| 1 | """Xquik X search source for the v3.0.0 last30days pipeline. |
| 2 | |
| 3 | Uses the Xquik REST API (https://xquik.com/api/v1) to search X/Twitter |
| 4 | with full engagement metrics (likes, retweets, replies, quotes, views, |
| 5 | bookmarks). Requires an API key from xquik.com. |
| 6 | """ |
| 7 | |
| 8 | from __future__ import annotations |
| 9 | |
| 10 | from datetime import datetime |
| 11 | from typing import Any, Dict, List, Optional |
| 12 | |
| 13 | from . import http, log |
| 14 | from .relevance import token_overlap_relevance as _compute_relevance |
| 15 | |
| 16 | # Per-process probe cache: (state, reason). state is "unset" until probed, then |
| 17 | # True (funded/working) | False (auth/payment failure) | None (inconclusive). |
| 18 | _probe_cache: tuple = ("unset", "") |
| 19 | |
| 20 | # Depth configurations: number of results to request per query |
| 21 | DEPTH_CONFIG = { |
| 22 | "quick": {"limit": 10, "queries": 1}, |
| 23 | "default": {"limit": 20, "queries": 2}, |
| 24 | "deep": {"limit": 40, "queries": 3}, |
| 25 | } |
| 26 | |
| 27 | _BASE_URL = "https://xquik.com/api/v1" |
| 28 | |
| 29 | |
| 30 | def _log(msg: str): |
| 31 | log.source_log("Xquik", msg, tty_only=False) |
| 32 | |
| 33 | |
| 34 | def _extract_core_subject(topic: str) -> str: |
| 35 | """Extract core subject for X search queries.""" |
| 36 | from .query import extract_core_subject |
| 37 | return extract_core_subject(topic, max_words=5, strip_suffixes=True) |
| 38 | |
| 39 | |
| 40 | def expand_xquik_queries(topic: str, depth: str) -> List[str]: |
| 41 | """Generate query variants based on depth. |
| 42 | |
| 43 | Args: |
| 44 | topic: Research topic |
| 45 | depth: "quick", "default", or "deep" |
| 46 | |
| 47 | Returns: |
| 48 | List of query strings (1 for quick, 2 for default, 3 for deep). |
| 49 | """ |
| 50 | core = _extract_core_subject(topic) |
| 51 | # Anti-bare-generic guard (#607): never let the core collapse to a single |
| 52 | # bare token when the topic carries more — a lone generic word floods X with |
| 53 | # off-topic collisions. Fall back to the full multi-word topic as the anchor. |
| 54 | topic_clean = topic.strip() |
| 55 | if len(core.split()) <= 1 and len(topic_clean.split()) > 1 and core.lower() != topic_clean.lower(): |
| 56 | core = topic_clean |
| 57 | queries = [core] |
| 58 | |
| 59 | # Add original topic if meaningfully different |
| 60 | if topic.lower().strip() != core.lower().strip(): |
| 61 | queries.append(topic.strip()) |
| 62 | |
| 63 | # Add compound term variant for deep searches |
| 64 | if len(queries) < 3: |
| 65 | from .query import extract_compound_terms |
| 66 | compounds = extract_compound_terms(topic) |
| 67 | if compounds: |
| 68 | or_parts = " OR ".join(f'"{t}"' for t in compounds[:3]) |
| 69 | queries.append(f"({or_parts})") |
| 70 | |
| 71 | cap = DEPTH_CONFIG.get(depth, DEPTH_CONFIG["default"])["queries"] |
| 72 | return queries[:cap] |
| 73 | |
| 74 | |
| 75 | def search_xquik( |
| 76 | topic: str, |
| 77 | from_date: str, |
| 78 | to_date: str, |
| 79 | depth: str = "default", |
| 80 | token: str = "", |
| 81 | ) -> Dict[str, Any]: |
| 82 | """Search X via Xquik REST API. |
| 83 | |
| 84 | Args: |
| 85 | topic: Search topic |
| 86 | from_date: Start date (YYYY-MM-DD) |
| 87 | to_date: End date (YYYY-MM-DD) |
| 88 | depth: Research depth - "quick", "default", or "deep" |
| 89 | token: Xquik API key |
| 90 | |
| 91 | Returns: |
| 92 | Dict with "items" list and optional "error" string. |
| 93 | """ |
| 94 | if not token: |
| 95 | return {"items": [], "error": "No XQUIK_API_KEY configured"} |
| 96 | |
| 97 | cfg = DEPTH_CONFIG.get(depth, DEPTH_CONFIG["default"]) |
| 98 | queries = expand_xquik_queries(topic, depth) |
| 99 | all_items: List[Dict[str, Any]] = [] |
| 100 | seen_ids: set[str] = set() |
| 101 | |
| 102 | for query_text in queries: |
| 103 | q = f"{query_text} since:{from_date} until:{to_date}" |
| 104 | items, auth_error = _execute_search( |
