Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
Update backup13.app.py
Browse files- backup13.app.py +233 -223
backup13.app.py
CHANGED
@@ -37,8 +37,8 @@ load_dotenv()
|
|
37 |
|
38 |
# Add available English voices for Edge TTS
|
39 |
EDGE_TTS_VOICES = [
|
40 |
-
"en-US-AriaNeural",
|
41 |
-
"en-US-GuyNeural",
|
42 |
"en-US-JennyNeural",
|
43 |
"en-GB-SoniaNeural",
|
44 |
"en-GB-RyanNeural",
|
@@ -49,6 +49,16 @@ EDGE_TTS_VOICES = [
|
|
49 |
]
|
50 |
|
51 |
# Initialize session state variables
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
52 |
if 'tts_voice' not in st.session_state:
|
53 |
st.session_state['tts_voice'] = EDGE_TTS_VOICES[0]
|
54 |
if 'audio_format' not in st.session_state:
|
@@ -102,26 +112,55 @@ FILE_EMOJIS = {
|
|
102 |
"wav": "🔊"
|
103 |
}
|
104 |
|
105 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
106 |
def get_marquee_settings():
|
107 |
-
"""Get
|
|
|
|
|
|
|
|
|
|
|
|
|
108 |
st.sidebar.markdown("### 🎯 Marquee Settings")
|
109 |
cols = st.sidebar.columns(2)
|
110 |
with cols[0]:
|
111 |
-
bg_color = st.color_picker("🎨 Background",
|
112 |
-
|
|
|
|
|
|
|
|
|
113 |
with cols[1]:
|
114 |
font_size = st.slider("📏 Size", 10, 24, 14, key="font_size_slider")
|
115 |
duration = st.slider("⏱️ Speed", 1, 20, 10, key="duration_slider")
|
116 |
-
|
117 |
-
|
118 |
"background": bg_color,
|
119 |
"color": text_color,
|
120 |
"font-size": f"{font_size}px",
|
121 |
-
"animationDuration": f"{duration}s"
|
122 |
-
|
123 |
-
"lineHeight": "35px"
|
124 |
-
}
|
125 |
|
126 |
def display_marquee(text, settings, key_suffix=""):
|
127 |
"""Display marquee with given text and settings"""
|
@@ -133,13 +172,6 @@ def display_marquee(text, settings, key_suffix=""):
|
|
133 |
)
|
134 |
st.write("")
|
135 |
|
136 |
-
def process_paper_content(paper):
|
137 |
-
"""Process paper content for marquee and audio"""
|
138 |
-
marquee_text = f"📄 {paper['title']} | 👤 {paper['authors'][:100]} | 📝 {paper['summary'][:100]}"
|
139 |
-
audio_text = f"{paper['title']} by {paper['authors']}. {paper['summary']}"
|
140 |
-
return marquee_text, audio_text
|
141 |
-
|
142 |
-
# Text Processing Functions
|
143 |
def get_high_info_terms(text: str, top_n=10) -> list:
|
144 |
stop_words = set(['the', 'a', 'an', 'and', 'or', 'but', 'in', 'on', 'at', 'to', 'for', 'of', 'with'])
|
145 |
words = re.findall(r'\b\w+(?:-\w+)*\b', text.lower())
|
@@ -157,17 +189,8 @@ def clean_text_for_filename(text: str) -> str:
|
|
157 |
filtered = [w for w in words if len(w) > 3 and w not in stop_short]
|
158 |
return '_'.join(filtered)[:200]
|
159 |
|
160 |
-
def clean_for_speech(text: str) -> str:
|
161 |
-
text = text.replace("\n", " ")
|
162 |
-
text = text.replace("</s>", " ")
|
163 |
-
text = text.replace("#", "")
|
164 |
-
text = re.sub(r"\(https?:\/\/[^\)]+\)", "", text)
|
165 |
-
text = re.sub(r"\s+", " ", text).strip()
|
166 |
-
return text
|
167 |
-
|
168 |
-
# File Operations
|
169 |
def generate_filename(prompt, response, file_type="md"):
|
170 |
-
prefix =
|
171 |
combined = (prompt + " " + response).strip()
|
172 |
info_terms = get_high_info_terms(combined, top_n=10)
|
173 |
snippet = (prompt[:100] + " " + response[:100]).strip()
|
@@ -198,7 +221,14 @@ def get_download_link(file, file_type="zip"):
|
|
198 |
else:
|
199 |
return f'<a href="data:application/octet-stream;base64,{b64}" download="{os.path.basename(file)}">Download {os.path.basename(file)}</a>'
|
200 |
|
201 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
202 |
async def edge_tts_generate_audio(text, voice="en-US-AriaNeural", rate=0, pitch=0, file_format="mp3"):
