TextModel-v1/gci_synth.py

import json
import random

TOPICS = {
    "agriculture": dict(
        tuples=[
            ("a wheat field", "nitrogen runoff from the upper terrace", "soil microbes convert the excess nitrogen into nitrous oxide",
             "the crop canopy yellows and grain yield drops", "kg of nitrogen per hectare", 180),
            ("a dairy herd", "a switch to a higher-protein feed mix", "rumen bacteria shift their fermentation balance",
             "milk yield per cow rises measurably", "grams of crude protein per kilogram of feed", 160),
            ("an orchard", "three consecutive nights below freezing", "cell walls in the blossom tissue rupture as ice forms",
             "the season's fruit set collapses", "consecutive hours below zero Celsius", 5),
        ],
        distractors=[
            "Crop rotation schedules are usually planned a full year in advance.",
            "Most combine harvesters are serviced at the start of the season.",
            "Farmers markets in the region open every Saturday morning.",
            "Irrigation pipes are typically replaced every eight to ten years.",
            "The regional agricultural fair is held in the first week of autumn.",
            "Tractor fuel prices are tracked weekly by the local cooperative.",
            "Seed catalogs are printed and distributed each winter.",
            "The county extension office publishes a soil-testing guide annually.",
        ],
    ),
    "archaeology": dict(
        tuples=[
            ("a coastal midden site", "a drop in sea level exposing a wider tidal flat", "shellfish became easier to harvest in bulk",
             "the shell-deposit layer thickens abruptly in the stratigraphy", "meters of relative sea level change", 2),
            ("a burial mound", "the introduction of bronze tools to the region", "grave goods shift from stone to metal implements",
             "the mound's later strata show a marked wealth increase", "years after the estimated trade contact date", 50),
            ("a hilltop settlement", "a prolonged drought recorded in nearby pollen cores", "residents abandoned the lower fields for terraced ones",
             "terrace wall construction spikes in the same layer", "consecutive dry seasons in the pollen record", 4),
        ],
        distractors=[
            "Radiocarbon dating labs typically report results within a few weeks.",
            "Site excavation permits must be renewed every field season.",
            "Most trowels used on digs are a standard four-inch size.",
            "The museum's ceramics collection is catalogued by region.",
            "Field notebooks are archived at the university at the end of each dig.",
            "Ground-penetrating radar surveys are usually done before excavation begins.",
            "Volunteer diggers are trained during the first two days on site.",
            "The site's screening station uses quarter-inch mesh sieves.",
        ],
    ),
    "architecture": dict(
        tuples=[
            ("a steel-framed tower", "a redesign that widened the base footprint", "wind load transfers more evenly through the lower trusses",
             "the building's sway amplitude drops in monitoring data", "millimeters of lateral sway per floor", 12),
            ("an old masonry chapel", "decades of rising groundwater", "the foundation stone slowly wicks moisture upward",
             "salt crystallization begins spalling the lower courses", "percent moisture content in the stone", 6),
            ("a glass atrium", "the addition of a double-skin facade", "trapped air between the layers buffers temperature swings",
             "interior heating costs fall noticeably each winter", "degrees Celsius of daily temperature swing", 8),
        ],
        distractors=[
            "Blueprints are typically archived digitally after project completion.",
            "Most building codes are revised on a multi-year cycle.",
            "Scaffolding rental rates vary by region and season.",
            "Structural steel is usually delivered pre-cut to the site.",
            "Site inspections are scheduled at fixed construction milestones.",
            "Architectural firms often specialize by building type.",
            "Concrete curing typically takes about four weeks to reach full strength.",
            "Elevator installation is usually one of the last interior trades.",
        ],
    ),
    "astronomy": dict(
        tuples=[
            ("a red giant star", "the exhaustion of core hydrogen fuel", "the core contracts while the outer layers expand and cool",
             "the star's luminosity climbs while its surface temperature falls", "solar radii of expansion", 100),
