 538 |
-
uploaded = upload_video_sdk(upload_path)
|
| 539 |
-
except Exception as e:
|
| 540 |
-
st.session_state["last_error"] = f"Upload failed for {upload_path}: {e}\n{traceback.format_exc()}"
|
| 541 |
-
st.error(f"Upload failed: {e}. Check the error log for more details.")
|
| 542 |
-
st.session_state["busy"] = False
|
| 543 |
-
raise
|
| 544 |
-
|
| 545 |
-
try:
|
| 546 |
-
processing_placeholder = st.empty()
|
| 547 |
-
processing_bar = processing_placeholder.progress(0)
|
| 548 |
-
start_time = time.time()
|
| 549 |
-
processed = wait_for_processed(uploaded, timeout=st.session_state.get("processing_timeout", 900))
|
| 550 |
-
elapsed = time.time() - start_time
|
| 551 |
-
timeout = st.session_state.get("processing_timeout", 900)
|
| 552 |
-
pct = min(100, int((elapsed / timeout) * 100)) if timeout > 0 else 0
|
| 553 |
-
processing_bar.progress(pct)
|
| 554 |
-
processing_placeholder.success("Processing complete")
|
| 555 |
-
except Exception as e:
|
| 556 |
-
st.session_state["last_error"] = f"Processing failed/wait timeout: {e}"
|
| 557 |
-
st.error("Video processing failed or timed out. See Last Error.")
|
| 558 |
-
st.session_state["busy"] = False
|
| 559 |
-
raise
|
| 560 |
-
|
| 561 |
-
st.session_state["uploaded_file"] = uploaded
|
| 562 |
-
st.session_state["processed_file"] = processed
|
| 563 |
-
st.session_state["last_local_path"] = current_path
|
| 564 |
-
st.session_state["file_hash"] = current_hash
|
| 565 |
-
|
| 566 |
prompt_text = (analysis_prompt.strip() or DEFAULT_PROMPT).strip()
|
| 567 |
-
out = ""
|
| 568 |
-
model_used = model_id
|
| 569 |
-
max_tokens = 2048 if "2.5" in model_used else 1024
|
| 570 |
-
est_tokens = max_tokens
|
| 571 |
-
|
| 572 |
-
# Try Agent first, fallback to Responses API
|
| 573 |
-
agent = maybe_create_agent(model_used)
|
| 574 |
-
debug_info = {"agent_attempted": False, "agent_ok": False, "agent_error": None, "agent_response_has_text": False}
|
| 575 |
-
if agent:
|
| 576 |
-
debug_info["agent_attempted"] = True
|
| 577 |
-
try:
|
| 578 |
-
with st.spinner("Generating description via Agent..."):
|
| 579 |
-
if not processed:
|
| 580 |
-
raise RuntimeError("Processed file missing for agent generation")
|
| 581 |
-
agent_response = agent.run(prompt_text, videos=[processed], safety_settings=safety_settings)
|
| 582 |
-
agent_text = getattr(agent_response, "content", None) or getattr(agent_response, "outputText", None) or None
|
| 583 |
-
if not agent_text:
|
| 584 |
-
if isinstance(agent_response, dict):
|
| 585 |
-
for k in ("content", "outputText", "text", "message"):
|
| 586 |
-
if k in agent_response and agent_response[k]:
|
| 587 |
-
agent_text = agent_response[k]
|
| 588 |
-
break
|
| 589 |
-
if agent_text and str(agent_text).strip():
|
| 590 |
-
out = str(agent_text).strip()
|
| 591 |
-
debug_info["agent_ok"] = True
|
| 592 |
-
debug_info["agent_response_has_text"] = True
|
| 593 |
-
else:
|
| 594 |
-
debug_info["agent_ok"] = False
|
| 595 |
-
except Exception as ae:
|
| 596 |
-