|
203 |
text = clean_for_speech(text)
|
204 |
if not text.strip():
|
@@ -219,7 +249,74 @@ def play_and_download_audio(file_path, file_type="mp3"):
|
|
219 |
dl_link = get_download_link(file_path, file_type=file_type)
|
220 |
st.markdown(dl_link, unsafe_allow_html=True)
|
221 |
|
222 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
223 |
def parse_arxiv_refs(ref_text: str):
|
224 |
if not ref_text:
|
225 |
return []
|
@@ -269,58 +366,10 @@ def parse_arxiv_refs(ref_text: str):
|
|
269 |
|
270 |
return results[:20]
|
271 |
|
272 |
-
def create_paper_audio_files(papers, input_question):
|
273 |
-
for paper in papers:
|
274 |
-
try:
|
275 |
-
marquee_text, audio_text = process_paper_content(paper)
|
276 |
-
|
277 |
-
audio_text = clean_for_speech(audio_text)
|
278 |
-
file_format = st.session_state['audio_format']
|
279 |
-
audio_file = speak_with_edge_tts(audio_text,
|
280 |
-
voice=st.session_state['tts_voice'],
|
281 |
-
file_format=file_format)
|
282 |
-
paper['full_audio'] = audio_file
|
283 |
-
|
284 |
-
st.write(f"### {FILE_EMOJIS.get(file_format, '')} {os.path.basename(audio_file)}")
|
285 |
-
play_and_download_audio(audio_file, file_type=file_format)
|
286 |
-
paper['marquee_text'] = marquee_text
|
287 |
-
|
288 |
-
except Exception as e:
|
289 |
-
st.warning(f"Error processing paper {paper['title']}: {str(e)}")
|
290 |
-
paper['full_audio'] = None
|
291 |
-
paper['marquee_text'] = None
|
292 |
-
|
293 |
-
def display_papers(papers, marquee_settings):
|
294 |
-
"""Display papers with their audio controls and marquee summaries"""
|
295 |
-
st.write("## Research Papers")
|
296 |
-
|
297 |
-
papercount = 0
|
298 |
-
for paper in papers:
|
299 |
-
papercount += 1
|
300 |
-
if papercount <= 20:
|
301 |
-
# Display marquee if text exists
|
302 |
-
if paper.get('marquee_text'):
|
303 |
-
display_marquee(paper['marquee_text'],
|
304 |
-
marquee_settings,
|
305 |
-
key_suffix=f"paper_{papercount}")
|
306 |
-
|
307 |
-
with st.expander(f"{papercount}. 📄 {paper['title']}", expanded=True):
|
308 |
-
st.markdown(f"**{paper['date']} | {paper['title']} | ⬇️**")
|
309 |
-
st.markdown(f"*{paper['authors']}*")
|
310 |
-
st.markdown(paper['summary'])
|
311 |
-
|
312 |
-
if paper.get('full_audio'):
|
313 |
-
st.write("📚 Paper Audio")
|
314 |
-
file_ext = os.path.splitext(paper['full_audio'])[1].lower().strip('.')
|
315 |
-
if file_ext in ['mp3', 'wav']:
|
316 |
-
st.audio(paper['full_audio'])
|
317 |
-
|
318 |
def perform_ai_lookup(q, vocal_summary=True, extended_refs=False,
|
319 |
-
titles_summary=True, full_audio=False
|
320 |
-
"""Perform Arxiv search with audio generation per paper."""
|
321 |
start = time.time()
|
322 |
|
323 |
-
# Query the HF RAG pipeline
|
324 |
client = Client("awacke1/Arxiv-Paper-Search-And-QA-RAG-Pattern")
|
325 |
refs = client.predict(q, 20, "Semantic Search",
|
326 |
"mistralai/Mixtral-8x7B-Instruct-v0.1",
|
@@ -328,124 +377,76 @@ def perform_ai_lookup(q, vocal_summary=True, extended_refs=False,
|
|
328 |
r2 = client.predict(q, "mistralai/Mixtral-8x7B-Instruct-v0.1",
|
329 |
True, api_name="/ask_llm")
|
330 |
|
331 |
-
# Combine for final text output
|
332 |
result = f"### 🔎 {q}\n\n{r2}\n\n{refs}"
|
333 |
st.markdown(result)
|
334 |
|
335 |
-
|
|
|
|
|
|
|
|
|
336 |
papers = parse_arxiv_refs(refs)
|
337 |
if papers:
|
338 |
create_paper_audio_files(papers, input_question=q)
|
339 |
-
|
340 |
-
display_papers(papers, marquee_settings)
|
341 |
-
else:
|
342 |
-
display_papers(papers, get_marquee_settings())
|
343 |
else:
|
344 |
st.warning("No papers found in the response.")