            ("a binary star system", "tidal forces from the closer companion", "the primary star's rotation gradually locks to the orbit",
             "the system's brightness curve becomes strictly periodic", "days of orbital period", 3),
            ("a protoplanetary disk", "a nearby supernova shockwave passing through", "dust grains compress into denser clumps",
             "planetesimal formation accelerates in the disk's outer edge", "astronomical units from the shock front", 50),
        ],
        distractors=[
            "Most large telescopes are scheduled for observing time months ahead.",
            "Star catalogs are updated periodically as new surveys complete.",
            "Amateur astronomy clubs often meet on the new moon each month.",
            "Satellite launch windows depend on orbital mechanics constraints.",
            "Observatory domes are usually repainted every several years.",
            "Telescope mirrors are re-coated with aluminum periodically.",
            "Astronomical almanacs are published a year in advance.",
            "Radio telescope arrays require precise time synchronization.",
        ],
    ),
    "automotive": dict(
        tuples=[
            ("a turbocharged engine", "prolonged high-load highway driving", "exhaust gas temperatures stay elevated for longer stretches",
             "turbocharger bearing wear accelerates measurably", "degrees Celsius above the rated exhaust temperature", 50),
            ("an electric vehicle battery pack", "repeated fast-charging cycles", "internal cell resistance rises with each cycle",
             "usable range drops noticeably over the following months", "fast-charge cycles logged", 200),
            ("a timing belt", "an extended interval past the service recommendation", "the rubber compound stiffens and micro-cracks form",
             "the belt eventually skips a tooth under load", "kilometers past the recommended interval", 20000),
        ],
        distractors=[
            "Tire rotations are usually recommended every ten thousand kilometers.",
            "Most dealerships offer scheduled maintenance reminders by text.",
            "Car paint colors are mixed to match a standard code system.",
            "Windshield wiper blades are typically replaced once a year.",
            "Vehicle registration renewals are handled annually in most regions.",
            "Cabin air filters are usually swapped during routine servicing.",
            "Most modern cars log diagnostic codes automatically.",
            "Showroom floor models are rotated seasonally.",
        ],
    ),
    "aviation": dict(
        tuples=[
            ("a commercial airliner wing", "repeated pressurization cycles over its service life", "microscopic fatigue cracks accumulate near rivet holes",
             "scheduled inspections start flagging crack growth", "pressurization cycles logged", 20000),
            ("a turboprop engine", "prolonged operation in dusty air", "sand particles erode the compressor blade edges",
             "fuel efficiency drops across the fleet's logbooks", "flight hours in high-dust conditions", 500),
            ("a glider", "a thermal column strengthening over open ground", "rising warm air lifts the aircraft faster than sink rate",
             "the pilot gains several hundred meters of altitude quickly", "meters per second of climb rate", 3),
        ],
        distractors=[
            "Airport runway markings are repainted on a routine schedule.",
            "Most airlines rotate cabin crew assignments monthly.",
            "Boarding passes are typically issued starting ninety minutes before departure.",
            "Aircraft livery redesigns happen every several years.",
            "Baggage handling systems are inspected regularly by ground staff.",
            "Flight simulators are recalibrated on a fixed maintenance cycle.",
            "Air traffic control shifts typically rotate every eight hours.",
            "Most airports publish noise abatement procedures publicly.",
        ],
    ),
    "chemistry": dict(
        tuples=[
            ("a buffered solution", "the slow addition of a strong acid", "the buffer's conjugate base is gradually consumed",
             "the pH begins to drop sharply past the buffer capacity", "moles of acid added per liter", 0.5),
            ("a polymer sample", "prolonged exposure to ultraviolet light", "chain scission breaks the polymer backbone",
             "the material becomes visibly brittle within weeks", "hours of UV exposure", 200),
            ("a metal alloy", "repeated thermal cycling in a furnace", "microstructural grain boundaries migrate and coarsen",
             "the alloy's tensile strength falls measurably", "thermal cycles applied", 100),
        ],
        distractors=[
