anjohan · April 27, 2026 19:49
diff --git a/mrsscheduler.py b/mrsscheduler.py
 """
 Script for going through all talks at MRS Spring 2026 and using AI to evaluate whether
 a talk is interesting or not. The script generates a .ics calendar file.
 This uses the Cerebras Cloud SDK as the AI endpoint, since this was the easiest
 to get running for me. Each run costs ~$0.25.

 This script was mostly written by Gemini 3.1 Pro within a few iterations.

 STEP 1:
    Download abstract book from https://www.mrs.org/docs/default-source/meetings-events/spring-meetings/2026/2026-mrs-spring-meeting-program.pdf

 STEP 2:
    Install PDF-to-markdown converter and run it (takes ~5 minutes):
        pip install markitdown[pdf]
        markitdown mrs.pdf > mrs.md

 STEP 3:
    Install Cerebras Cloud:
        pip install cerebras-cloud-sdk
    Sign up for cloud.cerebras.ai.
    Go to the billing page and select "pay as you go",
    and add $10 of value, otherwise you run out of tokens per minute/day.
    Get an API key (I think it changes when you switch from free tier).

 STEP 4:
    Edit the prompt_intro variable below with your interests.
    The example below finds 235 talks.

 STEP 5:
    Run this script (2-3 minutes) with your API key, which will produce schedule.ics
        CEREBRAS_API_KEY=YOURKEYHERE python mrsscheduler.py

 STEP 6:
    Import schedule.ics in your favorite calendar software
    Make a separate calendar, to make the events easy to
    delete in case your are not happy with the result.
    If using Google Calendar, make a separate MRS calendar,
    then import the events to that calendar.
    (Warning: It will by default add it to your main calendar.)
 """


 import os
 import re
 import uuid
 import multiprocessing
 from tqdm import tqdm
 from datetime import datetime, timedelta

 from cerebras.cloud.sdk import Cerebras

 client = Cerebras(
    api_key=os.environ.get("CEREBRAS_API_KEY")
 )

 # CONFIGURATION
 num_workers = 8 # within 1k requests/minute rate limit
 input_file = "mrs.md"
 output_file = "schedule.ics"

 prompt_intro = """I am interested methods development and computational aspects of molecular dynamics, atomistic simulations, machine learned interatomic potentials, and high-performance computing. I am not interested in pure application talks without methodological or computational novelty. I am also not interested in machine learning applied to domains that are not related to atomistic simulations. If relevant to ML potentials, I am also interested in computational quantum methods such as DFT."""


 def extract_talks_from_markdown(md_text):
    talks =[]
    md_text = re.sub(r'2026 MRS Spring Meeting\s+Current as of April 15, 2026', '', md_text)

    current_date, current_place, is_poster, current_talk = None, None, False, None
    state = 'SEARCHING'

    session_re = re.compile(r"^SESSION\s+[A-Z0-9.]+:\s+(.*)")
    talk_re = re.compile(r"^(\d{1,2}:\d{2}\s+[AP]M)\s+([*+^]*[A-Z0-9]+\.\d+\s*\.\d+)")
    break_re = re.compile(r"^(\d{1,2}:\d{2}\s+[AP]M)\s+(BREAK|PANEL DISCUSSION)")
    # Explicit regex to identify the known date bounds of the conference
    date_pattern = re.compile(r"^(Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday).*?(April|May)\s+\d{1,2},\s+2026")

    lines = md_text.split('\n')
    i = 0

    def process_header(header_text):
        """Treats the Title and Authors as one string, just stripping affiliations."""
        header_text = header_text.replace('\n', ' ').strip()
        if ';' in header_text:
            parts = header_text.split(';', 1)
            title_authors = parts[0].strip()
            affiliations = parts[1].strip()
        else:
            title_authors = header_text
            affiliations = ""

        return title_authors, "", affiliations

    while i < len(lines):
        line = lines[i].strip()

        if not line:
            if state == 'HEADER' and current_talk:
                t, a, aff = process_header(current_talk['raw_header'])
                current_talk['title'], current_talk['authors'], current_talk['affiliations'] = t, a, aff
                state = 'ABSTRACT'
            elif state == 'ABSTRACT' and current_talk:
                if current_talk['abstract'] and not current_talk['abstract'].endswith('\n\n'):
                    current_talk['abstract'] += '\n\n'
            i += 1
            continue

        match_session = session_re.match(line)
        if match_session:
            if current_talk and not is_poster:
                # Session ended. Current talk is last in session, end time remains None.
                current_talk['abstract'] = current_talk['abstract'].strip()
                talks.append(current_talk)
                current_talk = None
            is_poster = "Poster" in match_session.group(1).strip()