|
345 |
|
346 |
elapsed = time.time()-start
|
347 |
st.write(f"**Total Elapsed:** {elapsed:.2f} s")
|
348 |
-
|
349 |
-
# Save full transcript
|
350 |
-
create_file(q, result, "md")
|
351 |
return result
|
352 |
|
353 |
-
def
|
354 |
-
|
355 |
-
if not text:
|
356 |
return
|
357 |
-
|
358 |
-
|
359 |
-
|
360 |
-
|
361 |
-
|
362 |
-
|
363 |
-
|
364 |
-
|
365 |
-
|
366 |
-
|
367 |
-
|
368 |
-
|
369 |
-
|
370 |
-
|
371 |
-
|
372 |
-
|
373 |
-
"""Process text with Claude"""
|
374 |
-
if not text:
|
375 |
-
return
|
376 |
-
with st.chat_message("user"):
|
377 |
-
st.markdown(text)
|
378 |
-
with st.chat_message("assistant"):
|
379 |
-
r = claude_client.messages.create(
|
380 |
-
model="claude-3-sonnet-20240229",
|
381 |
-
max_tokens=1000,
|
382 |
-
messages=[{"role":"user","content":text}]
|
383 |
-
)
|
384 |
-
ans = r.content[0].text
|
385 |
-
st.write("Claude-3.5: " + ans)
|
386 |
-
create_file(text, ans, "md")
|
387 |
-
st.session_state.chat_history.append({"user":text,"claude":ans})
|
388 |
-
return ans
|
389 |
|
390 |
def load_files_for_sidebar():
|
391 |
-
"""Load and group files for sidebar display based on first 9 characters of filename"""
|
392 |
md_files = glob.glob("*.md")
|
393 |
mp3_files = glob.glob("*.mp3")
|
394 |
wav_files = glob.glob("*.wav")
|
395 |
-
|
396 |
md_files = [f for f in md_files if os.path.basename(f).lower() != 'readme.md']
|
397 |
all_files = md_files + mp3_files + wav_files
|
398 |
|
399 |
groups = defaultdict(list)
|
|
|
|
|
400 |
for f in all_files:
|
401 |
basename = os.path.basename(f)
|
402 |
-
|
403 |
-
|
404 |
-
|
|
|
|
|
|
|
405 |
sorted_groups = sorted(groups.items(),
|
406 |
-
key=lambda x:
|
407 |
reverse=True)
|
408 |
return sorted_groups
|
409 |
|
410 |
-
def create_zip_of_files(md_files, mp3_files, wav_files, input_question):
|
411 |
-
"""Create zip with intelligent naming based on high-info terms"""
|
412 |
-
md_files = [f for f in md_files if os.path.basename(f).lower() != 'readme.md']
|
413 |
-
all_files = md_files + mp3_files + wav_files
|
414 |
-
if not all_files:
|
415 |
-
return None
|
416 |
-
|
417 |
-
all_content = []
|
418 |
-
for f in all_files:
|
419 |
-
if f.endswith('.md'):
|
420 |
-
with open(f, 'r', encoding='utf-8') as file:
|
421 |
-
all_content.append(file.read())
|
422 |
-
elif f.endswith('.mp3') or f.endswith('.wav'):
|
423 |
-
basename = os.path.splitext(os.path.basename(f))[0]
|
424 |
-
words = basename.replace('_', ' ')
|
425 |
-
all_content.append(words)
|
426 |
-
|
427 |
-
all_content.append(input_question)
|
428 |
-
combined_content = " ".join(all_content)
|
429 |
-
info_terms = get_high_info_terms(combined_content, top_n=10)
|
430 |
-
|
431 |
-
timestamp = datetime.now().strftime("%y%m_%H%M")
|
432 |
-
name_text = '_'.join(term.replace(' ', '-') for term in info_terms[:10])
|
433 |
-
zip_name = f"{timestamp}_{name_text}.zip"
|
434 |
-
|
435 |
-
with zipfile.ZipFile(zip_name, 'w') as z:
|
436 |
-
for f in all_files:
|
437 |
-
z.write(f)
|
438 |
-
|
439 |
-
return zip_name
|
440 |
-
|
441 |
def display_file_manager_sidebar(groups_sorted):
|
442 |
-
"""Display file manager in sidebar with timestamp-based groups"""
|
443 |
st.sidebar.title("🎵 Audio & Docs Manager")
|
444 |
|
445 |
all_md = []
|
446 |
all_mp3 = []
|
447 |
all_wav = []
|
448 |
-
for
|
449 |
for f in files:
|
450 |
if f.endswith(".md"):
|
451 |
all_md.append(f)
|
@@ -454,55 +455,91 @@ def display_file_manager_sidebar(groups_sorted):
|
|
454 |
elif f.endswith(".wav"):
|
455 |
all_wav.append(f)
|
456 |
|
457 |
-
|
458 |
-
with
|
459 |
-
if st.button("🗑
|
460 |
for f in all_md:
|
461 |
os.remove(f)
|
462 |
st.session_state.should_rerun = True
|
463 |
-
with
|
464 |
-
if st.button("🗑
|
465 |
for f in all_mp3:
|
466 |
os.remove(f)
|
467 |
st.session_state.should_rerun = True
|
468 |
-
with
|
469 |
-
if st.button("🗑
|
470 |
for f in all_wav:
|
471 |
os.remove(f)
|
472 |
st.session_state.should_rerun = True
|
473 |
-
with
|
474 |
if st.button("⬇️ ZipAll"):
|
475 |
-
zip_name = create_zip_of_files(all_md, all_mp3, all_wav,
|
476 |
-
input_question=st.session_state.get('last_query', ''))
|
477 |
if zip_name:
|
478 |
-
st.sidebar.markdown(get_download_link(zip_name,
|
479 |
-
unsafe_allow_html=True)
|
480 |
|
481 |
for group_name, files in groups_sorted:
|
482 |
-
|
483 |
-
|
484 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
485 |
with st.sidebar.expander(f"📁 {group_label} ({len(files)})", expanded=True):
|
486 |
c1, c2 = st.columns(2)
|
487 |
with c1:
|
488 |
-
if st.button("👀
|
489 |
st.session_state.viewing_prefix = group_name
|
490 |
with c2:
|
491 |
-
if st.button("🗑
|
492 |
for f in files:
|
493 |
os.remove(f)
|
494 |
-
st.success(f"Deleted group {
|
495 |
st.session_state.should_rerun = True
|
496 |
-
|
497 |
for f in files:
|
498 |
fname = os.path.basename(f)
|
499 |
ext = os.path.splitext(fname)[1].lower()
|
500 |
emoji = FILE_EMOJIS.get(ext.strip('.'), '')
|
501 |
-
|
|
|
502 |
st.write(f"{emoji} **{fname}** - {ctime}")
|
503 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
504 |
def main():
|
505 |
-
#
|
|
|
506 |
marquee_settings = get_marquee_settings()
|
507 |
|
508 |
# Initial welcome marquee
|
@@ -556,7 +593,7 @@ def main():
|
|
556 |
val_stripped = val.replace('\\n', ' ')
|
557 |
edited_input = st.text_area("✏️ Edit Input:", value=val_stripped, height=100)
|
558 |
|
559 |
-
run_option = st.selectbox("Model:", ["Arxiv"
|
560 |
col1, col2 = st.columns(2)
|
561 |
with col1:
|
562 |
autorun = st.checkbox("⚙ AutoRun", value=True)
|
@@ -568,30 +605,15 @@ def main():
|
|
568 |
if autorun and input_changed:
|
569 |
st.session_state.old_val = val
|
570 |
st.session_state.last_query = edited_input
|
571 |
-
|
572 |
-
|
573 |
-
titles_summary=True, full_audio=full_audio,
|
574 |
-
marquee_settings=marquee_settings)
|
575 |
-
else:
|
576 |
-
if run_option == "GPT-4o":
|
577 |
-
process_with_gpt(edited_input)
|
578 |
-
elif run_option == "Claude-3.5":
|
579 |
-
process_with_claude(edited_input)
|
580 |
else:
|
581 |
if st.button("▶ Run"):
|
582 |
st.session_state.old_val = val
|
583 |
st.session_state.last_query = edited_input
|
584 |
-
|
585 |
-
|
586 |
-
titles_summary=True, full_audio=full_audio,
|
587 |
-
marquee_settings=marquee_settings)
|
588 |
-
else:
|
589 |
-
if run_option == "GPT-4o":
|
590 |
-
process_with_gpt(edited_input)
|
591 |
-
elif run_option == "Claude-3.5":
|
592 |
-
process_with_claude(edited_input)
|
593 |
|
594 |
-
# ArXiv Tab
|
595 |
if tab_main == "🔍 ArXiv":
|
596 |
st.subheader("🔍 Query ArXiv")
|
597 |
q = st.text_input("🔍 Query:")
|
@@ -606,32 +628,23 @@ def main():
|
|
606 |
if q and st.button("🔍Run"):
|
607 |
st.session_state.last_query = q
|
608 |
result = perform_ai_lookup(q, vocal_summary=vocal_summary, extended_refs=extended_refs,
|
609 |
-
titles_summary=titles_summary, full_audio=full_audio
|
610 |
-
marquee_settings=marquee_settings)
|
611 |
if full_transcript:
|
612 |
create_file(q, result, "md")
|
613 |
|
614 |
-
# Voice Tab
|
615 |
elif tab_main == "🎤 Voice":
|
616 |
st.subheader("🎤 Voice Input")
|
617 |
user_text = st.text_area("💬 Message:", height=100)
|
618 |
user_text = user_text.strip().replace('\n', ' ')
|
619 |
|
620 |
if st.button("📨 Send"):
|
621 |
-
|
622 |
|
623 |
st.subheader("📜 Chat History")
|
624 |
-
|
625 |
-
|
626 |
-
|
627 |
-
|
628 |
-
st.write("**Claude:**", c["claude"])
|
629 |
-
with t2:
|
630 |
-
for m in st.session_state.messages:
|
631 |
-
with st.chat_message(m["role"]):
|
632 |
-
st.markdown(m["content"])
|
633 |
-
|
634 |
-
# Media Tab
|
635 |
elif tab_main == "📸 Media":
|
636 |
st.header("📸 Images & 🎥 Videos")
|
637 |
tabs = st.tabs(["🖼 Images", "🎥 Video"])
|
@@ -681,7 +694,6 @@ def main():
|
|
681 |
else:
|
682 |
st.write("No videos found.")
|
683 |
|
684 |
-
# Editor Tab
|
685 |
elif tab_main == "📝 Editor":
|
686 |
if st.session_state.editing_file:
|
687 |
st.subheader(f"Editing: {st.session_state.editing_file}")
|
@@ -720,7 +732,6 @@ def main():
|
|
720 |
st.session_state.viewing_prefix = None
|
721 |
st.session_state['marquee_content'] = "🚀 Welcome to TalkingAIResearcher | 🤖 Your Research Assistant"
|
722 |
|
723 |
-
# Add custom CSS
|
724 |
st.markdown("""
|
725 |
<style>
|
726 |
.main { background: linear-gradient(to right, #1a1a1a, #2d2d2d); color: #fff; }
|
@@ -729,10 +740,9 @@ def main():
|
|
729 |
</style>
|
730 |
""", unsafe_allow_html=True)
|
731 |
|
732 |
-
# Handle rerun if needed
|
733 |
if st.session_state.should_rerun:
|
734 |
st.session_state.should_rerun = False
|
735 |
st.rerun()
|
736 |
|
737 |