            "Laboratory glassware is calibrated according to a fixed standard.",
            "Most reagent bottles include a manufacture and expiry date.",
            "Fume hood airflow is checked on a routine safety schedule.",
            "Chemical inventories are typically audited once a year.",
            "Titration equipment is rinsed with distilled water between trials.",
            "Safety data sheets are updated when formulations change.",
            "Most balances are recalibrated with certified reference weights.",
            "Waste solvent is collected in labeled containers by class.",
        ],
    ),
    "cooking": dict(
        tuples=[
            ("a pot of caramelizing sugar", "the temperature crossing into the amber range", "sucrose molecules break down into hundreds of new flavor compounds",
             "the sugar rapidly darkens and turns bitter if left further", "degrees Celsius on the sugar thermometer", 170),
            ("a sourdough starter", "several days of regular feeding at room temperature", "wild yeast and lactobacilli populations stabilize",
             "the dough begins to reliably double in size", "consecutive days of feeding", 5),
            ("a batch of bread dough", "over-proofing past the recommended window", "the gluten network stretches until it can no longer hold gas",
             "the loaf collapses instead of rising in the oven", "minutes past the recommended proof time", 45),
        ],
        distractors=[
            "Most home ovens run a few degrees off their dial setting.",
            "Recipe books are often organized by course or season.",
            "Kitchen knives should be sharpened on a regular schedule.",
            "Farmers markets tend to have the freshest produce early in the day.",
            "Most stand mixers include several interchangeable attachments.",
            "Spice jars are typically restocked every few months in a busy kitchen.",
            "Cast iron pans need occasional reseasoning to stay non-stick.",
            "Restaurant menus are usually revised on a seasonal basis.",
        ],
    ),
    "energy": dict(
        tuples=[
            ("a wind farm", "a sustained increase in average wind speed", "turbine blades pass into a higher-efficiency operating range",
             "the farm's total output climbs sharply that week", "meters per second of average wind speed", 12),
            ("a lithium-ion grid battery", "back-to-back deep discharge cycles", "internal electrode stress accumulates with each cycle",
             "the battery's usable capacity fades measurably", "deep discharge cycles logged", 300),
            ("a solar array", "a thin layer of dust accumulating on the panels", "less sunlight reaches the photovoltaic cells",
             "the array's daily output drops steadily until cleaning", "days since the last panel cleaning", 30),
        ],
        distractors=[
            "Utility meters are typically read on a monthly cycle.",
            "Most power plants schedule maintenance outages seasonally.",
            "Energy tariffs are usually reviewed by regulators annually.",
            "Transmission towers are inspected on a routine rotation.",
            "Grid operators publish demand forecasts a day ahead.",
            "Substation transformers are tested periodically for insulation health.",
            "Most utilities offer time-of-use billing plans.",
            "Renewable energy certificates are tracked in a public registry.",
        ],
    ),
    "finance": dict(
        tuples=[
            ("a corporate bond fund", "a central bank rate hike announcement", "existing bond prices adjust downward to match new yields",
             "the fund's net asset value drops within the trading day", "basis points of rate increase", 50),
            ("a retail brokerage", "a surge in first-time account signups", "order flow volume exceeds the matching engine's normal capacity",
             "trade execution latency spikes during peak hours", "thousands of new accounts opened", 50),
            ("a commodity futures desk", "a supply disruption reported upstream", "traders rapidly reprice the expected scarcity",
             "the futures contract gaps up at the next open", "percent reported supply shortfall", 10),
        ],
        distractors=[
            "Most stock exchanges publish a holiday trading calendar yearly.",
            "Quarterly earnings reports are typically released on a set schedule.",
            "Bank branch hours vary somewhat by location.",
            "Most brokerages send account statements on a monthly basis.",
            "Credit rating agencies review issuers periodically.",
            "Tax filing deadlines are set annually by the relevant authority.",
            "Financial regulators publish guidance documents from time to time.",
