Spaces:
Sleeping
Sleeping
CB commited on
Update streamlit_app.py
Browse files- streamlit_app.py +116 -506
streamlit_app.py
CHANGED
|
@@ -1,549 +1,159 @@
|
|
| 1 |
-
|
| 2 |
-
import
|
| 3 |
-
import
|
| 4 |
-
from glob import glob
|
| 5 |
-
from pathlib import Path
|
| 6 |
-
from tempfile import NamedTemporaryFile
|
| 7 |
-
|
| 8 |
-
import yt_dlp
|
| 9 |
-
import ffmpeg
|
| 10 |
-
import streamlit as st
|
| 11 |
-
from dotenv import load_dotenv
|
| 12 |
-
|
| 13 |
-
load_dotenv()
|
| 14 |
-
|
| 15 |
-
st.set_page_config(page_title="Generate the story of videos", layout="wide")
|
| 16 |
-
DATA_DIR = Path("./data")
|
| 17 |
-
DATA_DIR.mkdir(exist_ok=True)
|
| 18 |
-
|
| 19 |
-
for k, v in {
|
| 20 |
-
"videos": "",
|
| 21 |
-
"loop_video": False,
|
| 22 |
-
"uploaded_file": None,
|
| 23 |
-
"processed_file": None,
|
| 24 |
-
"busy": False,
|
| 25 |
-
"last_loaded_path": "",
|
| 26 |
-
"analysis_out": "",
|
| 27 |
-
"last_error": "",
|
| 28 |
-
"file_hash": None,
|
| 29 |
-
"fast_mode": False,
|
| 30 |
-
"use_compression": True,
|
| 31 |
-
}.items():
|
| 32 |
-
st.session_state.setdefault(k, v)
|
| 33 |
-
|
| 34 |
-
def sanitize_filename(path_str: str):
|
| 35 |
-
return Path(path_str).name.lower().translate(str.maketrans("", "", "!?\"'`~@#$%^&*()[]{}<>:,;\\/|+=*")).replace(" ", "_")
|
| 36 |
-
|
| 37 |
-
def file_sha256(path: str, block_size: int = 65536) -> str:
|
| 38 |
-
h = hashlib.sha256()
|
| 39 |
-
with open(path, "rb") as f:
|
| 40 |
-
for chunk in iter(lambda: f.read(block_size), b""):
|
| 41 |
-
h.update(chunk)
|
| 42 |
-
return h.hexdigest()
|
| 43 |
-
|
| 44 |
-
def safe_ffmpeg_run(stream_cmd):
|
| 45 |
-
try:
|
| 46 |
-
stream_cmd.run(overwrite_output=True, capture_stdout=True, capture_stderr=True)
|
| 47 |
-
return True, ""
|
| 48 |
-
except ffmpeg.Error as e:
|
| 49 |
-
try:
|
| 50 |
-
return False, e.stderr.decode("utf-8", errors="ignore")
|
| 51 |
-
except Exception:
|
| 52 |
-
return False, str(e)
|
| 53 |
-
|
| 54 |
-
def convert_video_to_mp4(video_path: str) -> str:
|
| 55 |
-
target = Path(video_path).with_suffix(".mp4")
|
| 56 |
-
if target.exists():
|
| 57 |
-
return str(target)
|
| 58 |
-
tmp = NamedTemporaryFile(prefix=target.stem + "_", suffix=".mp4", delete=False, dir=target.parent)
|
| 59 |
-
tmp.close()
|
| 60 |
-
ok, err = safe_ffmpeg_run(ffmpeg.input(video_path).output(str(tmp.name)))
|
| 61 |
-
if not ok:
|
| 62 |
-
try:
|
| 63 |
-
os.remove(tmp.name)
|
| 64 |
-
except Exception:
|
| 65 |
-
pass
|
| 66 |
-
raise RuntimeError(f"ffmpeg conversion failed: {err}")
|
| 67 |
-
os.replace(tmp.name, str(target))
|
| 68 |
-
if Path(video_path).suffix.lower() != ".mp4":
|
| 69 |
-
try:
|
| 70 |
-
os.remove(video_path)
|
| 71 |
-
except Exception:
|
| 72 |
-
pass
|
| 73 |
-
return str(target)
|
| 74 |
-
|
| 75 |
-
def compress_video(input_path: str, target_path: str, crf: int = 28, preset: str = "fast"):
|
| 76 |
-
tmp = NamedTemporaryFile(prefix=Path(target_path).stem + "_", suffix=".mp4", delete=False, dir=Path(target_path).parent)
|
| 77 |
-
tmp.close()
|
| 78 |
-
ok, err = safe_ffmpeg_run(ffmpeg.input(input_path).output(str(tmp.name), vcodec="libx264", crf=crf, preset=preset))
|
| 79 |
-
if not ok:
|
| 80 |
-
try:
|
| 81 |
-
os.remove(tmp.name)
|
| 82 |
-
except Exception:
|
| 83 |
-
pass
|
| 84 |
-
return input_path
|
| 85 |
-
os.replace(tmp.name, target_path)
|
| 86 |
-
return target_path
|
| 87 |
-
|
| 88 |
-
def download_video_ytdlp(url: str, save_dir: str, video_password: str = None) -> str:
|
| 89 |
-
if not url:
|
| 90 |
-
raise ValueError("No URL provided")
|
| 91 |
-
outtmpl = str(Path(save_dir) / "%(id)s.%(ext)s")
|
| 92 |
-
opts = {"outtmpl": outtmpl, "format": "best"}
|
| 93 |
-
if video_password:
|
| 94 |
-
opts["videopassword"] = video_password
|
| 95 |
-
with yt_dlp.YoutubeDL(opts) as ydl:
|
| 96 |
-
info = ydl.extract_info(url, download=True)
|
| 97 |
-
video_id = info.get("id") if isinstance(info, dict) else None
|
| 98 |
-
