Spaces:
Running
Running
File size: 11,059 Bytes
de28957 | 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 | from __future__ import annotations
import json
from typing import Any, Literal, Optional
from langchain_core.tools import tool
from psycopg.rows import dict_row
from app.db.postgres import get_readonly_connection
Metric = Literal[
"fake_news_count",
"articles_by_topic",
"sentiment_breakdown",
"propaganda_count",
"hate_speech_count",
"dialect_breakdown",
"article_lookup",
]
async def _fetch(sql: str, params: list[Any]) -> list[dict[str, Any]]:
async with get_readonly_connection() as conn:
async with conn.cursor(row_factory=dict_row) as cur:
await cur.execute(sql, params)
return await cur.fetchall()
def _date_filter(where: list[str], params: list[Any], date_col: str,
date_from: Optional[str], date_to: Optional[str]) -> None:
if date_from:
where.append(f"{date_col} >= %s")
params.append(date_from)
if date_to:
where.append(f"{date_col} <= %s")
params.append(date_to)
async def _fake_news_count(date_from, date_to, article_id, limit):
where: list[str] = []
params: list[Any] = []
_date_filter(where, params, "analyzed_at", date_from, date_to)
if article_id is not None:
where.append("article_id = %s")
params.append(article_id)
where_sql = f"WHERE {' AND '.join(where)}" if where else ""
breakdown_sql = f"""
SELECT verdict, COUNT(*) AS count
FROM article_fake_news
{where_sql}
GROUP BY verdict
ORDER BY count DESC
"""
rows = await _fetch(breakdown_sql, params)
refs_sql = f"""
SELECT article_id
FROM article_fake_news
{where_sql}
ORDER BY analyzed_at DESC
LIMIT %s
"""
refs = await _fetch(refs_sql, params + [limit])
return breakdown_sql, rows, refs
async def _articles_by_topic(date_from, date_to, topic, article_id, limit):
where: list[str] = []
params: list[Any] = []
_date_filter(where, params, "analyzed_at", date_from, date_to)
if topic:
where.append("(primary_topic = %s OR secondary_topic = %s)")
params.extend([topic, topic])
if article_id is not None:
where.append("article_id = %s")
params.append(article_id)
where_sql = f"WHERE {' AND '.join(where)}" if where else ""
breakdown_sql = f"""
SELECT primary_topic, COUNT(*) AS count
FROM article_topic
{where_sql}
GROUP BY primary_topic
ORDER BY count DESC
"""
rows = await _fetch(breakdown_sql, params)
refs_sql = f"""
SELECT article_id
FROM article_topic
{where_sql}
ORDER BY analyzed_at DESC
LIMIT %s
"""
refs = await _fetch(refs_sql, params + [limit])
return breakdown_sql, rows, refs
async def _sentiment_breakdown(date_from, date_to, sentiment, article_id, limit):
where: list[str] = []
params: list[Any] = []
_date_filter(where, params, "analyzed_at", date_from, date_to)
if sentiment:
where.append("primary_sentiment = %s")
params.append(sentiment)
if article_id is not None:
where.append("article_id = %s")
params.append(article_id)
where_sql = f"WHERE {' AND '.join(where)}" if where else ""
breakdown_sql = f"""
SELECT primary_sentiment, COUNT(*) AS count
FROM article_sentiment
{where_sql}
GROUP BY primary_sentiment
ORDER BY count DESC
"""
rows = await _fetch(breakdown_sql, params)
refs_sql = f"""
SELECT article_id
FROM article_sentiment
{where_sql}
ORDER BY analyzed_at DESC
LIMIT %s
"""
refs = await _fetch(refs_sql, params + [limit])
return breakdown_sql, rows, refs
async def _propaganda_count(date_from, date_to, article_id, limit):
where: list[str] = []
params: list[Any] = []
_date_filter(where, params, "analyzed_at", date_from, date_to)
if article_id is not None:
where.append("article_id = %s")
params.append(article_id)
where_sql = f"WHERE {' AND '.join(where)}" if where else ""
propaganda_where = where + ["is_propaganda = true"]
count_sql = f"""
SELECT COUNT(*) AS count
FROM article_propaganda
WHERE {' AND '.join(propaganda_where)}
"""
rows = await _fetch(count_sql, params)
technique_sql = f"""
SELECT technique, COUNT(*) AS count
FROM article_propaganda, unnest(techniques) AS technique
WHERE {' AND '.join(propaganda_where)}
GROUP BY technique
ORDER BY count DESC
LIMIT 10
"""
technique_rows = await _fetch(technique_sql, params)
rows[0]["technique_breakdown"] = technique_rows
refs_sql = f"""
SELECT article_id
FROM article_propaganda
WHERE {' AND '.join(propaganda_where)}
ORDER BY analyzed_at DESC
LIMIT %s
"""
refs = await _fetch(refs_sql, params + [limit])
return count_sql, rows, refs
async def _hate_speech_count(date_from, date_to, category, article_id, limit):
where: list[str] = []
params: list[Any] = []
_date_filter(where, params, "analyzed_at", date_from, date_to)
if category:
where.append("predicted_category = %s")
params.append(category)
if article_id is not None:
where.append("article_id = %s")
params.append(article_id)
where_sql = f"WHERE {' AND '.join(where)}" if where else ""
breakdown_sql = f"""
SELECT predicted_category, COUNT(*) AS count
FROM article_hate_speech
{where_sql}
GROUP BY predicted_category
ORDER BY count DESC
"""
rows = await _fetch(breakdown_sql, params)
refs_sql = f"""
SELECT article_id
FROM article_hate_speech
{where_sql}
ORDER BY analyzed_at DESC
LIMIT %s
"""
refs = await _fetch(refs_sql, params + [limit])
return breakdown_sql, rows, refs
async def _dialect_breakdown(date_from, date_to, dialect, article_id, limit):
where: list[str] = []
params: list[Any] = []
_date_filter(where, params, "analyzed_at", date_from, date_to)
if dialect:
where.append("detected_dialect = %s")
params.append(dialect)
if article_id is not None:
where.append("article_id = %s")
params.append(article_id)
where_sql = f"WHERE {' AND '.join(where)}" if where else ""
breakdown_sql = f"""
SELECT detected_dialect, COUNT(*) AS count
FROM article_dialect
{where_sql}
GROUP BY detected_dialect
ORDER BY count DESC
"""
rows = await _fetch(breakdown_sql, params)
refs_sql = f"""
SELECT article_id
FROM article_dialect
{where_sql}
ORDER BY analyzed_at DESC
LIMIT %s
"""
refs = await _fetch(refs_sql, params + [limit])
return breakdown_sql, rows, refs
async def _article_lookup(date_from, date_to, article_id, keyword, limit):
where: list[str] = []
params: list[Any] = []
_date_filter(where, params, "published_at", date_from, date_to)
if article_id is not None:
where.append("id = %s")
params.append(article_id)
if keyword:
where.append("title ILIKE %s")
params.append(f"%{keyword}%")
where_sql = f"WHERE {' AND '.join(where)}" if where else ""
sql = f"""
SELECT id, title, url, published_at
FROM news_articles
{where_sql}
ORDER BY published_at DESC
LIMIT %s
"""
rows = await _fetch(sql, params + [limit])
refs = [{"article_id": r["id"]} for r in rows]
return sql, rows, refs
@tool
async def sql_query_tool(
metric: Metric,
date_from: Optional[str] = None,
date_to: Optional[str] = None,
topic: Optional[str] = None,
sentiment: Optional[str] = None,
category: Optional[str] = None,
dialect: Optional[str] = None,
article_id: Optional[int] = None,
keyword: Optional[str] = None,
limit: int = 20,
) -> str:
"""
Run a pre-defined, read-only query over analyzed news articles.
metric: which statistic/lookup to run —
- fake_news_count: breakdown of fake-news verdicts
(SUPPORTED, REFUTED, PARTIALLY_TRUE, UNVERIFIABLE)
- articles_by_topic: article counts grouped by primary topic
- sentiment_breakdown: article counts grouped by sentiment
(positive, negative, neutral)
- propaganda_count: count of articles flagged as propaganda,
plus a technique frequency breakdown
- hate_speech_count: article counts grouped by hate-speech category
(none, other, origin, gender, religion)
- dialect_breakdown: article counts grouped by detected dialect
- article_lookup: look up specific articles by id or title keyword
date_from / date_to: ISO dates (YYYY-MM-DD). For all metrics except
article_lookup, these filter on the analysis timestamp
(`analyzed_at`) — i.e. when the ML model ran, not when the article
was published. For article_lookup they filter on `published_at`.
topic, sentiment, category, dialect: optional exact-match filters for
the corresponding metric.
article_id: restrict to a single article.
keyword: for article_lookup, a substring to search for in the title.
limit: max number of source article references / lookup rows to return.
Returns a JSON string: {"rows": [...], "sql_used": "...",
"source_refs": [{"article_id": ...}, ...]}
"""
handlers = {
"fake_news_count": lambda: _fake_news_count(date_from, date_to, article_id, limit),
"articles_by_topic": lambda: _articles_by_topic(date_from, date_to, topic, article_id, limit),
"sentiment_breakdown": lambda: _sentiment_breakdown(date_from, date_to, sentiment, article_id, limit),
"propaganda_count": lambda: _propaganda_count(date_from, date_to, article_id, limit),
"hate_speech_count": lambda: _hate_speech_count(date_from, date_to, category, article_id, limit),
"dialect_breakdown": lambda: _dialect_breakdown(date_from, date_to, dialect, article_id, limit),
"article_lookup": lambda: _article_lookup(date_from, date_to, article_id, keyword, limit),
}
handler = handlers.get(metric)
if handler is None:
return json.dumps({"rows": [], "sql_used": "", "source_refs": [],
"error": f"Unknown metric: {metric}"})
sql_used, rows, refs = await handler()
source_refs = [{"type": "article", "id": r["article_id"]} for r in refs]
return json.dumps(
{"rows": rows, "sql_used": sql_used.strip(), "source_refs": source_refs},
default=str,
)
|