debug_info["agent_error"] = f"{ae}"
|
| 597 |
-
|
| 598 |
-
if not out:
|
| 599 |
-
def generate_via_responses_api(prompt_text: str, processed, model_used: str, max_tokens: int = 1024, timeout: int = 300):
|
| 600 |
-
key = get_effective_api_key()
|
| 601 |
-
if not key:
|
| 602 |
-
raise RuntimeError("No API key provided")
|
| 603 |
-
if not HAS_GENAI or genai is None:
|
| 604 |
-
raise RuntimeError("Responses API not available")
|
| 605 |
-
genai.configure(api_key=key)
|
| 606 |
-
fname = file_name_or_id(processed)
|
| 607 |
-
if not fname:
|
| 608 |
-
raise RuntimeError("Uploaded file missing name/id")
|
| 609 |
-
|
| 610 |
-
system_msg = {"role": "system", "content": prompt_text}
|
| 611 |
-
user_msg = {"role": "user", "content": "Please summarize the attached video."}
|
| 612 |
-
|
| 613 |
-
call_variants = [
|
| 614 |
-
{"messages": [system_msg, user_msg], "files": [{"name": fname}], "safety_settings": safety_settings, "max_output_tokens": max_tokens},
|
| 615 |
-
{"input": [{"text": prompt_text, "files": [{"name": fname}]}], "safety_settings": safety_settings, "max_output_tokens": max_tokens},
|
| 616 |
-
]
|
| 617 |
-
|
| 618 |
-
last_exc = None
|
| 619 |
-
start = time.time()
|
| 620 |
-
backoff = 1.0
|
| 621 |
-
while True:
|
| 622 |
-
for payload in call_variants:
|
| 623 |
-
try:
|
| 624 |
-
response = genai.responses.generate(model=model_used, **payload)
|
| 625 |
-
return _normalize_genai_response(response)
|
| 626 |
-
except Exception as e:
|
| 627 |
-
last_exc = e
|
| 628 |
-
msg = str(e).lower()
|
| 629 |
-
if any(k in msg for k in ("internal", "unavailable", "deadlineexceeded", "deadline exceeded", "timeout", "rate limit")):
|
| 630 |
-
continue
|
| 631 |
-
raise
|
| 632 |
-
if time.time() - start > timeout:
|
| 633 |
-
raise TimeoutError("Responses.generate timed out")
|
| 634 |
-
time.sleep(backoff)
|
| 635 |
-
backoff = min(backoff * 2, 8.0)
|
| 636 |
-
|
| 637 |
-
def _normalize_genai_response(response):
|
| 638 |
-
outputs = []
|
| 639 |
-
if response is None:
|
| 640 |
-
return ""
|
| 641 |
-
text_pieces = []
|
| 642 |
-
try:
|
| 643 |
-
if isinstance(response, dict):
|
| 644 |
-
for key in ("output", "candidates", "items", "responses", "choices"):
|
| 645 |
-
val = response.get(key)
|
| 646 |
-
if isinstance(val, list) and val:
|
| 647 |
-
for item in val:
|
| 648 |
-
if isinstance(item, dict):
|
| 649 |
-
for k in ("content", "text", "message", "output_text", "output"):
|
| 650 |
-
t = item.get(k)
|
| 651 |
-
if t:
|
| 652 |
-
text_pieces.append(str(t).strip())
|
| 653 |
-
break
|
| 654 |
-
elif isinstance(item, str):
|
| 655 |
-
text_pieces.append(item.strip())
|
| 656 |
-
if not text_pieces:
|
| 657 |
-
for k in ("text", "message", "output_text"):
|
| 658 |
-
v = response.get(k)
|
| 659 |
-
if v:
|
| 660 |
-
text_pieces.append(str(v).strip())
|
| 661 |
-
break
|
| 662 |
-
else:
|
| 663 |
-
# SDK object: try attributes
|
| 664 |
-
for attr in ("output", "candidates", "text", "message", "content"):
|
| 665 |
-