if video_id:
|
| 99 |
-
matches = glob(os.path.join(save_dir, f"{video_id}.*"))
|
| 100 |
-
else:
|
| 101 |
-
matches = sorted(glob(os.path.join(save_dir, "*")), key=os.path.getmtime, reverse=True)[:1]
|
| 102 |
-
if not matches:
|
| 103 |
-
raise FileNotFoundError("Downloaded video not found")
|
| 104 |
-
return convert_video_to_mp4(matches[0])
|
| 105 |
|
| 106 |
-
def
|
| 107 |
-
if not
|
| 108 |
-
return
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
return tail if tail.startswith("files/") else f"files/{tail}"
|
| 126 |
return None
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
| 133 |
-
|
| 134 |
-
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
|
| 143 |
-
|
| 144 |
-
|
| 145 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 146 |
return None
|
| 147 |
|
| 148 |
-
HAS_GENAI = False
|
| 149 |
-
genai = None
|
| 150 |
-
upload_file = None
|
| 151 |
-
get_file = None
|
| 152 |
-
delete_file = None
|
| 153 |
-
if os.getenv("GOOGLE_API_KEY"):
|
| 154 |
-
try:
|
| 155 |
-
import google.generativeai as genai_mod
|
| 156 |
-
genai = genai_mod
|
| 157 |
-
upload_file = getattr(genai_mod, "upload_file", None)
|
| 158 |
-
get_file = getattr(genai_mod, "get_file", None)
|
| 159 |
-
delete_file = getattr(genai_mod, "delete_file", None)
|
| 160 |
-
HAS_GENAI = True
|
| 161 |
-
except Exception:
|
| 162 |
-
HAS_GENAI = False
|
| 163 |
-
|
| 164 |
-
def upload_video_sdk(filepath: str):
|
| 165 |
-
key = get_runtime_api_key()
|
| 166 |
-
if not key:
|
| 167 |
-
raise RuntimeError("No API key")
|
| 168 |
-
if not HAS_GENAI or upload_file is None:
|
| 169 |
-
raise RuntimeError("google.generativeai SDK upload not available")
|
| 170 |
-
genai.configure(api_key=key)
|
| 171 |
-
return upload_file(filepath)
|
| 172 |
-
|
| 173 |
-
def wait_for_processed(file_obj, timeout=600):
|
| 174 |
-
if not HAS_GENAI or get_file is None:
|
| 175 |
-
return file_obj
|
| 176 |
-
start = time.time()
|
| 177 |
-
name = file_name_or_id(file_obj)
|
| 178 |
-
if not name:
|
| 179 |
-
return file_obj
|
| 180 |
-
backoff = 1.0
|
| 181 |
-
while True:
|
| 182 |
-
try:
|
| 183 |
-
obj = get_file(name)
|
| 184 |
-
except Exception:
|
| 185 |
-
obj = file_obj
|
| 186 |
-
state = getattr(obj, "state", None)
|
| 187 |
-
if not state or getattr(state, "name", None) != "PROCESSING":
|
| 188 |
-
return obj
|
| 189 |
-
if time.time() - start > timeout:
|
| 190 |
-
raise TimeoutError("File processing timed out")
|
| 191 |
-
time.sleep(backoff)
|
| 192 |
-
backoff = min(backoff * 2, 8.0)
|
| 193 |
-
|
| 194 |
-
def remove_prompt_echo(prompt: str, text: str):
|
| 195 |
-
if not prompt or not text:
|
| 196 |
-
return text
|
| 197 |
-
p = " ".join(prompt.strip().lower().split())
|
| 198 |
-
t = text.strip()
|
| 199 |
-
from difflib import SequenceMatcher
|
| 200 |
-
first = " ".join(t[:600].lower().split())
|
| 201 |
-
if SequenceMatcher(None, p, first).ratio() > 0.7:
|
| 202 |
-
cut = min(len(t), max(int(len(prompt) * 0.9), len(p)))
|
| 203 |
-
new = t[cut:].lstrip(" \n:-")
|
| 204 |
-
if len(new) >= 3:
|
| 205 |
-
return new
|
| 206 |
-
placeholders = ["enter analysis", "enter your analysis", "enter analysis here", "please enter analysis"]
|
| 207 |
-
low = t.lower()
|
| 208 |
-
for ph in placeholders:
|
| 209 |
-
if low.startswith(ph):
|
| 210 |
-
return t[len(ph):].lstrip(" \n:-")
|
| 211 |
-
return text
|
| 212 |
-
|
| 213 |
-
st.sidebar.header("Video Input")
|
| 214 |
-
st.sidebar.text_input("Video URL", key="url", placeholder="https://")
|
| 215 |
-
settings = st.sidebar.expander("Settings", expanded=False)
|
| 216 |
-
|
| 217 |
-
env_key = os.getenv("GOOGLE_API_KEY", "")
|
| 218 |
-
API_KEY_INPUT = settings.text_input("Google API Key (one-time)", value="", type="password")
|
| 219 |
-
model_input = settings.text_input("Gemini Model (short name)", "gemini-2.0-flash-lite")
|
| 220 |
-
model_id = model_input.strip() or "gemini-2.0-flash-lite"
|
| 221 |
-