            "Most index funds rebalance their holdings quarterly.",
        ],
    ),
    "gardening": dict(
        tuples=[
            ("a raised vegetable bed", "a stretch of days above the seedlings' heat tolerance", "leaf stomata close to conserve water",
             "growth visibly stalls until temperatures ease", "consecutive days above the tolerance threshold", 3),
            ("a rose bush", "a sudden increase in humid, still air", "fungal spores find ideal conditions on the leaf surface",
             "black spot lesions spread across the foliage within days", "percent relative humidity sustained", 85),
            ("a compost pile", "the pile crossing its optimal carbon-to-nitrogen ratio", "thermophilic bacteria multiply rapidly",
             "the pile's internal temperature rises sharply", "carbon-to-nitrogen ratio by weight", 30),
        ],
        distractors=[
            "Garden hoses are usually stored indoors over winter.",
            "Most seed packets list a recommended planting depth.",
            "Compost bins come in a variety of standard sizes.",
            "Pruning shears should be cleaned between uses on different plants.",
            "Garden centers restock seasonal plants on a rotating basis.",
            "Mulch is typically reapplied once or twice a year.",
            "Most watering cans hold around two gallons.",
            "Raised beds are commonly built from untreated cedar.",
        ],
    ),
    "hardware": dict(
        tuples=[
            ("a data center rack", "sustained load pushing CPUs past their thermal design point", "cooling fans ramp to maximum and can no longer keep pace",
             "the rack's servers begin throttling clock speed", "degrees Celsius above the thermal design point", 15),
            ("a solid-state drive", "a high volume of small random writes", "flash cell wear accumulates unevenly across blocks",
             "write latency climbs noticeably as the drive ages", "terabytes written past the rated endurance", 100),
            ("a mechanical keyboard switch", "tens of thousands of actuations", "the internal spring gradually loses tension",
             "the keypress force needed drops noticeably", "thousands of actuations logged", 50),
        ],
        distractors=[
            "Most warehouses stock components in labeled bins by part number.",
            "Firmware updates are typically released on an irregular schedule.",
            "Server racks are usually bolted to the floor for stability.",
            "Cable management trays come in several standard widths.",
            "Most retailers offer a standard return window on hardware.",
            "Component datasheets are published by the manufacturer.",
            "Static-safe packaging is standard for sensitive electronics.",
            "Warehouse inventory counts are usually done quarterly.",
        ],
    ),
    "marine": dict(
        tuples=[
            ("a coral reef", "a sustained rise in sea surface temperature", "the coral expels its symbiotic algae under heat stress",
             "large sections of the reef turn visibly white within weeks", "degrees Celsius above the seasonal average", 2),
            ("a container ship's hull", "months of continuous operation without dry-docking", "barnacles and algae accumulate along the waterline",
             "fuel consumption per nautical mile rises steadily", "months since the last hull cleaning", 8),
            ("a coastal fishery", "a multi-year drop in plankton density", "juvenile fish find less food during their critical growth window",
             "the following season's catch reports decline sharply", "percent drop in plankton density", 30),
        ],
        distractors=[
            "Most harbors schedule dredging on a multi-year cycle.",
            "Fishing licenses are typically renewed annually.",
            "Cargo ships file manifests before entering port.",
            "Marine weather buoys report data at regular intervals.",
            "Lighthouse maintenance is usually handled by a coastal authority.",
            "Most ports have a designated pilot service for large vessels.",
            "Nautical charts are updated periodically by hydrographic offices.",
            "Ferry schedules are adjusted seasonally in most regions.",
        ],
    ),
    "music": dict(
        tuples=[
            ("a violin's spruce top", "a sustained drop in the practice room's humidity", "the wood fibers contract unevenly across the grain",
             "the instrument's tone becomes noticeably brighter and thinner", "percent relative humidity drop", 20),
            ("a concert hall", "an increase in the audience's seat occupancy", "sound absorption from clothing and bodies rises",