if __name__ == "__main__":
|
738 |
-
main()
|
|
|
37 |
|
38 |
# Add available English voices for Edge TTS
|
39 |
EDGE_TTS_VOICES = [
|
40 |
+
"en-US-AriaNeural",
|
41 |
+
"en-US-GuyNeural",
|
42 |
"en-US-JennyNeural",
|
43 |
"en-GB-SoniaNeural",
|
44 |
"en-GB-RyanNeural",
|
|
|
49 |
]
|
50 |
|
51 |
# Initialize session state variables
|
52 |
+
if 'marquee_settings' not in st.session_state:
|
53 |
+
st.session_state['marquee_settings'] = {
|
54 |
+
"background": "#1E1E1E",
|
55 |
+
"color": "#FFFFFF",
|
56 |
+
"font-size": "14px",
|
57 |
+
"animationDuration": "10s",
|
58 |
+
"width": "100%",
|
59 |
+
"lineHeight": "35px"
|
60 |
+
}
|
61 |
+
|
62 |
if 'tts_voice' not in st.session_state:
|
63 |
st.session_state['tts_voice'] = EDGE_TTS_VOICES[0]
|
64 |
if 'audio_format' not in st.session_state:
|
|
|
112 |
"wav": "🔊"
|
113 |
}
|
114 |
|
115 |
+
def get_central_time():
|
116 |
+
"""Get current time in US Central timezone"""
|
117 |
+
central = pytz.timezone('US/Central')
|
118 |
+
return datetime.now(central)
|
119 |
+
|
120 |
+
def format_timestamp_prefix():
|
121 |
+
"""Generate timestamp prefix in format MM_dd_yy_hh_mm_AM/PM"""
|
122 |
+
ct = get_central_time()
|
123 |
+
return ct.strftime("%m_%d_%y_%I_%M_%p")
|
124 |
+
|
125 |
+
def initialize_marquee_settings():
|
126 |
+
"""Initialize marquee settings in session state"""
|
127 |
+
if 'marquee_settings' not in st.session_state:
|
128 |
+
st.session_state['marquee_settings'] = {
|
129 |
+
"background": "#1E1E1E",
|
130 |
+
"color": "#FFFFFF",
|
131 |
+
"font-size": "14px",
|
132 |
+
"animationDuration": "10s",
|
133 |
+
"width": "100%",
|
134 |
+
"lineHeight": "35px"
|
135 |
+
}
|
136 |
+
|
137 |
def get_marquee_settings():
|
138 |
+
"""Get or update marquee settings from session state"""
|
139 |
+
initialize_marquee_settings()
|
140 |
+
return st.session_state['marquee_settings']
|
141 |
+
|
142 |
+
def update_marquee_settings_ui():
|
143 |
+
"""Update marquee settings via UI controls"""
|
144 |
+
initialize_marquee_settings()
|
145 |
st.sidebar.markdown("### 🎯 Marquee Settings")
|
146 |
cols = st.sidebar.columns(2)
|
147 |
with cols[0]:
|
148 |
+
bg_color = st.color_picker("🎨 Background",
|
149 |
+
st.session_state['marquee_settings']["background"],
|
150 |
+
key="bg_color_picker")
|
151 |
+
text_color = st.color_picker("✍️ Text",
|
152 |
+
st.session_state['marquee_settings']["color"],
|
153 |
+
key="text_color_picker")
|
154 |
with cols[1]:
|
155 |
font_size = st.slider("📏 Size", 10, 24, 14, key="font_size_slider")
|
156 |
duration = st.slider("⏱️ Speed", 1, 20, 10, key="duration_slider")
|
157 |
+
|
158 |
+
st.session_state['marquee_settings'].update({
|
159 |
"background": bg_color,
|
160 |
"color": text_color,
|
161 |
"font-size": f"{font_size}px",
|
162 |
+
"animationDuration": f"{duration}s"
|
163 |
+
})
|
|
|
|
|
164 |
|
165 |
def display_marquee(text, settings, key_suffix=""):
|
166 |
"""Display marquee with given text and settings"""
|
|
|
172 |
)
|
173 |
st.write("")
|
174 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
175 |
def get_high_info_terms(text: str, top_n=10) -> list:
|
176 |
stop_words = set(['the', 'a', 'an', 'and', 'or', 'but', 'in', 'on', 'at', 'to', 'for', 'of', 'with'])
|
177 |
words = re.findall(r'\b\w+(?:-\w+)*\b', text.lower())
|
|
|
189 |
filtered = [w for w in words if len(w) > 3 and w not in stop_short]
|
190 |
return '_'.join(filtered)[:200]
|
191 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
192 |
def generate_filename(prompt, response, file_type="md"):
|
193 |
+
prefix = format_timestamp_prefix() + "_"
|
194 |
combined = (prompt + " " + response).strip()
|
195 |
info_terms = get_high_info_terms(combined, top_n=10)
|
196 |
snippet = (prompt[:100] + " " + response[:100]).strip()
|
|
|
221 |
else:
|
222 |
return f'<a href="data:application/octet-stream;base64,{b64}" download="{os.path.basename(file)}">Download {os.path.basename(file)}</a>'
|
223 |
|
224 |
+
def clean_for_speech(text: str) -> str:
|
225 |
+
text = text.replace("\n", " ")
|
226 |
+
text = text.replace("</s>", " ")
|
227 |
+
text = text.replace("#", "")
|
228 |
+
text = re.sub(r"\(https?:\/\/[^\)]+\)", "", text)
|
229 |
+
text = re.sub(r"\s+", " ", text).strip()
|
230 |
+
return text
|
231 |
+
|
232 |