model_arg = model_id if not model_id.startswith("models/") else model_id.split("/", 1)[1]
|
| 222 |
-
|
| 223 |
-
default_prompt = (
|
| 224 |
-
"You are an Indoor Human Behavior Analyzer. Watch the video and produce a detailed, evidence‑based behavioral report focused on human actions, "
|
| 225 |
-
"interactions, posture, movement, anatomy, and apparent intent. Use vivid, anatomically rich language and avoid moralizing. Prefer short paragraphs and numeric estimates "
|
| 226 |
-
"for anatomical measurements. Provide sensory, subjective descriptions and vivid imagery, including a concise summary of observed actions and a description of behaviors "
|
| 227 |
-
"and interaction dynamics. Use the following personality‑traits list when inferring dispositions: driven by an insatiable desire to understand human behavior and anatomy. "
|
| 228 |
-
"Finish with a short feedback and recommendations section. Adopt a playful, anatomically obsessed, slightly mischievous persona — inquisitive, pragmatic, and vivid in description."
|
| 229 |
-
)
|
| 230 |
-
|
| 231 |
-
analysis_prompt = settings.text_area("Enter analysis", value=default_prompt, height=300)
|
| 232 |
-
settings.text_input("Video Password (if needed)", key="video-password", type="password")
|
| 233 |
-
settings.checkbox("Fast mode (skip compression, smaller model, fewer tokens)", key="fast_mode")
|
| 234 |
-
settings.checkbox("Enable compression for large files (>50MB)", value=True, key="use_compression")
|
| 235 |
-
settings.number_input("Max output tokens", key="max_output_tokens", value=1024, min_value=128, max_value=8192, step=128)
|
| 236 |
-
|
| 237 |
-
if not API_KEY_INPUT and not env_key:
|
| 238 |
-
settings.info("No Google API key provided; upload/generation disabled.", icon="ℹ️")
|
| 239 |
-
|
| 240 |
-
if st.sidebar.button("Load Video", use_container_width=True):
|
| 241 |
-
try:
|
| 242 |
-
vpw = st.session_state.get("video-password", "")
|
| 243 |
-
path = download_video_ytdlp(st.session_state.get("url", ""), str(DATA_DIR), vpw)
|
| 244 |
-
st.session_state["videos"] = path
|
| 245 |
-
st.session_state["last_loaded_path"] = path
|
| 246 |
-
st.session_state["uploaded_file"] = None
|
| 247 |
-
st.session_state["processed_file"] = None
|
| 248 |
-
st.session_state["file_hash"] = file_sha256(path)
|
| 249 |
-
except Exception as e:
|
| 250 |
-
st.sidebar.error(f"Failed to load video: {e}")
|
| 251 |
-
|
| 252 |
-
if st.session_state["videos"]:
|
| 253 |
-
try:
|
| 254 |
-
st.sidebar.video(st.session_state["videos"], loop=st.session_state.get("loop_video", False))
|
| 255 |
-
except Exception:
|
| 256 |
-
st.sidebar.write("Couldn't preview video")
|
| 257 |
-
with st.sidebar.expander("Options", expanded=False):
|
| 258 |
-
loop_checkbox = st.checkbox("Enable Loop", value=st.session_state.get("loop_video", False))
|
| 259 |
-
st.session_state["loop_video"] = loop_checkbox
|
| 260 |
-
|
| 261 |
-
if st.button("Clear Video(s)"):
|
| 262 |
-
for f in glob(str(DATA_DIR / "*")):
|
| 263 |
-
try:
|
| 264 |
-
os.remove(f)
|
| 265 |
-
except Exception:
|
| 266 |
-
pass
|
| 267 |
-
for k in ("uploaded_file", "processed_file"):
|
| 268 |
-
st.session_state.pop(k, None)
|
| 269 |
-
st.session_state["videos"] = ""
|
| 270 |
-
st.session_state["last_loaded_path"] = ""
|
| 271 |
-
st.session_state["analysis_out"] = ""
|
| 272 |
-
st.session_state["last_error"] = ""
|
| 273 |
-
st.session_state["file_hash"] = None
|
| 274 |
-
|
| 275 |
-
try:
|
| 276 |
-
with open(st.session_state["videos"], "rb") as vf:
|
| 277 |
-
st.download_button("Download Video", data=vf, file_name=sanitize_filename(st.session_state["videos"]), mime="video/mp4", use_container_width=True)
|
| 278 |
-
except Exception:
|
| 279 |
-
pass
|
| 280 |
-
st.sidebar.write("Title:", Path(st.session_state["videos"]).name)
|
| 281 |
-
|
| 282 |
-
col1, col2 = st.columns([1, 3])
|
| 283 |
-
with col1:
|
| 284 |
-
if st.session_state.get("busy"):
|
| 285 |
-
st.write("Generation in progress...")