             "the room's reverberation time shortens measurably", "percent of seats occupied", 80),
            ("a guitar string", "hundreds of hours of playing", "metal fatigue builds at the string's contact points",
             "intonation drifts noticeably higher up the neck", "hours of cumulative playing time", 100),
        ],
        distractors=[
            "Most concert venues sell tickets starting months in advance.",
            "Instrument cases are typically lined with soft padding.",
            "Sheet music publishers release new arrangements regularly.",
            "Orchestras usually rehearse on a fixed weekly schedule.",
            "Most recording studios book sessions by the hour.",
            "Music festivals are often scheduled around the summer season.",
            "Piano tuners typically visit on a seasonal basis.",
            "Streaming platforms update their charts weekly.",
        ],
    ),
    "photography": dict(
        tuples=[
            ("a camera sensor", "a long exposure in warm ambient conditions", "thermal noise accumulates across the sensor's photosites",
             "grain becomes visibly obvious in the shadow regions", "seconds of exposure time", 30),
            ("a batch of darkroom prints", "developer chemistry used past its stated capacity", "silver halide reduction slows unevenly across the tray",
             "the later prints come out visibly flatter in contrast", "prints processed past rated capacity", 20),
            ("a telephoto lens", "repeated rapid temperature changes between indoor and outdoor shoots", "internal condensation forms between the glass elements",
             "autofocus accuracy degrades until the lens fully equalizes", "degrees Celsius of temperature swing", 15),
        ],
        distractors=[
            "Most camera bodies receive firmware updates periodically.",
            "Photography clubs often meet monthly to review member work.",
            "Memory cards should be formatted in-camera before a shoot.",
            "Studio lighting kits usually include several modifiers.",
            "Print labs typically offer same-day service for standard sizes.",
            "Lens caps come in a range of standard thread sizes.",
            "Most galleries rotate their exhibited work seasonally.",
            "Camera straps are usually sold separately from the body.",
        ],
    ),
    "physiology": dict(
        tuples=[
            ("a marathon runner's muscles", "the glycogen stores dropping past a critical level", "the body shifts toward burning fat for fuel less efficiently",
             "pace drops sharply in the final miles", "percent of glycogen stores remaining", 20),
            ("the human eye", "a sudden move from bright sunlight into a dark room", "rod cells slowly regenerate their light-sensitive pigment",
             "night vision gradually improves over several minutes", "minutes spent in the dark", 10),
            ("a person's core temperature", "prolonged exposure to cold water", "peripheral blood vessels constrict to preserve heat",
             "shivering intensifies sharply once core temperature drops further", "degrees Celsius drop in core temperature", 2),
        ],
        distractors=[
            "Most clinics schedule routine checkups on an annual basis.",
            "Blood donation centers typically operate by appointment.",
            "Standard hospital shifts often run twelve hours.",
            "Vaccination schedules are usually published by health authorities.",
            "Most gyms offer orientation sessions for new members.",
            "Medical textbooks are revised every several years.",
            "Physical therapy sessions are commonly booked weekly.",
            "Hospitals typically restock supplies on a set schedule.",
        ],
    ),
    "sports": dict(
        tuples=[
            ("a marathon field of runners", "an unexpected rise in race-day temperature", "the body's cooling demand outpaces sweat evaporation",
             "average finishing times slow noticeably across the field", "degrees Celsius above the forecast temperature", 8),
            ("a basketball team's shooting", "a long losing streak affecting practice attendance", "shot repetition drops below the usual weekly volume",
             "free-throw percentage declines over the following games", "percent drop in practice attendance", 30),
            ("a cyclist's legs", "repeated high-intensity interval sessions", "muscle fibers adapt by increasing capillary density",
             "sustained power output improves over the following weeks", "interval sessions completed", 12),
        ],
        distractors=[
            "Most stadiums are inspected before the season opener.",