            # Look ahead for a valid date line
            current_date, current_place = None, None
            for j in range(i + 1, min(i + 10, len(lines))):
                lookahead_line = lines[j].strip()
                if date_pattern.match(lookahead_line):
                    current_date = lookahead_line
                    # The next non-empty, non-event line is generally the location
                    for k in range(j + 1, min(j + 5, len(lines))):
                        place_line = lines[k].strip()
                        if place_line and not talk_re.match(place_line) and not break_re.match(place_line):
                            current_place = place_line
                            break
                    break

            state = 'SEARCHING'
            i += 1
            continue

        match_break = break_re.match(line)
        if match_break:
            time_str, break_type = match_break.groups()
            if current_talk and not is_poster:
                current_talk['end_time_str'] = time_str.strip()  # The break signifies the end of the previous talk
                current_talk['abstract'] = current_talk['abstract'].strip()
                talks.append(current_talk)
                current_talk = None
            state = 'SEARCHING'
            i += 1
            continue

        match_talk = talk_re.match(line)
        if match_talk:
            time_str, talk_id = match_talk.groups()

            if current_talk and not is_poster:
                current_talk['end_time_str'] = time_str.strip()  # The new talk signifies the end of the previous talk
                current_talk['abstract'] = current_talk['abstract'].strip()
                talks.append(current_talk)

            if not is_poster:
                current_talk = {
                    'talk_id': talk_id.strip(),
                    'time': time_str.strip(),
                    'end_time_str': None, # To be dynamically populated
                    'date': current_date,
                    'place': current_place,
                    'title': '', 'authors': '', 'affiliations': '', 'raw_header': '', 'abstract': ''
                }
                state = 'HEADER'
            else:
                current_talk, state = None, 'SEARCHING'
            i += 1
            continue

        if not is_poster and current_talk:
            if state == 'HEADER': current_talk['raw_header'] += line + " "
            elif state == 'ABSTRACT':
                if line.startswith("SYMPOSIUM BI") or line.startswith("Symposium Organizers"):
                    state = 'SEARCHING'
                    continue
                if current_talk['abstract'] and not current_talk['abstract'].endswith('\n'):
                    current_talk['abstract'] += ' '
                current_talk['abstract'] += line
        i += 1

    if current_talk and not is_poster:
        if state == 'HEADER':
            t, a, aff = process_header(current_talk['raw_header'])
            current_talk['title'], current_talk['authors'], current_talk['affiliations'] = t, a, aff
        current_talk['abstract'] = current_talk['abstract'].strip()
        talks.append(current_talk)

    for t in talks: t.pop('raw_header', None)
    return talks


 def is_interesting(talk):
    """Evaluate a single talk using AI"""

    prompt_template = prompt_intro + "\n" +  """

    Answering only yes or no, do you think I would be interested in the following talk?

    Title and Authors: {title}
    Abstract: {abstract}
    """

    prompt = prompt_template.format(
            title=talk['title'],
            abstract=talk['abstract']
            )

    try:
        completion = client.chat.completions.create(
                messages=[{"role":"user","content":prompt}],
                model="llama3.1-8b",
                max_completion_tokens=1024,
                temperature=0.2,
                top_p=1,
                stream=False
                )

        answer = completion.choices[0].message.content
        # Clean response: lowercased and stripped of punctuation
        answer = re.sub(r'[^a-z]', '', answer.strip().lower())

        if answer == 'yes':
            return True
    except Exception as e:
        print(f"Error calling API for talk '{talk['title'][:30]}...': {e}")
    return False


 def get_interesting_talks(talks):
    """Use AI to evaluate each talk based on the user's prompt."""
    interesting =[]

    print(f"Evaluating {len(talks)} talks using LLM...")
    with multiprocessing.Pool(processes=num_workers) as pool:
        results = pool.imap(is_interesting, talks)

        for i, result in enumerate(tqdm(results, total=len(talks))):
            if result:
                interesting.append(talks[i])

    return interesting


 def parse_datetime(date_str, time_str):
    """Converts the date and time strings into datetime objects."""
    if not date_str or not time_str:
        return None

    try:
        # Extract purely the valid date using regex to avoid artifacts
        match = re.search(r'(April|May)\s+\d{1,2},\s+2026', date_str)
        if not match:
            return None