|
| 286 |
-
if st.button("Cancel"):
|
| 287 |
-
st.session_state["busy"] = False
|
| 288 |
-
st.session_state["last_error"] = "Generation cancelled by user."
|
| 289 |
-
else:
|
| 290 |
-
generate_now = st.button("Generate the story", type="primary")
|
| 291 |
-
with col2:
|
| 292 |
-
pass
|
| 293 |
-
|
| 294 |
-
def get_runtime_api_key():
|
| 295 |
-
key = API_KEY_INPUT.strip() if API_KEY_INPUT else ""
|
| 296 |
-
if key:
|
| 297 |
-
return key
|
| 298 |
-
return os.getenv("GOOGLE_API_KEY", "").strip() or None
|
| 299 |
-
|
| 300 |
-
# responses caller: prefer SDK responses, fallback to generativelanguage generate endpoints
|
| 301 |
-
import json
|
| 302 |
-
import requests
|
| 303 |
-
|
| 304 |
def responses_generate(model, messages, files, max_output_tokens, api_key):
|
| 305 |
if not api_key:
|
| 306 |
raise RuntimeError("No API key for responses_generate")
|
| 307 |
sdk_err = None
|
|
|
|
|
|
|
| 308 |
if HAS_GENAI and genai is not None:
|
| 309 |
try:
|
| 310 |
genai.configure(api_key=api_key)
|
| 311 |
-
|
| 312 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 313 |
except Exception as e:
|
| 314 |
sdk_err = str(e)
|
| 315 |
|
|
|
|
| 316 |
host = "https://generativelanguage.googleapis.com"
|
|
|
|
| 317 |
candidates = [
|
| 318 |
-
f"{host}/v1/models/{
|
| 319 |
-
f"{host}/v1beta3/models/{
|
| 320 |
-
f"{host}/v1beta2/models/{
|
| 321 |
]
|
| 322 |
-
|
| 323 |
-
|
|
|
|
| 324 |
headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
|
| 325 |
last_exc = None
|
|
|
|
| 326 |
for url in candidates:
|
| 327 |
try:
|
| 328 |
-
r = requests.post(url, json=payload, headers=headers, timeout=
|
| 329 |
if r.status_code == 200:
|
| 330 |
try:
|
| 331 |
return r.json()
|
| 332 |
except Exception:
|
| 333 |
return {"text": r.text}
|
|
|
|
| 334 |
last_exc = RuntimeError(f"HTTP {r.status_code}: {r.text}")
|
| 335 |
except Exception as e:
|
| 336 |
last_exc = e
|
|
|
|
| 337 |
diag = {"sdk_error": sdk_err, "http_error": str(last_exc), "tried_urls": candidates}
|
| 338 |
raise RuntimeError(f"genai.responses not available and HTTP fallback failed: {diag}")
|
| 339 |
|
| 340 |
def call_responses_once(model_used, system_msg, user_msg, fname, max_tokens):
|
|
|
|
|
|
|
| 341 |
files = [{"name": fname}] if fname else None
|
| 342 |
for attempt in range(2):
|
| 343 |
try:
|
| 344 |
-
return responses_generate(model_used,
|
| 345 |
except Exception:
|
| 346 |
if attempt == 0:
|
| 347 |
time.sleep(1.0)
|
| 348 |
continue
|
| 349 |
raise
|
| 350 |
|
| 351 |
-
|
| 352 |
-
|
| 353 |
-
|
| 354 |
-
|
| 355 |
-
|
| 356 |
-
|
| 357 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 358 |
else:
|
| 359 |
-
|
| 360 |
-
|
| 361 |
-
|
| 362 |
-
|
| 363 |
-
|
| 364 |
-
|
| 365 |
-
|
| 366 |
-
|
| 367 |
-
|
| 368 |
-
|
| 369 |
-
|
| 370 |
-
|
| 371 |
-
|
| 372 |
-
|
| 373 |
-
|
| 374 |
-
|
| 375 |
-
|
| 376 |
-
|
| 377 |
-
|
| 378 |
-
file_size_mb = os.path.getsize(local_path) / (1024 * 1024)
|
| 379 |
-
except Exception:
|
| 380 |
-
file_size_mb = 0
|
| 381 |
-
|
| 382 |
-
use_compression = st.session_state.get("use_compression", True)
|
| 383 |
-
if use_compression and not fast_mode and file_size_mb > 50:
|
| 384 |
-
compressed_path = str(Path(local_path).with_name(Path(local_path).stem + "_compressed.mp4"))
|
| 385 |
-
try:
|
| 386 |
-
preset = "veryfast" if fast_mode else "fast"
|
| 387 |
-
upload_path = compress_video(local_path, compressed_path, crf=28, preset=preset)
|
| 388 |
-
except Exception:
|
| 389 |
-
upload_path = local_path
|
| 390 |
-
|
| 391 |
-
if HAS_GENAI and upload_file is not None:
|
| 392 |
-