            "Ticket prices for major events often rise closer to game day.",
            "League schedules are typically published months in advance.",
            "Team equipment is usually replaced on a yearly cycle.",
            "Most broadcasters rotate commentary teams by season.",
            "Referee certifications are renewed periodically.",
            "Youth leagues commonly practice twice a week.",
            "Sports medicine clinics often see a rise in visits after tournaments.",
        ],
    ),
    "textiles": dict(
        tuples=[
            ("a wool sweater", "repeated washing in hot water", "the fiber's scale structure locks together under agitation",
             "the garment shrinks noticeably after a handful of washes", "hot washes completed", 5),
            ("a length of dyed cotton", "extended exposure to direct sunlight", "UV energy breaks down the dye's chromophore bonds",
             "the fabric visibly fades within a season outdoors", "weeks of continuous sun exposure", 8),
            ("a synthetic rope", "sustained load near its rated breaking strength", "polymer chains slowly creep and elongate under tension",
             "the rope's diameter thins measurably before failure", "percent of rated breaking strength sustained", 80),
        ],
        distractors=[
            "Most fabric mills run on a standard weekday production schedule.",
            "Sewing machines require periodic oiling for smooth operation.",
            "Textile trade shows are usually held twice a year.",
            "Dye lots can vary slightly between production batches.",
            "Most tailors take measurements at the first fitting.",
            "Fabric bolts are typically sold by the meter or yard.",
            "Washing labels are sewn in during final garment assembly.",
            "Cotton harvests are usually processed within weeks of picking.",
        ],
    ),
    "weather": dict(
        tuples=[
            ("a coastal town", "a sustained drop in atmospheric pressure offshore", "warm moist air rises rapidly ahead of the front",
             "rainfall rates spike sharply within the hour", "hectopascals of pressure drop", 10),
            ("a mountain valley", "cold air pooling overnight in the low ground", "a temperature inversion traps surface pollutants",
             "morning visibility drops sharply until the sun burns it off", "degrees Celsius of inversion strength", 6),
            ("a wheat-growing region", "several consecutive weeks without meaningful rainfall", "topsoil moisture falls below the crop's wilting point",
             "regional yield forecasts are revised downward", "consecutive weeks without rain", 6),
        ],
        distractors=[
            "Weather stations typically report data every ten minutes.",
            "Most forecasting models are updated several times a day.",
            "Meteorologists often specialize in a particular region.",
            "Weather balloons are usually launched twice daily.",
            "Storm shutters are commonly installed before the season starts.",
            "Local news broadcasts typically include a weather segment nightly.",
            "Rain gauges need periodic calibration to stay accurate.",
            "Climate records are archived by national weather agencies.",
        ],
    ),
    "wildlife": dict(
        tuples=[
            ("a migratory bird population", "an early snowmelt shifting the season's timing", "insect hatches occur before the birds arrive to feed on them",
             "chick survival rates drop that breeding season", "days earlier than the historical snowmelt date", 12),
            ("a deer herd", "a sharp increase in a nearby predator's population", "deer shift toward denser cover during daylight hours",
             "browsing damage patterns move deeper into the forest interior", "percent increase in predator sightings", 40),
            ("a coral-reef fish population", "a marine heatwave lasting several weeks", "the fish's usual prey species retreat to deeper water",
             "the population's average body condition declines that season", "weeks of sustained heatwave conditions", 6),
        ],
        distractors=[
            "Wildlife surveys are typically conducted on a seasonal schedule.",
            "Most nature reserves restrict access during nesting season.",
            "Camera traps are usually checked every few weeks.",
            "Conservation permits are renewed on an annual basis.",
            "Field biologists often tag animals for long-term tracking.",
            "National parks publish visitor guidelines each season.",
            "Most zoos rotate enrichment activities for their animals.",
            "Migration routes are mapped using satellite tracking data.",
        ],
    ),
}

