File size: 12,428 Bytes
67f0e56 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 | #!/usr/bin/env python3
"""Generate emotion-labeled stories using Gemini API.
171 emotions x 100 topics x 10 stories = 171,000 stories.
Concurrent API calls (up to 100), SQLite WAL for storage,
saves both raw API output and parsed stories.
Run:
python -m full_replication.generate_stories
python -m full_replication.generate_stories --test
python -m full_replication.generate_stories --workers 50
"""
import argparse
import os
import re
import sqlite3
import threading
import time
from concurrent.futures import ThreadPoolExecutor, as_completed
from dotenv import load_dotenv
from google import genai
from google.genai import types
from tqdm import tqdm
from full_replication.config import EMOTIONS, TOPICS, STORY_PROMPT
load_dotenv(os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), ".env"))
DB_PATH = os.path.join(os.path.dirname(os.path.abspath(__file__)), "data", "stories.db")
MODEL = "gemini-2.0-flash-lite"
STORIES_PER_CALL = 10
# Appended to Anthropic's prompt to enforce strict output format
FORMAT_SUFFIX = """
OUTPUT FORMAT: Start directly with [story 1] — no preamble, no introductions, no explanations, no commentary. Output ONLY the stories separated by [story N] markers. Nothing else."""
# Thread-local storage for DB connections
_local = threading.local()
_db_lock = threading.Lock()
def get_db():
"""Get thread-local DB connection with WAL mode."""
if not hasattr(_local, "conn"):
_local.conn = sqlite3.connect(DB_PATH, timeout=30)
_local.conn.execute("PRAGMA journal_mode=WAL")
_local.conn.execute("PRAGMA busy_timeout=10000")
return _local.conn
def init_db():
"""Create tables if they don't exist."""
os.makedirs(os.path.dirname(DB_PATH), exist_ok=True)
conn = sqlite3.connect(DB_PATH)
conn.execute("PRAGMA journal_mode=WAL")
conn.executescript("""
CREATE TABLE IF NOT EXISTS api_calls (
id INTEGER PRIMARY KEY AUTOINCREMENT,
emotion TEXT NOT NULL,
topic_idx INTEGER NOT NULL,
topic TEXT NOT NULL,
raw_response TEXT,
status TEXT DEFAULT 'pending',
error TEXT,
created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
UNIQUE(emotion, topic_idx)
);
CREATE TABLE IF NOT EXISTS stories (
id INTEGER PRIMARY KEY AUTOINCREMENT,
api_call_id INTEGER NOT NULL,
emotion TEXT NOT NULL,
topic_idx INTEGER NOT NULL,
topic TEXT NOT NULL,
story_idx INTEGER NOT NULL,
text TEXT NOT NULL,
created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
FOREIGN KEY (api_call_id) REFERENCES api_calls(id),
UNIQUE(emotion, topic_idx, story_idx)
);
CREATE INDEX IF NOT EXISTS idx_stories_emotion ON stories(emotion);
CREATE INDEX IF NOT EXISTS idx_api_calls_status ON api_calls(status);
CREATE TABLE IF NOT EXISTS stories_clean (
id INTEGER PRIMARY KEY AUTOINCREMENT,
api_call_id INTEGER NOT NULL,
emotion TEXT NOT NULL,
topic_idx INTEGER NOT NULL,
topic TEXT NOT NULL,
story_idx INTEGER NOT NULL,
text TEXT NOT NULL,
created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
FOREIGN KEY (api_call_id) REFERENCES api_calls(id),
UNIQUE(emotion, topic_idx, story_idx)
);
CREATE INDEX IF NOT EXISTS idx_stories_clean_emotion ON stories_clean(emotion);
""")
conn.commit()
conn.close()
def get_completed():
"""Return set of (emotion, topic_idx) already in stories_clean."""