        clean_date = match.group(0)
        combined = f"{clean_date} {time_str.strip()}"
        return datetime.strptime(combined, "%B %d, %Y %I:%M %p")
    except Exception as e:
        print(f"Warning: Could not parse datetime ({date_str} {time_str}) - {e}")
        return None


 def get_start_end_dt(talk):
    """Helper function to calculate the duration of the talk."""
    start_dt = parse_datetime(talk['date'], talk['time'])
    if not start_dt:
        return None, None

    if talk.get('end_time_str'):
        end_dt = parse_datetime(talk['date'], talk['end_time_str'])
        # Safeguard if AM/PM boundary is crossed unpredictably
        if not end_dt or end_dt <= start_dt:
            end_dt = start_dt + timedelta(minutes=15)
    else:
        # Default to 15 mins if it's the last talk before a session ends
        end_dt = start_dt + timedelta(minutes=15)

    return start_dt, end_dt


 def fold_ics_line(line):
    """Folds lines longer than 75 characters per RFC 5545 specification."""
    folded =[]
    while len(line) > 75:
        folded.append(line[:75])
        line = " " + line[75:]
    folded.append(line)
    return "\r\n".join(folded)


 def generate_ics(talks, filename="schedule.ics"):
    """Generates an iCalendar (.ics) file from the filtered talks."""
    ics_lines =[
        "BEGIN:VCALENDAR",
        "VERSION:2.0",
        "PRODID:-//User Name//Conference Schedule 1.0//EN",
        "CALSCALE:GREGORIAN"
    ]

    for talk in talks:
        start_dt, end_dt = get_start_end_dt(talk)
        if not start_dt or not end_dt:
            continue

        # Datetime formatted for ICS (floating local time)
        dtstart = start_dt.strftime("%Y%m%dT%H%M%S")
        dtend = end_dt.strftime("%Y%m%dT%H%M%S")

        # Escape special characters for valid ICS text using raw strings to prevent SyntaxWarning
        escape = lambda s: str(s).replace('\\', r'\\').replace(';', r'\;').replace(',', r'\,').replace('\n', r'\n')

        summary = escape(talk['title'])
        location = escape(talk['place'])
        description = escape(f"Title/Authors: {talk['title']}\n\n{talk['abstract']}")
        uid = str(uuid.uuid4()) + "@mrsschedule"

        event_lines =[
            "BEGIN:VEVENT",
            f"UID:{uid}",
            f"DTSTAMP:{datetime.now().strftime('%Y%m%dT%H%M%SZ')}",
            f"DTSTART:{dtstart}",
            f"DTEND:{dtend}",
            f"SUMMARY:{summary}",
            f"LOCATION:{location}",
            f"DESCRIPTION:{description}",
            "END:VEVENT"
        ]

        for line in event_lines:
            ics_lines.append(fold_ics_line(line))

    ics_lines.append("END:VCALENDAR")

    with open(filename, 'w', encoding='utf-8') as f:
        # ICS files strictly require \r\n line endings
        f.write("\r\n".join(ics_lines))


 if __name__ == "__main__":
    # 1. Read the markdown content
    with open(input_file, "r", encoding="utf-8") as file:
        markdown_data = file.read()

    # 2. Extract talks
    all_talks = extract_talks_from_markdown(markdown_data)

    # 3. Filter using LLM API
    my_talks = get_interesting_talks(all_talks)

    # 4. Print simple format to console
    print("\n" + "="*50)
    print("      YOUR PERSONALIZED CONFERENCE ITINERARY")
    print("="*50 + "\n")

    if not my_talks:
        print("No talks matching your interests were found.")
    else:
        for t in my_talks:
            start_dt, end_dt = get_start_end_dt(t)
            time_display = f"{t['time']} (could not calculate duration)"
            if start_dt and end_dt:
                duration = int((end_dt - start_dt).total_seconds() / 60)
                time_display = f"{start_dt.strftime('%I:%M %p')} - {end_dt.strftime('%I:%M %p')} ({duration} min)"

            print(f"📌 {t['title']}")
            print(f"🕒 {t['date']} @ {time_display}")
            print(f"📍 {t['place']}")
            print("-" * 50)

        # 5. Export to .ics
        generate_ics(my_talks, filename=output_file)
        print(f"\n✅ Created '{output_file}' with {len(my_talks)} events.")
        print("   You can double-click this file to import it into your calendar.")
	"""
	Script for going through all talks at MRS Spring 2026 and using AI to evaluate whether
	a talk is interesting or not. The script generates a .ics calendar file.
	This uses the Cerebras Cloud SDK as the AI endpoint, since this was the easiest
	to get running for me. Each run costs ~$0.25.