TEMPLATES = {
    "A": dict(
        label="Threshold-based change",
        opening=[
            "In {subject}, {driver} began building up over time.",
            "{subject_cap} experienced a steady buildup driven by {driver}.",
        ],
        middle=[
            "As the buildup continued, {mechanism}.",
            "Over the following period, {mechanism}.",
            "This kept accumulating until {mechanism}.",
        ],
        closing=[
            "Once the value passed roughly {threshold} {unit}, {outcome}.",
            "After crossing about {threshold} {unit}, {outcome}.",
        ],
    ),
    "B": dict(
        label="Comparative experiments",
        opening=[
            "Researchers compared {subject} under two conditions after introducing {driver}.",
            "A side-by-side trial on {subject} isolated the effect of {driver}.",
        ],
        middle=[
            "In the treated condition, {mechanism}.",
            "Compared to the control group, {mechanism}.",
            "The treated samples showed that {mechanism}.",
        ],
        closing=[
            "By the time the difference reached about {threshold} {unit}, {outcome}.",
            "Once the measured gap hit roughly {threshold} {unit}, {outcome}.",
        ],
    ),
    "C": dict(
        label="Cumulative exposure",
        opening=[
            "{subject_cap} was subjected to repeated exposure to {driver}.",
            "Over many repeated cycles, {subject} was affected by {driver}.",
        ],
        middle=[
            "With each additional cycle, {mechanism}.",
            "The cumulative effect meant that {mechanism}.",
            "Repetition gradually meant that {mechanism}.",
        ],
        closing=[
            "After accumulating about {threshold} {unit}, {outcome}.",
            "Once the cumulative total reached {threshold} {unit}, {outcome}.",
        ],
    ),
    "D": dict(
        label="Sudden spike events",
        opening=[
            "A sudden spike in {driver} hit {subject} without much warning.",
            "{subject_cap} was hit by an abrupt spike tied to {driver}.",
        ],
        middle=[
            "In the immediate aftermath, {mechanism}.",
            "Within a short window, {mechanism}.",
            "The sudden change meant that {mechanism}.",
        ],
        closing=[
            "Since the spike exceeded {threshold} {unit}, {outcome}.",
            "Because the spike passed {threshold} {unit}, {outcome}.",
        ],
    ),
    "E": dict(
        label="Threshold-crossing dynamics",
        opening=[
            "{subject_cap} was tracked closely as {driver} slowly changed.",
            "Monitoring of {subject} focused on the slow trend in {driver}.",
        ],
        middle=[
            "As the trend continued, {mechanism}.",
            "Approaching the critical point, {mechanism}.",
            "Just before the crossing, {mechanism}.",
        ],
        closing=[
            "The moment it crossed {threshold} {unit}, {outcome}.",
            "Right after crossing {threshold} {unit}, {outcome}.",
        ],
    ),
    "F": dict(
        label="Feedback loops",
        opening=[
            "A feedback loop formed in {subject} once {driver} took hold.",
            "{subject_cap} developed a self-reinforcing loop starting from {driver}.",
        ],
        middle=[
            "Each pass through the loop meant that {mechanism}.",
            "The loop reinforced itself as {mechanism}.",
            "With every iteration, {mechanism}.",
        ],
        closing=[
            "Once the loop's strength passed {threshold} {unit}, {outcome}.",
            "After the feedback exceeded {threshold} {unit}, {outcome}.",
        ],
    ),
}