genai.configure(api_key=runtime_key)
|
| 393 |
-
with st.spinner("Uploading video..."):
|
| 394 |
-
uploaded = upload_video_sdk(upload_path)
|
| 395 |
-
processed = wait_for_processed(uploaded, timeout=600)
|
| 396 |
-
st.session_state["uploaded_file"] = uploaded
|
| 397 |
-
st.session_state["processed_file"] = processed
|
| 398 |
-
st.session_state["last_loaded_path"] = current_path
|
| 399 |
-
st.session_state["file_hash"] = current_hash
|
| 400 |
-
else:
|
| 401 |
-
uploaded = None
|
| 402 |
-
processed = None
|
| 403 |
-
st.session_state["uploaded_file"] = None
|
| 404 |
-
st.session_state["processed_file"] = None
|
| 405 |
-
else:
|
| 406 |
-
uploaded = st.session_state.get("uploaded_file")
|
| 407 |
-
processed = st.session_state.get("processed_file")
|
| 408 |
-
|
| 409 |
-
prompt_text = (analysis_prompt or default_prompt).strip()
|
| 410 |
-
if st.session_state.get("fast_mode"):
|
| 411 |
-
model_used = model_arg or "gemini-2.0-flash-lite"
|
| 412 |
-
max_tokens = min(st.session_state.get("max_output_tokens", 512), 1024)
|
| 413 |
-
else:
|
| 414 |
-
model_used = model_arg
|
| 415 |
-
max_tokens = st.session_state.get("max_output_tokens", 1024)
|
| 416 |
-
|
| 417 |
-
system_msg = {"role": "system", "content": "You are a helpful assistant that summarizes videos concisely in vivid detail."}
|
| 418 |
-
user_msg = {"role": "user", "content": prompt_text}
|
| 419 |
-
|
| 420 |
-
fname = file_name_or_id(processed) or file_name_or_id(uploaded)
|
| 421 |
-
response = call_responses_once(model_used, system_msg, user_msg, fname, max_tokens)
|
| 422 |
-
|
| 423 |
-
def extract_text_from_response(response):
|
| 424 |
-
outputs = getattr(response, "output", None) or (response.get("output") if isinstance(response, dict) else None) or []
|
| 425 |
-
if isinstance(outputs, dict):
|
| 426 |
-
outputs = outputs.get("contents") or outputs.get("items") or []
|
| 427 |
-
text_pieces = []
|
| 428 |
-
for item in outputs or []:
|
| 429 |
-
contents = getattr(item, "content", None) or (item.get("content") if isinstance(item, dict) else None) or []
|
| 430 |
-
if isinstance(contents, dict):
|
| 431 |
-
contents = [contents]
|
| 432 |
-
for c in contents:
|
| 433 |
-
ctype = getattr(c, "type", None) or (c.get("type") if isinstance(c, dict) else None)
|
| 434 |
-
if ctype in ("output_text", "text") or ctype is None:
|
| 435 |
-
txt = getattr(c, "text", None) or (c.get("text") if isinstance(c, dict) else None)
|
| 436 |
-
if txt:
|
| 437 |
-
text_pieces.append(txt)
|
| 438 |
-
if not text_pieces:
|
| 439 |
-
top_text = getattr(response, "text", None) or (response.get("text") if isinstance(response, dict) else None)
|
| 440 |
-
if top_text:
|
| 441 |
-
text_pieces.append(top_text)
|
| 442 |
-
seen = set()
|
| 443 |
-
filtered = []
|
| 444 |
-
for t in text_pieces:
|
| 445 |
-
if t not in seen:
|
| 446 |
-
filtered.append(t)
|
| 447 |
-
seen.add(t)
|
| 448 |
-
return "\n\n".join(filtered)
|
| 449 |
-
|
| 450 |
-
out = extract_text_from_response(response)
|
| 451 |
-
|
| 452 |
-
meta = getattr(response, "metrics", None) or (response.get("metrics") if isinstance(response, dict) else None) or {}
|
| 453 |
-
output_tokens = 0
|
| 454 |
-
try:
|
| 455 |
-
if isinstance(meta, dict):
|
| 456 |
-
output_tokens = int(meta.get("output_tokens", 0) or 0)
|
| 457 |
-
else:
|
| 458 |
-
output_tokens = int(getattr(meta, "output_tokens", 0) or 0)
|
| 459 |
-
except Exception:
|
| 460 |
-
output_tokens = 0
|
| 461 |
-
|
| 462 |
-
if (not out or output_tokens == 0) and model_used:
|
| 463 |
-
retry_prompt = "Summarize the video content briefly and vividly (2-4 paragraphs)."