	This script was mostly written by Gemini 3.1 Pro within a few iterations.

	STEP 1:
	Download abstract book from https://www.mrs.org/docs/default-source/meetings-events/spring-meetings/2026/2026-mrs-spring-meeting-program.pdf

	STEP 2:
	Install PDF-to-markdown converter and run it (takes ~5 minutes):
	pip install markitdown[pdf]
	markitdown mrs.pdf > mrs.md

	STEP 3:
	Install Cerebras Cloud:
	pip install cerebras-cloud-sdk
	Sign up for cloud.cerebras.ai.
	Go to the billing page and select "pay as you go",
	and add $10 of value, otherwise you run out of tokens per minute/day.
	Get an API key (I think it changes when you switch from free tier).

	STEP 4:
	Edit the prompt_intro variable below with your interests.
	The example below finds 235 talks.

	STEP 5:
	Run this script (2-3 minutes) with your API key, which will produce schedule.ics
	CEREBRAS_API_KEY=YOURKEYHERE python mrsscheduler.py

	STEP 6:
	Import schedule.ics in your favorite calendar software
	Make a separate calendar, to make the events easy to
	delete in case your are not happy with the result.
	If using Google Calendar, make a separate MRS calendar,
	then import the events to that calendar.
	(Warning: It will by default add it to your main calendar.)
	"""


	import os
	import re
	import uuid
	import multiprocessing
	from tqdm import tqdm
	from datetime import datetime, timedelta

	from cerebras.cloud.sdk import Cerebras

	client = Cerebras(
	api_key=os.environ.get("CEREBRAS_API_KEY")
	)

	# CONFIGURATION
	num_workers = 8 # within 1k requests/minute rate limit
	input_file = "mrs.md"
	output_file = "schedule.ics"

	prompt_intro = """I am interested methods development and computational aspects of molecular dynamics, atomistic simulations, machine learned interatomic potentials, and high-performance computing. I am not interested in pure application talks without methodological or computational novelty. I am also not interested in machine learning applied to domains that are not related to atomistic simulations. If relevant to ML potentials, I am also interested in computational quantum methods such as DFT."""


	def extract_talks_from_markdown(md_text):
	talks =[]
	md_text = re.sub(r'2026 MRS Spring Meeting\s+Current as of April 15, 2026', '', md_text)

	current_date, current_place, is_poster, current_talk = None, None, False, None
	state = 'SEARCHING'

	session_re = re.compile(r"^SESSION\s+[A-Z0-9.]+:\s+(.*)")
	talk_re = re.compile(r"^(\d{1,2}:\d{2}\s+[AP]M)\s+([+^][A-Z0-9]+\.\d+\s*\.\d+)")
	break_re = re.compile(r"^(\d{1,2}:\d{2}\s+[AP]M)\s+(BREAK\|PANEL DISCUSSION)")
	# Explicit regex to identify the known date bounds of the conference
	date_pattern = re.compile(r"^(Monday\|Tuesday\|Wednesday\|Thursday\|Friday\|Saturday\|Sunday).*?(April\|May)\s+\d{1,2},\s+2026")

	lines = md_text.split('\n')
	i = 0

	def process_header(header_text):
	"""Treats the Title and Authors as one string, just stripping affiliations."""
	header_text = header_text.replace('\n', ' ').strip()
	if ';' in header_text:
	parts = header_text.split(';', 1)
	title_authors = parts[0].strip()
	affiliations = parts[1].strip()
	else:
	title_authors = header_text
	affiliations = ""

	return title_authors, "", affiliations

	while i < len(lines):
	line = lines[i].strip()

	if not line:
	if state == 'HEADER' and current_talk:
	t, a, aff = process_header(current_talk['raw_header'])
	current_talk['title'], current_talk['authors'], current_talk['affiliations'] = t, a, aff
	state = 'ABSTRACT'
	elif state == 'ABSTRACT' and current_talk:
	if current_talk['abstract'] and not current_talk['abstract'].endswith('\n\n'):
	current_talk['abstract'] += '\n\n'
	i += 1
	continue

	match_session = session_re.match(line)
	if match_session:
	if current_talk and not is_poster:
	# Session ended. Current talk is last in session, end time remains None.
	current_talk['abstract'] = current_talk['abstract'].strip()
	talks.append(current_talk)
	current_talk = None
	is_poster = "Poster" in match_session.group(1).strip()