v = getattr(response, attr, None)
|
| 666 |
-
if v:
|
| 667 |
-
text_pieces.append(str(v).strip())
|
| 668 |
-
if not text_pieces:
|
| 669 |
-
text_pieces.append(str(response))
|
| 670 |
-
except Exception:
|
| 671 |
-
text_pieces.append(str(response))
|
| 672 |
-
seen = set()
|
| 673 |
-
filtered = []
|
| 674 |
-
for t in text_pieces:
|
| 675 |
-
if not isinstance(t, str):
|
| 676 |
-
continue
|
| 677 |
-
if t and t not in seen:
|
| 678 |
-
filtered.append(t)
|
| 679 |
-
seen.add(t)
|
| 680 |
-
return "\n\n".join(filtered).strip()
|
| 681 |
-
|
| 682 |
-
try:
|
| 683 |
-
with st.spinner("Generating description via Responses API..."):
|
| 684 |
-
out = generate_via_responses_api(prompt_text, processed, model_used, max_tokens=max_tokens, timeout=st.session_state.get("generation_timeout", 300))
|
| 685 |
-
except Exception as e:
|
| 686 |
-
st.session_state["last_error"] = f"Responses API error: {e}\nDebug: {debug_info}"
|
| 687 |
-
st.error("An error occurred while generating the story. You can try Generate again; the uploaded video will be reused.")
|
| 688 |
-
out = ""
|
| 689 |
-
|
| 690 |
-
if out:
|
| 691 |
-
out = remove_prompt_echo(prompt_text, out)
|
| 692 |
-
p = prompt_text
|
| 693 |
-
if p and out.strip().lower().startswith(p.lower()):
|
| 694 |
-
out = out.strip()[len(p):].lstrip(" \n:-")
|
| 695 |
-
placeholders = ["enter analysis", "enter your analysis", "enter analysis here", "please enter analysis"]
|
| 696 |
-
low = out.strip().lower()
|
| 697 |
-
for ph in placeholders:
|
| 698 |
-
if low.startswith(ph):
|
| 699 |
-
out = out.strip()[len(ph):].lstrip(" \n:-")
|
| 700 |
-
break
|
| 701 |
-
out = out.strip()
|
| 702 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 703 |
st.session_state["analysis_out"] = out
|
| 704 |
st.session_state["last_error"] = ""
|
| 705 |
st.subheader("Analysis Result")
|
| 706 |
-
st.markdown(out
|
| 707 |
-
st.caption(f"Est. max tokens: {est_tokens}")
|
| 708 |
-
|
| 709 |
except Exception as e:
|
| 710 |
-
st.session_state["last_error"] = f"{e}\n{traceback.format_exc()}"
|
| 711 |
-
st.error("An error occurred while generating the story
|
| 712 |
finally:
|
| 713 |
st.session_state["busy"] = False
|
|
|
|
| 1 |
# streamlit_app.py
|
| 2 |
+
|
| 3 |
import os
|
| 4 |
import time
|
| 5 |
import string
|
| 6 |
import hashlib
|
| 7 |
import traceback
|
| 8 |
+
import base64
|
| 9 |
from glob import glob
|
| 10 |
from pathlib import Path
|
| 11 |
|
|
|
|
| 30 |
# google.generativeai SDK (guarded)
|
| 31 |
try:
|
| 32 |
import google.generativeai as genai
|
|
|
|
|
|
|
|
|
|
| 33 |
HAS_GENAI = True
|
| 34 |
except Exception:
|
| 35 |
genai = None
|
|
|
|
| 36 |
HAS_GENAI = False
|
| 37 |
|
| 38 |
st.set_page_config(page_title="Generate the story of videos", layout="wide")
|
|
|
|
| 145 |
final = convert_video_to_mp4(chosen)
|
| 146 |
return final
|
| 147 |
|
| 148 |
+
def remove_prompt_echo(prompt: str, text: str, check_len: int = 600, ratio_threshold: float = 0.68):
|
| 149 |
+
if not prompt or not text:
|
| 150 |
+
return text
|
| 151 |
+
a = " ".join(prompt.strip().lower().split())
|
| 152 |
+
b_full = text.strip()
|
| 153 |
+
b = " ".join(b_full[:check_len].lower().split())
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 154 |
try:
|
| 155 |
+
from difflib import SequenceMatcher
|
| 156 |
+
ratio = SequenceMatcher(None, a, b).ratio()
|
|
|
|
|
|
|
| 157 |
except Exception:
|
| 158 |
+
ratio = 0.0
|
| 159 |
+
if ratio >= ratio_threshold:
|
| 160 |
+
cut = min(len(b_full), max(int(len(prompt) * 0.9), len(a)))
|
| 161 |
+
new_text = b_full[cut:].lstrip(" \n:-")
|
| 162 |
+
if len(new_text) >= 3:
|
| 163 |
+
return new_text
|
| 164 |
+
placeholders = ["enter analysis", "enter your analysis", "enter analysis here", "please enter analysis"]
|
| 165 |
+
low = b_full.strip().lower()
|
| 166 |
+
for ph in placeholders:
|
| 167 |
+
if low.startswith(ph):
|
| 168 |
+
return b_full[len(ph):].lstrip(" \n:-")
|
| 169 |
+
return text
|
| 170 |
+
|
| 171 |
+
def compress_video_if_large(local_path: str, threshold_mb: int = 200):
|
| 172 |
+
try:
|
| 173 |
+
file_size_mb = os.path.getsize(local_path) / (1024 * 1024)
|
| 174 |
+
except Exception:
|
| 175 |
+
st.session_state["last_error"] = "Failed to stat file before compression"
|
| 176 |
+
return local_path, False
|
| 177 |
+
|
| 178 |
+
if file_size_mb <= threshold_mb:
|
| 179 |
+
return local_path, False
|
| 180 |
|
| 181 |
+
p = Path(local_path)
|
| 182 |
+
compressed_name = f"{p.stem}_compressed.mp4"
|
| 183 |
+
compressed_path = str(p.with_name(compressed_name))
|
| 184 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 185 |
try:
|
| 186 |
+
result = compress_video(local_path, compressed_path, crf=28, preset="fast")
|
| 187 |
+
if result and os.path.exists(result) and os.path.getsize(result) > 0:
|
| 188 |
+
return result, True
|
| 189 |
+
return local_path, False
|
| 190 |
except Exception:
|
| 191 |
+
st.session_state["last_error"] = "Video compression failed"
|
| 192 |
+
return local_path, False
|
| 193 |
+
|
| 194 |
+
# ---- Inline-video generation via base64 (bypass upload) ----
|
| 195 |
+
def generate_with_inline_video(local_path: str, prompt: str, model_used: str, timeout: int = 300):
|
| 196 |
+
# Read the video bytes
|
| 197 |
+
with open(local_path, "rb") as f:
|
| 198 |
+
video_bytes = f.read()
|
| 199 |
+
b64 = base64.b64encode(video_bytes).decode("utf-8")
|
| 200 |
+
video_part = {
|
| 201 |
+
"inline_data": {
|
| 202 |
+
"mime_type": "video/mp4",
|
| 203 |
+
"data": b64
|
| 204 |
+
}
|
| 205 |
+
}
|
| 206 |
+
contents = [prompt, video_part]
|
| 207 |
+
|
| 208 |
+
# Use new client API if present
|
| 209 |
try:
|
| 210 |
+
client = genai.Client()
|
| 211 |
+
resp = client.models.generate_content(
|
| 212 |
+
model=model_used,
|
| 213 |
+
contents=contents,
|
| 214 |
+
generation_config={"max_output_tokens": 1024}
|
| 215 |
+
)
|
| 216 |
except Exception:
|
| 217 |
+
# Fallback older style
|
| 218 |
+
resp = genai.GenerativeModel(model_used).generate_content(contents)