| 105 | q, cfg["limit"], token, |
| 106 | label=query_text, id_prefix="XQ", |
| 107 | seen_ids=seen_ids, relevance_query=query_text, |
| 108 | index_offset=len(all_items), |
| 109 | ) |
| 110 | if auth_error: |
| 111 | # Auth/payment failure is fatal for the whole source (e.g. 401/403, |
| 112 | # and 402-unpaid surfaced via U5 diagnose) — return it so the caller |
| 113 | # settles honestly instead of silently empty. |
| 114 | return {"items": [], "error": auth_error} |
| 115 | all_items.extend(items) |
| 116 | |
| 117 | return {"items": all_items} |
| 118 | |
| 119 | |
| 120 | def _execute_search( |
| 121 | q: str, |
| 122 | limit: int, |
| 123 | token: str, |
| 124 | *, |
| 125 | label: str, |
| 126 | id_prefix: str, |
| 127 | seen_ids: set[str], |
| 128 | relevance_query: str, |
| 129 | index_offset: int = 0, |
| 130 | ) -> tuple[List[Dict[str, Any]], str | None]: |
| 131 | """Run one Xquik search call and parse its tweets. |
| 132 | |
| 133 | Returns ``(items, auth_error)``. ``auth_error`` is a non-empty string only |
| 134 | on a fatal auth/payment failure (401/403); transient/HTTP errors log and |
| 135 | return ``([], None)`` so one bad lane never discards another's results. |
| 136 | ``relevance_query`` (the topic) is what items are scored against — for the |
| 137 | handle lanes that differs from the search query (``from:handle``). |
| 138 | ``index_offset`` keeps item ids unique across multiple calls that share an |
| 139 | accumulator (multi-query topic search, per-handle lanes). |
| 140 | """ |
| 141 | full_url = f"{_BASE_URL}/x/tweets/search?q={_url_encode(q)}&queryType=Top&limit={limit}" |
| 142 | _log(f"Searching: {label}") |
| 143 | try: |
| 144 | request_headers = {"X-Api-Key": token} |
| 145 | response = http.get(full_url, headers=request_headers, timeout=30, retries=2) |
| 146 | except http.HTTPError as exc: |
| 147 | status = getattr(exc, "status_code", None) |
| 148 | if status == 402: |
| 149 | # Unpaid key — fatal for the source, and surfaced on the real search |
| 150 | # path (not just --diagnose) so a live run reports it instead of |
| 151 | # settling silently empty. |
| 152 | return [], "Xquik key unpaid (402)" |
| 153 | if status in (401, 403): |
| 154 | return [], f"Xquik auth failed ({status})" |
| 155 | _log(f"HTTP error for '{label}': {exc}") |
| 156 | return [], None |
| 157 | except Exception as exc: |
| 158 | _log(f"Error for '{label}': {exc}") |
| 159 | return [], None |
| 160 | |
| 161 | tweets = response.get("tweets", []) |
| 162 | if not isinstance(tweets, list): |
| 163 | return [], None |
| 164 | items: List[Dict[str, Any]] = [] |
| 165 | for tweet in tweets: |
| 166 | if not isinstance(tweet, dict): |
| 167 | continue |
| 168 | tweet_id = str(tweet.get("id", "")) |
| 169 | if tweet_id in seen_ids: |
| 170 | continue |
| 171 | seen_ids.add(tweet_id) |
| 172 | item = _parse_tweet(tweet, index_offset + len(items), relevance_query, id_prefix=id_prefix) |
| 173 | if item: |
| 174 | items.append(item) |
| 175 | return items, None |
| 176 | |
| 177 | |
| 178 | def _is_own(url: str, handle: str) -> bool: |
| 179 | """True when a tweet URL is authored by ``handle`` (their own post). |
| 180 | |
| 181 | Used by the ABOUT lane to drop the subject's own tweets so only mentions |
| 182 | *by others* remain. Handles both x.com and twitter.com permalinks. |
| 183 | """ |
| 184 | u = (url or "").lower() |
| 185 | h = handle.lower().lstrip("@").strip() |
| 186 | return bool(h) and (f"x.com/{h}/status" in u or f"twitter.com/{h}/status" in u) |
| 187 | |
| 188 | |
| 189 | def search_handles( |
| 190 | handles: List[str], |
| 191 | topic: str, |
| 192 | from_date: str, |
| 193 | to_date: str, |
| 194 | *, |
| 195 | count_per: int = 8, |
| 196 | token: str = "", |
| 197 | ) -> List[Dict[str, Any]]: |
| 198 | """FROM lane: tweets authored BY each handle (their own timeline). |
| 199 | |
| 200 | The topic is NOT AND'd into the query (that was the from:-AND bug, #610) — |