	# Look ahead for a valid date line
	current_date, current_place = None, None
	for j in range(i + 1, min(i + 10, len(lines))):
	lookahead_line = lines[j].strip()
	if date_pattern.match(lookahead_line):
	current_date = lookahead_line
	# The next non-empty, non-event line is generally the location
	for k in range(j + 1, min(j + 5, len(lines))):
	place_line = lines[k].strip()
	if place_line and not talk_re.match(place_line) and not break_re.match(place_line):
	current_place = place_line
	break
	break

	state = 'SEARCHING'
	i += 1
	continue

	match_break = break_re.match(line)
	if match_break:
	time_str, break_type = match_break.groups()
	if current_talk and not is_poster:
	current_talk['end_time_str'] = time_str.strip() # The break signifies the end of the previous talk
	current_talk['abstract'] = current_talk['abstract'].strip()
	talks.append(current_talk)
	current_talk = None
	state = 'SEARCHING'
	i += 1
	continue

	match_talk = talk_re.match(line)
	if match_talk:
	time_str, talk_id = match_talk.groups()

	if current_talk and not is_poster:
	current_talk['end_time_str'] = time_str.strip() # The new talk signifies the end of the previous talk
	current_talk['abstract'] = current_talk['abstract'].strip()
	talks.append(current_talk)

	if not is_poster:
	current_talk = {
	'talk_id': talk_id.strip(),
	'time': time_str.strip(),
	'end_time_str': None, # To be dynamically populated
	'date': current_date,
	'place': current_place,
	'title': '', 'authors': '', 'affiliations': '', 'raw_header': '', 'abstract': ''
	}
	state = 'HEADER'
	else:
	current_talk, state = None, 'SEARCHING'
	i += 1
	continue

	if not is_poster and current_talk:
	if state == 'HEADER': current_talk['raw_header'] += line + " "
	elif state == 'ABSTRACT':
	if line.startswith("SYMPOSIUM BI") or line.startswith("Symposium Organizers"):
	state = 'SEARCHING'
	continue
	if current_talk['abstract'] and not current_talk['abstract'].endswith('\n'):
	current_talk['abstract'] += ' '
	current_talk['abstract'] += line
	i += 1

	if current_talk and not is_poster:
	if state == 'HEADER':
	t, a, aff = process_header(current_talk['raw_header'])
	current_talk['title'], current_talk['authors'], current_talk['affiliations'] = t, a, aff
	current_talk['abstract'] = current_talk['abstract'].strip()
	talks.append(current_talk)

	for t in talks: t.pop('raw_header', None)
	return talks


	def is_interesting(talk):
	"""Evaluate a single talk using AI"""

	prompt_template = prompt_intro + "\n" + """

	Answering only yes or no, do you think I would be interested in the following talk?

	Title and Authors: {title}
	Abstract: {abstract}
	"""

	prompt = prompt_template.format(
	title=talk['title'],
	abstract=talk['abstract']
	)

	try:
	completion = client.chat.completions.create(
	messages=[{"role":"user","content":prompt}],
	model="llama3.1-8b",
	max_completion_tokens=1024,
	temperature=0.2,
	top_p=1,
	stream=False
	)

	answer = completion.choices[0].message.content
	# Clean response: lowercased and stripped of punctuation
	answer = re.sub(r'[^a-z]', '', answer.strip().lower())

	if answer == 'yes':
	return True
	except Exception as e:
	print(f"Error calling API for talk '{talk['title'][:30]}...': {e}")
	return False


	def get_interesting_talks(talks):
	"""Use AI to evaluate each talk based on the user's prompt."""
	interesting =[]

	print(f"Evaluating {len(talks)} talks using LLM...")
	with multiprocessing.Pool(processes=num_workers) as pool:
	results = pool.imap(is_interesting, talks)

	for i, result in enumerate(tqdm(results, total=len(talks))):
	if result:
	interesting.append(talks[i])

	return interesting


	def parse_datetime(date_str, time_str):
	"""Converts the date and time strings into datetime objects."""
	if not date_str or not time_str:
	return None

	try:
	# Extract purely the valid date using regex to avoid artifacts
	match = re.search(r'(April\|May)\s+\d{1,2},\s+2026', date_str)
	if not match:
	return None