|
| 219 |
+
|
| 220 |
+
text = getattr(resp, "text", None) or getattr(resp, "output_text", None) or str(resp)
|
| 221 |
+
return text
|
| 222 |
+
|
| 223 |
+
# ---- Main UI & logic ----
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 224 |
|
| 225 |
# Reset when URL changes (compare against last_url only)
|
| 226 |
current_url = st.session_state.get("url", "")
|
| 227 |
if current_url != st.session_state.get("last_url"):
|
| 228 |
if st.session_state.get("last_url"):
|
| 229 |
+
# clear old video state
|
| 230 |
+
st.session_state.pop("uploaded_file", None)
|
| 231 |
+
st.session_state.pop("processed_file", None)
|
| 232 |
+
st.session_state["videos"] = ""
|
| 233 |
+
st.session_state["last_local_path"] = ""
|
| 234 |
+
st.session_state["analysis_out"] = ""
|
| 235 |
+
st.session_state["last_error"] = ""
|
| 236 |
+
st.session_state["file_hash"] = None
|
| 237 |
+
for f in glob(str(DATA_DIR / "*")):
|
| 238 |
+
try:
|
| 239 |
+
os.remove(f)
|
| 240 |
+
except Exception:
|
| 241 |
+
pass
|
| 242 |
st.session_state["last_url"] = current_url
|
| 243 |
|
| 244 |
+
# Sidebar UI
|
| 245 |
st.sidebar.header("Video Input")
|
| 246 |
st.sidebar.text_input("Video URL", key="url", placeholder="https://")
|
| 247 |
|
| 248 |
settings_exp = st.sidebar.expander("Settings", expanded=False)
|
| 249 |
+
model_choice = settings_exp.selectbox("Select model", options=MODEL_OPTIONS,
|
| 250 |
+
index=MODEL_OPTIONS.index(DEFAULT_MODEL) if DEFAULT_MODEL in MODEL_OPTIONS else 0)
|
| 251 |
if model_choice == "custom":
|
| 252 |
model_input = settings_exp.text_input("Custom model id", value=DEFAULT_MODEL, key="model_input")
|
| 253 |
model_selected = model_input.strip() or DEFAULT_MODEL
|
|
|
|
| 277 |
|
| 278 |
key_source = "session" if st.session_state.get("api_key") else ".env" if os.getenv("GOOGLE_API_KEY") else "none"
|
| 279 |
settings_exp.caption(f"Using API key from: {key_source}")
|
| 280 |
+
if not st.session_state.get("api_key") and not os.getenv("GOOGLE_API_KEY"):
|
| 281 |
+
settings_exp.warning("No Google API key provided; generation disabled.", icon="⚠️")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 282 |
|
| 283 |
+
# Buttons & layout
|
| 284 |
col1, col2 = st.columns([1, 3])
|
| 285 |
with col1:
|
| 286 |
+
generate_now = st.button("Generate the story", type="primary")
|
| 287 |
with col2:
|
| 288 |
if not st.session_state.get("videos"):
|
| 289 |
st.info("Load a video first (sidebar) to enable generation.", icon="ℹ️")
|
|
|
|
| 298 |
st.session_state["last_local_path"] = path
|
| 299 |
st.session_state.pop("uploaded_file", None)
|
| 300 |
st.session_state.pop("processed_file", None)
|
| 301 |
+
st.session_state["file_hash"] = file_sha256(path)
|
|
|
|
|
|
|
|
|
|
| 302 |
except Exception as e:
|
| 303 |
st.sidebar.error(f"Failed to load video: {e}")
|
| 304 |
|
|
|
|
| 317 |
st.session_state["loop_video"] = loop_checkbox
|
| 318 |
|
| 319 |
if st.button("Clear Video(s)"):
|
| 320 |
+
# Clear video state
|
| 321 |
+