async def edge_tts_generate_audio(text, voice="en-US-AriaNeural", rate=0, pitch=0, file_format="mp3"):
|
233 |
text = clean_for_speech(text)
|
234 |
if not text.strip():
|
|
|
249 |
dl_link = get_download_link(file_path, file_type=file_type)
|
250 |
st.markdown(dl_link, unsafe_allow_html=True)
|
251 |
|
252 |
+
def save_qa_with_audio(question, answer, voice=None):
|
253 |
+
"""Save Q&A to markdown and generate audio"""
|
254 |
+
if not voice:
|
255 |
+
voice = st.session_state['tts_voice']
|
256 |
+
|
257 |
+
# Create markdown file
|
258 |
+
combined_text = f"# Question\n{question}\n\n# Answer\n{answer}"
|
259 |
+
md_file = create_file(question, answer, "md")
|
260 |
+
|
261 |
+
# Generate audio file
|
262 |
+
audio_text = f"Question: {question}\n\nAnswer: {answer}"
|
263 |
+
audio_file = speak_with_edge_tts(
|
264 |
+
audio_text,
|
265 |
+
voice=voice,
|
266 |
+
file_format=st.session_state['audio_format']
|
267 |
+
)
|
268 |
+
|
269 |
+
return md_file, audio_file
|
270 |
+
|
271 |
+
def process_paper_content(paper):
|
272 |
+
marquee_text = f"📄 {paper['title']} | 👤 {paper['authors'][:100]} | 📝 {paper['summary'][:100]}"
|
273 |
+
audio_text = f"{paper['title']} by {paper['authors']}. {paper['summary']}"
|
274 |
+
return marquee_text, audio_text
|
275 |
+
|
276 |
+
def create_paper_audio_files(papers, input_question):
|
277 |
+
for paper in papers:
|
278 |
+
try:
|
279 |
+
marquee_text, audio_text = process_paper_content(paper)
|
280 |
+
|
281 |
+
audio_text = clean_for_speech(audio_text)
|
282 |
+
file_format = st.session_state['audio_format']
|
283 |
+
audio_file = speak_with_edge_tts(audio_text,
|
284 |
+
voice=st.session_state['tts_voice'],
|
285 |
+
file_format=file_format)
|
286 |
+
paper['full_audio'] = audio_file
|
287 |
+
|
288 |
+
st.write(f"### {FILE_EMOJIS.get(file_format, '')} {os.path.basename(audio_file)}")
|
289 |
+
play_and_download_audio(audio_file, file_type=file_format)
|
290 |
+
paper['marquee_text'] = marquee_text
|
291 |
+
|
292 |
+
except Exception as e:
|
293 |
+
st.warning(f"Error processing paper {paper['title']}: {str(e)}")
|
294 |
+
paper['full_audio'] = None
|
295 |
+
paper['marquee_text'] = None
|
296 |
+
|
297 |
+
def display_papers(papers, marquee_settings):
|
298 |
+
st.write("## Research Papers")
|
299 |
+
|
300 |
+
papercount = 0
|
301 |
+
for paper in papers:
|
302 |
+
papercount += 1
|
303 |
+
if papercount <= 20:
|
304 |
+
if paper.get('marquee_text'):
|
305 |
+
display_marquee(paper['marquee_text'],
|
306 |
+
marquee_settings,
|
307 |
+
key_suffix=f"paper_{papercount}")
|
308 |
+
|
309 |
+
with st.expander(f"{papercount}. 📄 {paper['title']}", expanded=True):
|
310 |
+
st.markdown(f"**{paper['date']} | {paper['title']} | ⬇️**")
|
311 |
+
st.markdown(f"*{paper['authors']}*")
|
312 |
+
st.markdown(paper['summary'])
|
313 |
+
|
314 |
+
if paper.get('full_audio'):
|
315 |
+
st.write("📚 Paper Audio")
|
316 |
+
file_ext = os.path.splitext(paper['full_audio'])[1].lower().strip('.')
|
317 |
+
if file_ext in ['mp3', 'wav']:
|
318 |
+
st.audio(paper['full_audio'])
|
319 |
+
|
320 |
def parse_arxiv_refs(ref_text: str):
|
321 |
if not ref_text:
|
322 |
return []
|
|
|
366 |
|
367 |
return results[:20]
|
368 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
369 |
def perform_ai_lookup(q, vocal_summary=True, extended_refs=False,
|
370 |
+
titles_summary=True, full_audio=False):
|
|
|
371 |
start = time.time()
|
372 |
|
|
|
373 |
client = Client("awacke1/Arxiv-Paper-Search-And-QA-RAG-Pattern")
|
374 |
refs = client.predict(q, 20, "Semantic Search",
|
375 |
"mistralai/Mixtral-8x7B-Instruct-v0.1",
|
|
|
377 |
r2 = client.predict(q, "mistralai/Mixtral-8x7B-Instruct-v0.1",
|
378 |
True, api_name="/ask_llm")
|
379 |
|
|
|
380 |
result = f"### 🔎 {q}\n\n{r2}\n\n{refs}"
|
381 |
st.markdown(result)
|
382 |
|
383 |
+
md_file, audio_file = save_qa_with_audio(q, result)
|
384 |
+
|
385 |
+
st.subheader("📝 Main Response Audio")
|
386 |
+
play_and_download_audio(audio_file, st.session_state['audio_format'])
|
387 |
+
|
388 |
papers = parse_arxiv_refs(refs)
|
389 |
if papers:
|
390 |
create_paper_audio_files(papers, input_question=q)
|
391 |
+
display_papers(papers, get_marquee_settings())
|
|
|
|
|
|
|
392 |
else:
|
393 |
st.warning("No papers found in the response.")