DIFFICULTIES = {"easy": 2, "medium": 3, "hard": 4}

QUESTION_FORMS = [
    "What was the underlying cause behind the change described for {subject}?",
    "Which factor most directly explains the outcome observed in {subject}?",
    "According to the passage, what triggered the eventual change in {subject}?",
]


def _cap(s):
    return s[0].upper() + s[1:]


def _jitter(value):
    if isinstance(value, float):
        return round(value * random.uniform(0.85, 1.15), 2)
    return max(1, int(round(value * random.uniform(0.85, 1.15))))


def _render_chain(topic, template_id, n_edges, rng):
    tpl = TEMPLATES[template_id]
    subj, driver, mech, outcome, unit, threshold = rng.choice(TOPICS[topic]["tuples"])
    threshold = _jitter(threshold)
    slots = dict(subject=subj, subject_cap=_cap(subj), driver=driver, mechanism=mech,
                 outcome=outcome, unit=unit, threshold=threshold)

    n_nodes = n_edges + 1
    sentences = [rng.choice(tpl["opening"]).format(**slots)]
    for _ in range(n_nodes - 2):
        sentences.append(rng.choice(tpl["middle"]).format(**slots))
    sentences.append(rng.choice(tpl["closing"]).format(**slots))
    return sentences, slots


def generate_item(item_id, rng):
    topic = rng.choice(list(TOPICS.keys()))
    template_id = rng.choice(list(TEMPLATES.keys()))
    difficulty = rng.choice(list(DIFFICULTIES.keys()))
    n_edges = DIFFICULTIES[difficulty]

    related_sentences, slots = _render_chain(topic, template_id, n_edges, rng)
    n_related = len(related_sentences)

    bank = list(TOPICS[topic]["distractors"])
    rng.shuffle(bank)
    distractor_sentences = bank[:n_related] if len(bank) >= n_related else (bank * 2)[:n_related]

    # (kind, text, chain_order) -- chain_order is the sentence's position in the
    # causal chain (0..n_related-1) for related sentences, None for distractors.
    # Tracked explicitly rather than re-derived by text lookup, since two related
    # sentences can render identical text (a "middle" phrasing can repeat when a
    # long chain draws the same template variant twice).
    seg_specs = []
    for order, s in enumerate(related_sentences):
        seg_specs.append(("related", s, order))
    for s in distractor_sentences:
        seg_specs.append(("distractor", s, None))
    rng.shuffle(seg_specs)

    segments = []
    context_parts = []
    cursor = 0
    related_ids, unrelated_ids = [], []
    related_seg_id_in_order = []
    for i, (kind, text, chain_order) in enumerate(seg_specs):
        if context_parts:
            cursor += 1  # joining space
        start = cursor
        end = start + len(text)
        seg_id = i
        segments.append({"id": seg_id, "charStart": start, "charEnd": end})
        if kind == "related":
            related_ids.append(seg_id)
            related_seg_id_in_order.append((chain_order, seg_id))
        else:
            unrelated_ids.append(seg_id)
        context_parts.append(text)
        cursor = end

    context = " ".join(context_parts)

    related_seg_id_in_order.sort()
    ordered_related_ids = [seg_id for _, seg_id in related_seg_id_in_order]
    key_link_pairs = [[ordered_related_ids[i], ordered_related_ids[i + 1]] for i in range(len(ordered_related_ids) - 1)]

    question = rng.choice(QUESTION_FORMS).format(subject=slots["subject"])
    reference_answer = f"{_cap(slots['driver'])}, via {slots['mechanism']}, led to {slots['outcome']}."

    return {
        "id": f"synth-{item_id:06d}",
        "topic": topic,
        "topicLabel": topic.capitalize(),
        "templateId": template_id,
        "difficulty": difficulty,
        "question": question,
        "referenceAnswer": reference_answer,
        "context": context,
        "segments": segments,
        "relatedSegmentIds": related_ids,
        "unrelatedSegmentIds": unrelated_ids,
        "keyLinkPairs": key_link_pairs,
        "meta": {
            "subject": slots["subject"], "driver": slots["driver"], "mechanism": slots["mechanism"],
            "outcome": slots["outcome"], "threshold": slots["threshold"], "unit": slots["unit"],
        },
    }


def generate_dataset(n, seed=0):
    rng = random.Random(seed)
    return [generate_item(i, rng) for i in range(n)]


if __name__ == "__main__":
    import sys
    n = int(sys.argv[1]) if len(sys.argv) > 1 else 10
    items = generate_dataset(n, seed=42)
    for it in items[:3]:
        print(json.dumps(it, indent=2))
    print(f"\ngenerated {len(items)} items")