st.session_state.pop("uploaded_file", None)
|
| 322 |
+
st.session_state.pop("processed_file", None)
|
| 323 |
+
st.session_state["videos"] = ""
|
| 324 |
+
st.session_state["last_local_path"] = ""
|
| 325 |
+
st.session_state["analysis_out"] = ""
|
| 326 |
+
st.session_state["last_error"] = ""
|
| 327 |
+
st.session_state["file_hash"] = None
|
| 328 |
+
for f in glob(str(DATA_DIR / "*")):
|
| 329 |
+
try:
|
| 330 |
+
os.remove(f)
|
| 331 |
+
except Exception:
|
| 332 |
+
pass
|
| 333 |
|
| 334 |
try:
|
| 335 |
with open(st.session_state["videos"], "rb") as vf:
|
| 336 |
+
st.download_button("Download Video", data=vf,
|
| 337 |
+
file_name=sanitize_filename(st.session_state["videos"]),
|
| 338 |
+
mime="video/mp4", use_container_width=True)
|
| 339 |
except Exception:
|
| 340 |
st.sidebar.error("Failed to prepare download")
|
| 341 |
|
|
|
|
| 344 |
file_size_mb = os.path.getsize(st.session_state["videos"]) / (1024 * 1024)
|
| 345 |
st.sidebar.caption(f"File size: {file_size_mb:.1f} MB")
|
| 346 |
if file_size_mb > st.session_state.get("compress_threshold_mb", 200):
|
| 347 |
+
st.sidebar.warning(
|
| 348 |
+
f"Large file detected — it may exceed inline size limits (>{st.session_state.get('compress_threshold_mb')} MB).",
|
| 349 |
+
icon="⚠️"
|
| 350 |
+
)
|
| 351 |
except Exception:
|
| 352 |
pass
|
| 353 |
|
| 354 |
+
# Generation / analysis
|
| 355 |
if generate_now and not st.session_state.get("busy"):
|
| 356 |
if not st.session_state.get("videos"):
|
| 357 |
st.error("No video loaded. Use 'Load Video' in the sidebar.")
|
| 358 |
else:
|
| 359 |
+
key = st.session_state.get("api_key") or os.getenv("GOOGLE_API_KEY")
|
| 360 |
+
if not key:
|
| 361 |
st.error("Google API key not set.")
|
| 362 |
else:
|
| 363 |
+
st.session_state["busy"] = True
|
| 364 |
try:
|
| 365 |
+
genai.configure(api_key=key)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 366 |
model_id = (st.session_state.get("model_input") or model_selected or DEFAULT_MODEL).strip()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 367 |
prompt_text = (analysis_prompt.strip() or DEFAULT_PROMPT).strip()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 368 |
|
| 369 |
+
with st.spinner("Generating analysis (inline video)…"):
|
| 370 |
+
out = generate_with_inline_video(
|
| 371 |
+
st.session_state["videos"], prompt_text, model_id,
|
| 372 |
+
timeout=st.session_state.get("generation_timeout", 300)
|
| 373 |
+
)
|
| 374 |
+
|
| 375 |
+
out = remove_prompt_echo(prompt_text, out)
|
| 376 |
st.session_state["analysis_out"] = out
|
| 377 |
st.session_state["last_error"] = ""
|
| 378 |
st.subheader("Analysis Result")
|
| 379 |
+
st.markdown(out or "No analysis returned.")
|
|
|
|
|
|
|
| 380 |
except Exception as e:
|
| 381 |
+
st.session_state["last_error"] = f"Inline generation error: {e}\n{traceback.format_exc()}"
|
| 382 |
+
st.error("An error occurred while generating the story using inline video. Check the error log.")
|
| 383 |
finally:
|
| 384 |
st.session_state["busy"] = False
|