conn = sqlite3.connect(DB_PATH, timeout=30)
conn.execute("PRAGMA journal_mode=WAL")
rows = conn.execute(
"SELECT DISTINCT emotion, topic_idx FROM stories_clean"
).fetchall()
conn.close()
return set(rows)
_PREAMBLE_RE = re.compile(
r'^(Here\s+are|Here\s+is|Below\s+are|These\s+are|The\s+following|I\'ve\s+written|Sure|Okay)',
re.IGNORECASE
)
def is_preamble(text):
"""Check if text is model preamble rather than an actual story."""
return bool(_PREAMBLE_RE.match(text.strip()))
def clean_story(text):
"""Strip leading markdown bold/headers and trailing junk."""
text = text.strip()
# Remove leading **Title** or ## Title lines
text = re.sub(r'^(?:\*\*[^*]+\*\*|#{1,3}\s+.+)\s*\n', '', text).strip()
return text
def parse_stories(text, expected_count=10):
"""Parse model output into individual stories."""
min_stories = max(2, expected_count // 2)
# Strategy 1: [story N] markers
parts = re.split(r'\[story\s*\d+\]', text, flags=re.IGNORECASE)
parts = [p.strip() for p in parts if p.strip() and len(p.strip()) > 50]
if len(parts) >= min_stories:
return parts
# Strategy 2: Numbered patterns
parts = re.split(r'(?:^|\n)\s*(?:\*{0,2}(?:Story\s+)?\d+[\.\):\*]{1,3}\s*\*{0,2})', text, flags=re.IGNORECASE)
parts = [p.strip() for p in parts if p.strip() and len(p.strip()) > 50]
if len(parts) >= min_stories:
return parts
# Strategy 3: Double newline separation
parts = re.split(r'\n\s*\n', text)
parts = [p.strip() for p in parts if p.strip() and len(p.strip()) > 50]
if len(parts) >= min_stories:
return parts
# Fallback
if len(text.strip()) > 100:
return [text.strip()]
return []
def generate_one(client, emotion, topic_idx, topic):
"""Generate stories for one emotion x topic, save to DB."""
prompt = STORY_PROMPT.format(
n_stories=STORIES_PER_CALL,
topic=topic,
emotion=emotion,
) + FORMAT_SUFFIX
db = get_db()
raw_response = None
error = None
status = "error"
try:
response = client.models.generate_content(
model=MODEL,
contents=prompt,
config=types.GenerateContentConfig(
temperature=0.9,
top_p=0.95,
top_k=64,
max_output_tokens=4096,
),
)
raw_response = response.text
stories = parse_stories(raw_response, STORIES_PER_CALL)
if not stories:
error = "no stories parsed"
status = "error"
else:
status = "done"
except Exception as e:
error = str(e)[:500]
stories = []
# Save to DB
with _db_lock:
try:
cursor = db.execute(
"""INSERT OR REPLACE INTO api_calls
(emotion, topic_idx, topic, raw_response, status, error)
VALUES (?, ?, ?, ?, ?, ?)""",
(emotion, topic_idx, topic, raw_response, status, error),
)
api_call_id = cursor.lastrowid
for i, story_text in enumerate(stories):
db.execute(
"""INSERT OR REPLACE INTO stories
(api_call_id, emotion, topic_idx, topic, story_idx, text)
VALUES (?, ?, ?, ?, ?, ?)""",
(api_call_id, emotion, topic_idx, topic, i, story_text),
)
# Write clean versions (skip preamble, clean formatting)
clean_idx = 0
for story_text in stories:
if is_preamble(story_text):
continue
cleaned = clean_story(story_text)
if len(cleaned) > 50:
db.execute(
"""INSERT OR REPLACE INTO stories_clean
(api_call_id, emotion, topic_idx, topic, story_idx, text)
VALUES (?, ?, ?, ?, ?, ?)""",
(api_call_id, emotion, topic_idx, topic, clean_idx, cleaned),
)
clean_idx += 1
db.commit()
except Exception as e:
db.rollback()
error = str(e)[:500]
return {
"emotion": emotion,
"topic_idx": topic_idx,
"n_stories": len(stories),
"status": status,
"error": error,
}
def backfill_clean():
"""Re-parse all existing stories into stories_clean table."""