| 201 | we pull the raw timeline and use ``topic`` for relevance ranking only. |
| 202 | Returns a flat list of item dicts (mirrors ``bird_x.search_handles``). |
| 203 | """ |
| 204 | if not token or not handles: |
| 205 | return [] |
| 206 | items: List[Dict[str, Any]] = [] |
| 207 | seen_ids: set[str] = set() |
| 208 | for raw in handles: |
| 209 | handle = str(raw).lstrip("@").strip() |
| 210 | if not handle: |
| 211 | continue |
| 212 | q = f"from:{handle} since:{from_date} until:{to_date}" |
| 213 | got, auth_error = _execute_search( |
| 214 | q, count_per, token, |
| 215 | label=f"from:{handle}", id_prefix="XF", |
| 216 | seen_ids=seen_ids, relevance_query=topic, |
| 217 | index_offset=len(items), |
| 218 | ) |
| 219 | if auth_error: |
| 220 | break # fatal auth/payment failure — stop, keep what we have |
| 221 | items.extend(got) |
| 222 | return items |
| 223 | |
| 224 | |
| 225 | def search_mentions( |
| 226 | handles: List[str], |
| 227 | from_date: str, |
| 228 | to_date: str, |
| 229 | *, |
| 230 | topic: str = "", |
| 231 | count_per: int = 5, |
| 232 | token: str = "", |
| 233 | ) -> List[Dict[str, Any]]: |
| 234 | """ABOUT lane: tweets mentioning each handle, authored by OTHERS. |
| 235 | |
| 236 | Queries ``@handle`` then drops the handle's own tweets (``_is_own``) so only |
| 237 | third-party mentions remain. Returns a flat list of item dicts. |
| 238 | """ |
| 239 | if not token or not handles: |
| 240 | return [] |
| 241 | items: List[Dict[str, Any]] = [] |
| 242 | seen_ids: set[str] = set() |
| 243 | for raw in handles: |
| 244 | handle = str(raw).lstrip("@").strip() |
| 245 | if not handle: |
| 246 | continue |
| 247 | q = f"@{handle} since:{from_date} until:{to_date}" |
| 248 | got, auth_error = _execute_search( |
| 249 | q, count_per, token, |
| 250 | label=f"@{handle}", id_prefix="XA", |
| 251 | seen_ids=seen_ids, relevance_query=topic, |
| 252 | index_offset=len(items), |
| 253 | ) |
| 254 | if auth_error: |
| 255 | break |
| 256 | items.extend(it for it in got if not _is_own(it.get("url", ""), handle)) |
| 257 | return items |
| 258 | |
| 259 | |
| 260 | def probe_works(token: str, timeout: int = 8) -> Optional[bool]: |
| 261 | """Cheap runtime check that the xquik key actually returns data. |
| 262 | |
| 263 | Mirrors ``bird_x.probe_works`` for the key-based X path so ``--diagnose`` |
| 264 | reflects reality instead of static key presence. Returns True |
| 265 | (funded/working), False (a clear auth/payment failure — 401/403, or 402 |
| 266 | when the key is configured but unpaid), or None (inconclusive: timeout / |
| 267 | transient HTTP) so callers fail open. |
| 268 | The human-readable reason is available via ``probe_reason()``. Cached per |
| 269 | process so repeated diagnose calls don't re-probe. |
| 270 | """ |
| 271 | global _probe_cache |
| 272 | if _probe_cache[0] != "unset": |
| 273 | return _probe_cache[0] |
| 274 | if not token: |
| 275 | _probe_cache = (False, "no XQUIK_API_KEY configured") |
| 276 | return False |
| 277 | from datetime import timedelta, timezone |
| 278 | since = (datetime.now(timezone.utc) - timedelta(days=30)).strftime("%Y-%m-%d") |
| 279 | # @x (the platform's own account) posts frequently, so a no-error response |
| 280 | # means the key works even if this particular window is quiet. |
| 281 | q = f"from:x since:{since}" |
| 282 | full_url = f"{_BASE_URL}/x/tweets/search?q={_url_encode(q)}&queryType=Top&limit=1" |
| 283 | try: |
| 284 | request_headers = {"X-Api-Key": token} |
| 285 | http.get(full_url, headers=request_headers, timeout=timeout, retries=0) |
| 286 | except http.HTTPError as exc: |
| 287 | status = getattr(exc, "status_code", None) |
| 288 | if status == 402: |
| 289 | _probe_cache = (False, "xquik key unpaid (402)") |
| 290 | elif status in (401, 403): |
| 291 | _probe_cache = (False, f"xquik auth failed ({status})") |
| 292 | else: |
| 293 | # 5xx / unexpected status — inconclusive, don't report a false-down. |