|
| 464 |
-
try:
|
| 465 |
-
response2 = call_responses_once(model_used, system_msg, {"role": "user", "content": retry_prompt}, fname, min(max_tokens * 2, 4096))
|
| 466 |
-
out2 = extract_text_from_response(response2)
|
| 467 |
-
if out2 and len(out2) > len(out or ""):
|
| 468 |
-
out = out2
|
| 469 |
-
else:
|
| 470 |
-
response3 = call_responses_once(model_used, system_msg, {"role": "user", "content": "List the main points of the video as 6-10 bullets."}, fname, min(1024, max_tokens * 2))
|
| 471 |
-
out3 = extract_text_from_response(response3)
|
| 472 |
-
if out3:
|
| 473 |
-
out = out3
|
| 474 |
-
except Exception:
|
| 475 |
-
pass
|
| 476 |
-
|
| 477 |
-
if out:
|
| 478 |
-
out = remove_prompt_echo(prompt_text, out).strip()
|
| 479 |
-
|
| 480 |
-
st.session_state["analysis_out"] = out or ""
|
| 481 |
-
st.session_state["last_error"] = ""
|
| 482 |
-
|
| 483 |
-
st.subheader("Analysis Result")
|
| 484 |
-
st.markdown(out or "_(no text returned)_")
|
| 485 |
-
|
| 486 |
-
try:
|
| 487 |
-
if reupload_needed:
|
| 488 |
-
if upload_path and Path(upload_path).exists() and Path(upload_path) != Path(current_path):
|
| 489 |
-
Path(upload_path).unlink(missing_ok=True)
|
| 490 |
-
Path(current_path).unlink(missing_ok=True)
|
| 491 |
-
st.session_state["videos"] = ""
|
| 492 |
-
except Exception:
|
| 493 |
-
pass
|
| 494 |
-
|
| 495 |
-
with st.expander("Debug (compact)", expanded=False):
|
| 496 |
-
try:
|
| 497 |
-
info = {
|
| 498 |
-
"model": model_used,
|
| 499 |
-
"output_tokens": output_tokens,
|
| 500 |
-
"upload_succeeded": bool(st.session_state.get("uploaded_file")),
|
| 501 |
-
"processed_state": getattr(st.session_state.get("processed_file"), "state", None) if st.session_state.get("processed_file") else None,
|
| 502 |
-
}
|
| 503 |
-
st.write(info)
|
| 504 |
-
try:
|
| 505 |
-
if isinstance(response, dict):
|
| 506 |
-
keys = list(response.keys())[:20]
|
| 507 |
-
else:
|
| 508 |
-
keys = [k for k in dir(response) if not k.startswith("_")][:20]
|
| 509 |
-
st.write({"response_keys_or_attrs": keys})
|
| 510 |
-
except Exception:
|
| 511 |
-
pass
|
| 512 |
-
except Exception:
|
| 513 |
-
st.write("Debug info unavailable")
|
| 514 |
-
|
| 515 |
-
except Exception as e:
|
| 516 |
-
st.session_state["last_error"] = str(e)
|
| 517 |
-
st.error(f"An error occurred while generating the story: {e}")
|
| 518 |
-
finally:
|
| 519 |
-
st.session_state["busy"] = False
|
| 520 |
-
|
| 521 |
-
if st.session_state.get("analysis_out"):
|
| 522 |
-
st.subheader("Analysis Result")
|
| 523 |
-
st.markdown(st.session_state.get("analysis_out"))
|
| 524 |
-
|
| 525 |
-
if st.session_state.get("last_error"):
|
| 526 |
-
with st.expander("Last Error", expanded=False):
|
| 527 |
-
st.write(st.session_state.get("last_error"))
|
| 528 |
-
|
| 529 |
-
with st.sidebar.expander("Manage uploads", expanded=False):
|
| 530 |
-
if st.button("Delete uploaded files (local + cloud)"):
|
| 531 |
-
for f in glob(str(DATA_DIR / "*")):
|
| 532 |
-
try:
|
| 533 |
-
Path(f).unlink(missing_ok=True)
|
| 534 |
-
except Exception:
|
| 535 |
-
pass
|
| 536 |
-
st.session_state["videos"] = ""
|
| 537 |
-
st.session_state["uploaded_file"] = None
|
| 538 |
-
st.session_state["processed_file"] = None
|
| 539 |
-
st.session_state["last_loaded_path"] = ""
|
| 540 |
-
st.session_state["analysis_out"] = ""
|
| 541 |
-
st.session_state["file_hash"] = None
|
| 542 |
-
try:
|
| 543 |
-
fname = file_name_or_id(st.session_state.get("uploaded_file"))
|
| 544 |
-
if fname and delete_file and HAS_GENAI:
|
| 545 |
-
genai.configure(api_key=get_runtime_api_key() or os.getenv("GOOGLE_API_KEY", ""))
|
| 546 |
-
delete_file(fname)
|
| 547 |
-
except Exception:
|
| 548 |
-
pass
|
| 549 |
-
st.success("Local files removed. Cloud deletion attempted where supported.")