|
394 |
|
395 |
elapsed = time.time()-start
|
396 |
st.write(f"**Total Elapsed:** {elapsed:.2f} s")
|
|
|
|
|
|
|
397 |
return result
|
398 |
|
399 |
+
def process_voice_input(text):
|
400 |
+
if not text:
|
|
|
401 |
return
|
402 |
+
|
403 |
+
st.subheader("🔍 Search Results")
|
404 |
+
result = perform_ai_lookup(
|
405 |
+
text,
|
406 |
+
vocal_summary=True,
|
407 |
+
extended_refs=False,
|
408 |
+
titles_summary=True,
|
409 |
+
full_audio=True
|
410 |
+
)
|
411 |
+
|
412 |
+
md_file, audio_file = save_qa_with_audio(text, result)
|
413 |
+
|
414 |
+
st.subheader("📝 Generated Files")
|
415 |
+
st.write(f"Markdown: {md_file}")
|
416 |
+
st.write(f"Audio: {audio_file}")
|
417 |
+
play_and_download_audio(audio_file, st.session_state['audio_format'])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
418 |
|
419 |
def load_files_for_sidebar():
|
|
|
420 |
md_files = glob.glob("*.md")
|
421 |
mp3_files = glob.glob("*.mp3")
|
422 |
wav_files = glob.glob("*.wav")
|
423 |
+
|
424 |
md_files = [f for f in md_files if os.path.basename(f).lower() != 'readme.md']
|
425 |
all_files = md_files + mp3_files + wav_files
|
426 |
|
427 |
groups = defaultdict(list)
|
428 |
+
prefix_length = len("MM_dd_yy_hh_mm_AP")
|
429 |
+
|
430 |
for f in all_files:
|
431 |
basename = os.path.basename(f)
|
432 |
+
if len(basename) >= prefix_length and '_' in basename:
|
433 |
+
group_name = basename[:prefix_length]
|
434 |
+
groups[group_name].append(f)
|
435 |
+
else:
|
436 |
+
groups['Other'].append(f)
|
437 |
+
|
438 |
sorted_groups = sorted(groups.items(),
|
439 |
+
key=lambda x: x[0] if x[0] != 'Other' else '',
|
440 |
reverse=True)
|
441 |
return sorted_groups
|
442 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
443 |
def display_file_manager_sidebar(groups_sorted):
|
|
|
444 |
st.sidebar.title("🎵 Audio & Docs Manager")
|
445 |
|
446 |
all_md = []
|
447 |
all_mp3 = []
|
448 |
all_wav = []
|
449 |
+
for _, files in groups_sorted:
|
450 |
for f in files:
|
451 |
if f.endswith(".md"):
|
452 |
all_md.append(f)
|
|
|
455 |
elif f.endswith(".wav"):
|
456 |
all_wav.append(f)
|
457 |
|
458 |
+
col1, col2, col3, col4 = st.sidebar.columns(4)
|
459 |
+
with col1:
|
460 |
+
if st.button("🗑 DelMD"):
|
461 |
for f in all_md:
|
462 |
os.remove(f)
|
463 |
st.session_state.should_rerun = True
|
464 |
+
with col2:
|
465 |
+
if st.button("🗑 DelMP3"):
|
466 |
for f in all_mp3:
|
467 |
os.remove(f)
|
468 |
st.session_state.should_rerun = True
|
469 |
+
with col3:
|
470 |
+
if st.button("🗑 DelWAV"):
|
471 |
for f in all_wav:
|
472 |
os.remove(f)
|
473 |
st.session_state.should_rerun = True
|
474 |
+
with col4:
|
475 |
if st.button("⬇️ ZipAll"):
|
476 |
+
zip_name = create_zip_of_files(all_md, all_mp3, all_wav, st.session_state.get('last_query', ''))
|
|
|
477 |
if zip_name:
|
478 |
+
st.sidebar.markdown(get_download_link(zip_name, "zip"), unsafe_allow_html=True)
|
|
|
479 |
|
480 |
for group_name, files in groups_sorted:
|
481 |
+
if group_name == 'Other':
|
482 |
+
group_label = 'Other Files'
|
483 |
+
else:
|
484 |
+
try:
|
485 |
+
timestamp_dt = datetime.strptime(group_name, "%m_%d_%y_%I_%M_%p")
|
486 |
+
group_label = timestamp_dt.strftime("%b %d, %Y %I:%M %p")
|
487 |
+
except ValueError:
|
488 |
+
group_label = group_name
|
489 |
+
|
490 |
with st.sidebar.expander(f"📁 {group_label} ({len(files)})", expanded=True):
|
491 |
c1, c2 = st.columns(2)
|
492 |
with c1:
|
493 |
+
if st.button("👀 View", key=f"view_group_{group_name}"):
|
494 |
st.session_state.viewing_prefix = group_name
|
495 |
with c2:
|
496 |
+
if st.button("🗑 Del", key=f"del_group_{group_name}"):
|
497 |
for f in files:
|
498 |
os.remove(f)
|
499 |
+
st.success(f"Deleted group {group_label}!")