conn = sqlite3.connect(DB_PATH, timeout=30)
conn.execute("PRAGMA journal_mode=WAL")
# Get all api_calls that are done
calls = conn.execute(
"SELECT id, emotion, topic_idx, topic FROM api_calls WHERE status = 'done'"
).fetchall()
cleaned_total = 0
skipped_total = 0
for api_call_id, emotion, topic_idx, topic in calls:
rows = conn.execute(
"SELECT story_idx, text FROM stories WHERE api_call_id = ? ORDER BY story_idx",
(api_call_id,)
).fetchall()
clean_idx = 0
for _, story_text in rows:
if is_preamble(story_text):
skipped_total += 1
continue
cleaned = clean_story(story_text)
if len(cleaned) > 50:
conn.execute(
"""INSERT OR REPLACE INTO stories_clean
(api_call_id, emotion, topic_idx, topic, story_idx, text)
VALUES (?, ?, ?, ?, ?, ?)""",
(api_call_id, emotion, topic_idx, topic, clean_idx, cleaned),
)
clean_idx += 1
cleaned_total += 1
conn.commit()
print(f"Backfill complete: {cleaned_total} clean stories, {skipped_total} preambles skipped")
conn.close()
def main():
parser = argparse.ArgumentParser()
parser.add_argument("--test", action="store_true", help="Single call test")
parser.add_argument("--workers", type=int, default=100, help="Concurrent workers")
parser.add_argument("--backfill", action="store_true", help="Backfill stories_clean from existing stories")
args = parser.parse_args()
init_db()
if args.backfill:
backfill_clean()
return
api_key = os.environ.get("GEMINI_API_KEY")
if not api_key:
print("ERROR: GEMINI_API_KEY not found in .env")
return
completed = get_completed()
# Build work queue
tasks = []
for emotion in EMOTIONS:
topics = TOPICS
for ti, topic in enumerate(topics):
if (emotion, ti) not in completed:
tasks.append((emotion, ti, topic))
total = len(EMOTIONS) * len(TOPICS)
done = total - len(tasks)
if args.test:
tasks = tasks[:1]
print(f"TEST MODE: 1 call only")
print(f"=== Story Generation (Gemini API) ===")
print(f"Total: {total} calls ({STORIES_PER_CALL} stories each)")
print(f"Done: {done}, Remaining: {len(tasks)}")
print(f"Workers: {min(args.workers, len(tasks))}")
if not tasks:
print("All stories already generated.")
return
client = genai.Client(api_key=api_key)
errors = 0
total_stories = 0
workers = min(args.workers, len(tasks))
with ThreadPoolExecutor(max_workers=workers) as executor:
futures = {
executor.submit(generate_one, client, emotion, ti, topic): (emotion, ti)
for emotion, ti, topic in tasks
}
with tqdm(total=len(tasks), desc="Generating", unit="call") as pbar:
for future in as_completed(futures):
result = future.result()
total_stories += result["n_stories"]
if result["status"] == "error":
errors += 1
pbar.update(1)
pbar.set_postfix(
stories=total_stories,
errors=errors,
rate=f"{total_stories/(pbar.n or 1)*STORIES_PER_CALL:.0f}/call"
)
# Summary
conn = sqlite3.connect(DB_PATH, timeout=30)
total_stories_db = conn.execute("SELECT COUNT(*) FROM stories").fetchone()[0]
total_clean_db = conn.execute("SELECT COUNT(*) FROM stories_clean").fetchone()[0]
total_calls_done = conn.execute("SELECT COUNT(*) FROM api_calls WHERE status='done'").fetchone()[0]
total_errors = conn.execute("SELECT COUNT(*) FROM api_calls WHERE status='error'").fetchone()[0]
conn.close()
print(f"\n=== COMPLETE ===")
print(f"API calls: {total_calls_done} done, {total_errors} errors")
print(f"Stories (raw): {total_stories_db}")
print(f"Stories (clean): {total_clean_db}")
print(f"DB: {DB_PATH}")
if __name__ == "__main__":
main()
|