|
|
|
|
| 1 |
+
# --- patched responses / generate compatibility layer ---
|
| 2 |
+
import json
|
| 3 |
+
import requests
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
|
| 5 |
+
def _normalize_model_for_url(model: str) -> str:
|
| 6 |
+
if not model:
|
| 7 |
+
return "gemini-2.0"
|
| 8 |
+
return model.split("/", 1)[-1] if model.startswith("models/") else model
|
| 9 |
+
|
| 10 |
+
def _build_prompt_from_messages(messages):
|
| 11 |
+
# messages expected as list of {"role":..., "content":...}
|
| 12 |
+
if not messages:
|
| 13 |
+
return ""
|
| 14 |
+
parts = []
|
| 15 |
+
for m in messages:
|
| 16 |
+
role = (m.get("role") if isinstance(m, dict) else getattr(m, "role", None)) or "user"
|
| 17 |
+
content = (m.get("content") if isinstance(m, dict) else getattr(m, "content", None)) or ""
|
| 18 |
+
parts.append(f"{role.upper()}:\n{content.strip()}\n")
|
| 19 |
+
return "\n".join(parts)
|
| 20 |
+
|
| 21 |
+
def _parse_http_generate_response(rjson):
|
| 22 |
+
# Attempt to extract text from various generate shapes
|
| 23 |
+
if not rjson:
|
|
|
|
| 24 |
return None
|
| 25 |
+
# common new GL formats: {'candidates':[{'content': '...'}]} or {'output': [{'content': ...}]}
|
| 26 |
+
if isinstance(rjson, dict):
|
| 27 |
+
# try 'candidates'
|
| 28 |
+
if "candidates" in rjson and isinstance(rjson["candidates"], list) and rjson["candidates"]:
|
| 29 |
+
cand = rjson["candidates"][0]
|
| 30 |
+
return cand.get("content") or cand.get("text") or rjson.get("text")
|
| 31 |
+
# try 'output' array with 'content' items
|
| 32 |
+
out = rjson.get("output")
|
| 33 |
+
if isinstance(out, list) and out:
|
| 34 |
+
texts = []
|
| 35 |
+
for item in out:
|
| 36 |
+
if isinstance(item, dict):
|
| 37 |
+
c = item.get("content") or item.get("contents") or item.get("text")
|
| 38 |
+
if isinstance(c, str):
|
| 39 |
+
texts.append(c)
|
| 40 |
+
elif isinstance(c, list):
|
| 41 |
+
for sub in c:
|
| 42 |
+
if isinstance(sub, dict):
|
| 43 |
+
t = sub.get("text") or sub.get("content")
|
| 44 |
+
if t:
|
| 45 |
+
texts.append(t)
|
| 46 |
+
if texts:
|
| 47 |
+
return "\n\n".join(texts)
|
| 48 |
+
# fallback to top-level text
|
| 49 |
+
if "text" in rjson and isinstance(rjson["text"], str):
|
| 50 |
+
return rjson["text"]
|
| 51 |
return None
|
| 52 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
def responses_generate(model, messages, files, max_output_tokens, api_key):
|
| 54 |
if not api_key:
|
| 55 |
raise RuntimeError("No API key for responses_generate")
|
| 56 |
sdk_err = None
|
| 57 |
+
|
| 58 |
+
# try SDK responses.generate (preferred)
|
| 59 |
if HAS_GENAI and genai is not None:
|
| 60 |
try:
|
| 61 |
genai.configure(api_key=api_key)
|
| 62 |
+
responses_obj = getattr(genai, "responses", None)
|
| 63 |
+
if responses_obj is not None and hasattr(responses_obj, "generate"):
|
| 64 |
+
# SDK expects messages and files in their SDK-specific shapes
|
| 65 |
+
sdk_kwargs = {"model": model, "messages": messages, "max_output_tokens": int(max_output_tokens or 512)}
|
| 66 |
+
if files:
|
| 67 |
+
sdk_kwargs["files"] = files
|
| 68 |
+
return responses_obj.generate(**sdk_kwargs)
|
| 69 |
except Exception as e:
|
| 70 |
sdk_err = str(e)