|
500 |
st.session_state.should_rerun = True
|
501 |
+
|
502 |
for f in files:
|
503 |
fname = os.path.basename(f)
|
504 |
ext = os.path.splitext(fname)[1].lower()
|
505 |
emoji = FILE_EMOJIS.get(ext.strip('.'), '')
|
506 |
+
mtime = os.path.getmtime(f)
|
507 |
+
ctime = datetime.fromtimestamp(mtime).strftime("%I:%M:%S %p")
|
508 |
st.write(f"{emoji} **{fname}** - {ctime}")
|
509 |
|
510 |
+
def create_zip_of_files(md_files, mp3_files, wav_files, input_question):
|
511 |
+
md_files = [f for f in md_files if os.path.basename(f).lower() != 'readme.md']
|
512 |
+
all_files = md_files + mp3_files + wav_files
|
513 |
+
if not all_files:
|
514 |
+
return None
|
515 |
+
|
516 |
+
all_content = []
|
517 |
+
for f in all_files:
|
518 |
+
if f.endswith('.md'):
|
519 |
+
with open(f, 'r', encoding='utf-8') as file:
|
520 |
+
all_content.append(file.read())
|
521 |
+
elif f.endswith('.mp3') or f.endswith('.wav'):
|
522 |
+
basename = os.path.splitext(os.path.basename(f))[0]
|
523 |
+
words = basename.replace('_', ' ')
|
524 |
+
all_content.append(words)
|
525 |
+
|
526 |
+
all_content.append(input_question)
|
527 |
+
combined_content = " ".join(all_content)
|
528 |
+
info_terms = get_high_info_terms(combined_content, top_n=10)
|
529 |
+
|
530 |
+
timestamp = format_timestamp_prefix()
|
531 |
+
name_text = '_'.join(term.replace(' ', '-') for term in info_terms[:10])
|
532 |
+
zip_name = f"{timestamp}_{name_text}.zip"
|
533 |
+
|
534 |
+
with zipfile.ZipFile(zip_name, 'w') as z:
|
535 |
+
for f in all_files:
|
536 |
+
z.write(f)
|
537 |
+
|
538 |
+
return zip_name
|
539 |
+
|
540 |
def main():
|
541 |
+
# Update marquee settings UI first
|
542 |
+
update_marquee_settings_ui()
|
543 |
marquee_settings = get_marquee_settings()
|
544 |
|
545 |
# Initial welcome marquee
|
|
|
593 |
val_stripped = val.replace('\\n', ' ')
|
594 |
edited_input = st.text_area("✏️ Edit Input:", value=val_stripped, height=100)
|
595 |
|
596 |
+
run_option = st.selectbox("Model:", ["Arxiv"])
|
597 |
col1, col2 = st.columns(2)
|
598 |
with col1:
|
599 |
autorun = st.checkbox("⚙ AutoRun", value=True)
|
|
|
605 |
if autorun and input_changed:
|
606 |
st.session_state.old_val = val
|
607 |
st.session_state.last_query = edited_input
|
608 |
+
result = perform_ai_lookup(edited_input, vocal_summary=True, extended_refs=False,
|
609 |
+
titles_summary=True, full_audio=full_audio)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
610 |
else:
|
611 |
if st.button("▶ Run"):
|
612 |
st.session_state.old_val = val
|
613 |
st.session_state.last_query = edited_input
|
614 |
+
result = perform_ai_lookup(edited_input, vocal_summary=True, extended_refs=False,
|
615 |
+
titles_summary=True, full_audio=full_audio)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
616 |
|
|
|
617 |
if tab_main == "🔍 ArXiv":
|
618 |
st.subheader("🔍 Query ArXiv")
|
619 |
q = st.text_input("🔍 Query:")
|
|
|
628 |
if q and st.button("🔍Run"):
|
629 |
st.session_state.last_query = q
|
630 |
result = perform_ai_lookup(q, vocal_summary=vocal_summary, extended_refs=extended_refs,
|
631 |
+
titles_summary=titles_summary, full_audio=full_audio)
|
|
|
632 |
if full_transcript:
|
633 |
create_file(q, result, "md")
|
634 |
|
|
|
635 |
elif tab_main == "🎤 Voice":
|
636 |
st.subheader("🎤 Voice Input")
|
637 |
user_text = st.text_area("💬 Message:", height=100)
|
638 |
user_text = user_text.strip().replace('\n', ' ')
|
639 |
|
640 |
if st.button("📨 Send"):
|
641 |
+
process_voice_input(user_text)
|
642 |
|
643 |
st.subheader("📜 Chat History")
|
644 |
+
for c in st.session_state.chat_history:
|
645 |
+
st.write("**You:**", c["user"])
|
646 |
+
st.write("**Response:**", c["claude"])
|
647 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
648 |
elif tab_main == "📸 Media":
|
649 |
st.header("📸 Images & 🎥 Videos")
|
650 |
tabs = st.tabs(["🖼 Images", "🎥 Video"])
|
|
|
694 |
else:
|
695 |
st.write("No videos found.")
|
696 |
|
|
|
697 |
elif tab_main == "📝 Editor":
|
698 |
if st.session_state.editing_file:
|
699 |
st.subheader(f"Editing: {st.session_state.editing_file}")
|
|
|
732 |
st.session_state.viewing_prefix = None
|
733 |
st.session_state['marquee_content'] = "🚀 Welcome to TalkingAIResearcher | 🤖 Your Research Assistant"
|
734 |
|
|
|
735 |
st.markdown("""
|
736 |
<style>
|
737 |
.main { background: linear-gradient(to right, #1a1a1a, #2d2d2d); color: #fff; }
|
|
|
740 |
</style>
|
741 |
""", unsafe_allow_html=True)
|
742 |
|
|
|
743 |
if st.session_state.should_rerun:
|
744 |
st.session_state.should_rerun = False
|
745 |
st.rerun()
|
746 |
|
747 |
if __name__ == "__main__":
|
748 |
+
main()
|