	clean_date = match.group(0)
	combined = f"{clean_date} {time_str.strip()}"
	return datetime.strptime(combined, "%B %d, %Y %I:%M %p")
	except Exception as e:
	print(f"Warning: Could not parse datetime ({date_str} {time_str}) - {e}")
	return None


	def get_start_end_dt(talk):
	"""Helper function to calculate the duration of the talk."""
	start_dt = parse_datetime(talk['date'], talk['time'])
	if not start_dt:
	return None, None

	if talk.get('end_time_str'):
	end_dt = parse_datetime(talk['date'], talk['end_time_str'])
	# Safeguard if AM/PM boundary is crossed unpredictably
	if not end_dt or end_dt <= start_dt:
	end_dt = start_dt + timedelta(minutes=15)
	else:
	# Default to 15 mins if it's the last talk before a session ends
	end_dt = start_dt + timedelta(minutes=15)

	return start_dt, end_dt


	def fold_ics_line(line):
	"""Folds lines longer than 75 characters per RFC 5545 specification."""
	folded =[]
	while len(line) > 75:
	folded.append(line[:75])
	line = " " + line[75:]
	folded.append(line)
	return "\r\n".join(folded)


	def generate_ics(talks, filename="schedule.ics"):
	"""Generates an iCalendar (.ics) file from the filtered talks."""
	ics_lines =[
	"BEGIN:VCALENDAR",
	"VERSION:2.0",
	"PRODID:-//User Name//Conference Schedule 1.0//EN",
	"CALSCALE:GREGORIAN"
	]

	for talk in talks:
	start_dt, end_dt = get_start_end_dt(talk)
	if not start_dt or not end_dt:
	continue

	# Datetime formatted for ICS (floating local time)
	dtstart = start_dt.strftime("%Y%m%dT%H%M%S")
	dtend = end_dt.strftime("%Y%m%dT%H%M%S")

	# Escape special characters for valid ICS text using raw strings to prevent SyntaxWarning
	escape = lambda s: str(s).replace('\\', r'\\').replace(';', r'\;').replace(',', r'\,').replace('\n', r'\n')

	summary = escape(talk['title'])
	location = escape(talk['place'])
	description = escape(f"Title/Authors: {talk['title']}\n\n{talk['abstract']}")
	uid = str(uuid.uuid4()) + "@mrsschedule"

	event_lines =[
	"BEGIN:VEVENT",
	f"UID:{uid}",
	f"DTSTAMP:{datetime.now().strftime('%Y%m%dT%H%M%SZ')}",
	f"DTSTART:{dtstart}",
	f"DTEND:{dtend}",
	f"SUMMARY:{summary}",
	f"LOCATION:{location}",
	f"DESCRIPTION:{description}",
	"END:VEVENT"
	]

	for line in event_lines:
	ics_lines.append(fold_ics_line(line))

	ics_lines.append("END:VCALENDAR")

	with open(filename, 'w', encoding='utf-8') as f:
	# ICS files strictly require \r\n line endings
	f.write("\r\n".join(ics_lines))


	if __name__ == "__main__":
	# 1. Read the markdown content
	with open(input_file, "r", encoding="utf-8") as file:
	markdown_data = file.read()

	# 2. Extract talks
	all_talks = extract_talks_from_markdown(markdown_data)

	# 3. Filter using LLM API
	my_talks = get_interesting_talks(all_talks)

	# 4. Print simple format to console
	print("\n" + "="*50)
	print(" YOUR PERSONALIZED CONFERENCE ITINERARY")
	print("="*50 + "\n")

	if not my_talks:
	print("No talks matching your interests were found.")
	else:
	for t in my_talks:
	start_dt, end_dt = get_start_end_dt(t)
	time_display = f"{t['time']} (could not calculate duration)"
	if start_dt and end_dt:
	duration = int((end_dt - start_dt).total_seconds() / 60)
	time_display = f"{start_dt.strftime('%I:%M %p')} - {end_dt.strftime('%I:%M %p')} ({duration} min)"

	print(f"📌 {t['title']}")
	print(f"🕒 {t['date']} @ {time_display}")
	print(f"📍 {t['place']}")
	print("-" * 50)

	# 5. Export to .ics
	generate_ics(my_talks, filename=output_file)
	print(f"\n✅ Created '{output_file}' with {len(my_talks)} events.")
	print(" You can double-click this file to import it into your calendar.")
No results found