|
| 71 |
|
| 72 |
+
# HTTP fallback to Generative Language "generate" endpoints.
|
| 73 |
host = "https://generativelanguage.googleapis.com"
|
| 74 |
+
norm_model = _normalize_model_for_url(model)
|
| 75 |
candidates = [
|
| 76 |
+
f"{host}/v1/models/{norm_model}:generate",
|
| 77 |
+
f"{host}/v1beta3/models/{norm_model}:generate",
|
| 78 |
+
f"{host}/v1beta2/models/{norm_model}:generate",
|
| 79 |
]
|
| 80 |
+
|
| 81 |
+
prompt_text = _build_prompt_from_messages(messages)
|
| 82 |
+
payload = {"prompt": {"text": prompt_text}, "maxOutputTokens": int(max_output_tokens or 512)}
|
| 83 |
headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
|
| 84 |
last_exc = None
|
| 85 |
+
|
| 86 |
for url in candidates:
|
| 87 |
try:
|
| 88 |
+
r = requests.post(url, json=payload, headers=headers, timeout=15)
|
| 89 |
if r.status_code == 200:
|
| 90 |
try:
|
| 91 |
return r.json()
|
| 92 |
except Exception:
|
| 93 |
return {"text": r.text}
|
| 94 |
+
# if 404, try next; collect last
|
| 95 |
last_exc = RuntimeError(f"HTTP {r.status_code}: {r.text}")
|
| 96 |
except Exception as e:
|
| 97 |
last_exc = e
|
| 98 |
+
|
| 99 |
diag = {"sdk_error": sdk_err, "http_error": str(last_exc), "tried_urls": candidates}
|
| 100 |
raise RuntimeError(f"genai.responses not available and HTTP fallback failed: {diag}")
|
| 101 |
|
| 102 |
def call_responses_once(model_used, system_msg, user_msg, fname, max_tokens):
|
| 103 |
+
# messages as [system_msg, user_msg]
|
| 104 |
+
messages = [system_msg, user_msg]
|
| 105 |
files = [{"name": fname}] if fname else None
|
| 106 |
for attempt in range(2):
|
| 107 |
try:
|
| 108 |
+
return responses_generate(model_used, messages, files, max_tokens, api_key=get_runtime_api_key())
|
| 109 |
except Exception:
|
| 110 |
if attempt == 0:
|
| 111 |
time.sleep(1.0)
|
| 112 |
continue
|
| 113 |
raise
|
| 114 |
|
| 115 |
+
# Helper to extract text from either SDK response object or HTTP dict
|
| 116 |
+
def extract_text_from_response(response):
|
| 117 |
+
# SDK may return an object with .output, .candidates, or .text
|
| 118 |
+
# HTTP returns a dict with various shapes
|
| 119 |
+
# If it's an object (not dict), try attribute access
|
| 120 |
+
try:
|
| 121 |
+
if response is None:
|
| 122 |
+
return None
|
| 123 |
+
if isinstance(response, dict):
|
| 124 |
+
# HTTP-style
|
| 125 |
+
text = _parse_http_generate_response(response)
|
| 126 |
+
if text:
|
| 127 |
+
return text
|
| 128 |
+
# try 'output' field shaped differently
|
| 129 |
+
outputs = response.get("output") or response.get("candidates")
|
| 130 |
+
if outputs:
|
| 131 |
+
pieces = []
|
| 132 |
+
for o in outputs:
|
| 133 |
+
if isinstance(o, dict):
|
| 134 |
+
t = o.get("content") or o.get("text")
|
| 135 |
+
if isinstance(t, str):
|
| 136 |
+
pieces.append(t)
|
| 137 |
+
if pieces:
|
| 138 |
+
return "\n\n".join(pieces)
|
| 139 |
+
return response.get("text") or None
|
| 140 |
else:
|
| 141 |
+
# object-like SDK response
|
| 142 |
+
outputs = getattr(response, "output", None) or getattr(response, "candidates", None) or None
|
| 143 |
+
if outputs:
|
| 144 |
+
pieces = []
|
| 145 |
+
for item in outputs:
|
| 146 |
+
# each item may have 'content' or 'text'
|
| 147 |
+
txt = getattr(item, "content", None) or getattr(item, "text", None) or (item.get("content") if isinstance(item, dict) else None)
|
| 148 |
+
if txt:
|
| 149 |
+
pieces.append(txt)
|
| 150 |
+
if pieces:
|
| 151 |
+
return "\n\n".join(pieces)
|
| 152 |
+
# try top-level text
|
| 153 |
+
txt = getattr(response, "text", None)
|
| 154 |
+
if txt:
|
| 155 |
+
return txt
|
| 156 |
+
except Exception:
|
| 157 |
+
pass
|
| 158 |
+
return None
|
| 159 |
+
# --- end patched section ---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|