| 294 | _probe_cache = (None, f"xquik probe inconclusive ({status})") |
| 295 | return _probe_cache[0] |
| 296 | except Exception as exc: |
| 297 | _probe_cache = (None, f"xquik probe inconclusive ({type(exc).__name__})") |
| 298 | return None |
| 299 | _probe_cache = (True, "ok") |
| 300 | return True |
| 301 | |
| 302 | |
| 303 | def probe_reason() -> str: |
| 304 | """Human-readable reason for the last ``probe_works`` result (or '').""" |
| 305 | return _probe_cache[1] |
| 306 | |
| 307 | |
| 308 | def search_and_enrich( |
| 309 | topic: str, |
| 310 | from_date: str, |
| 311 | to_date: str, |
| 312 | depth: str = "default", |
| 313 | token: str = "", |
| 314 | ) -> Dict[str, Any]: |
| 315 | """Search X via Xquik and return results. |
| 316 | |
| 317 | Xquik API returns full engagement data by default, so no separate |
| 318 | enrichment step is needed. |
| 319 | """ |
| 320 | return search_xquik(topic, from_date, to_date, depth=depth, token=token) |
| 321 | |
| 322 | |
| 323 | def parse_xquik_response(response: Dict[str, Any]) -> List[Dict[str, Any]]: |
| 324 | """Extract items from search response. |
| 325 | |
| 326 | Args: |
| 327 | response: Response dict from search_xquik() |
| 328 | |
| 329 | Returns: |
| 330 | List of normalized item dicts. |
| 331 | """ |
| 332 | return response.get("items", []) |
| 333 | |
| 334 | |
| 335 | def _parse_tweet( |
| 336 | tweet: Dict[str, Any], index: int, query: str, id_prefix: str = "XQ" |
| 337 | ) -> Dict[str, Any] | None: |
| 338 | """Parse a single tweet from the API response into the standard item format.""" |
| 339 | author = tweet.get("author") or {} |
| 340 | username = str(author.get("username", "")).lstrip("@") |
| 341 | tweet_id = str(tweet.get("id", "")) |
| 342 | |
| 343 | # Build URL |
| 344 | url = "" |
| 345 | if username and tweet_id: |
| 346 | url = f"https://x.com/{username}/status/{tweet_id}" |
| 347 | if not url: |
| 348 | return None |
| 349 | |
| 350 | # Parse date |
| 351 | date = None |
| 352 | created_at = tweet.get("createdAt") or "" |
| 353 | if created_at: |
| 354 | try: |
| 355 | if len(created_at) > 10 and created_at[10] == "T": |
| 356 | dt = datetime.fromisoformat(created_at.replace("Z", "+00:00")) |
| 357 | else: |
| 358 | dt = datetime.strptime(created_at, "%a %b %d %H:%M:%S %z %Y") |
| 359 | date = dt.strftime("%Y-%m-%d") |
| 360 | except (ValueError, TypeError): |
| 361 | pass |
| 362 | |
| 363 | text = str(tweet.get("text", "")).strip()[:500] |
| 364 | |
| 365 | # Leading-run @mentions = who the post is directed at (reply target). Shared |
| 366 | # parser with bird so the first-party interaction signal fires for xquik too. |
| 367 | from .query import leading_mentions |
| 368 | mentioned_handles = leading_mentions(text) |
| 369 | |
| 370 | # Build engagement dict with full metrics |
| 371 | engagement = { |
| 372 | "likes": _safe_int(tweet.get("likeCount")), |
| 373 | "reposts": _safe_int(tweet.get("retweetCount")), |
| 374 | "replies": _safe_int(tweet.get("replyCount")), |
| 375 | "quotes": _safe_int(tweet.get("quoteCount")), |
| 376 | "views": _safe_int(tweet.get("viewCount")), |
| 377 | "bookmarks": _safe_int(tweet.get("bookmarkCount")), |
| 378 | } |
| 379 | |
| 380 | return { |
| 381 | "id": f"{id_prefix}{index + 1}", |
| 382 | "text": text, |
| 383 | "url": url, |
| 384 | "author_handle": username, |
| 385 | "date": date, |
| 386 | "engagement": engagement, |
| 387 | "mentioned_handles": mentioned_handles, |
| 388 | "relevance": _compute_relevance(query, text) if query else 0.7, |
| 389 | "why_relevant": "", |
| 390 | } |
| 391 | |
| 392 | |
| 393 | def _safe_int(value: Any) -> int | None: |
| 394 | """Convert value to int, returning None on failure.""" |
| 395 | if value is None: |
| 396 | return None |
| 397 | try: |
| 398 | return int(value) |
| 399 | except (ValueError, TypeError): |
| 400 | return None |
| 401 | |
| 402 | |
| 403 | def _url_encode(text: str) -> str: |
| 404 | """URL-encode a string using stdlib.""" |
| 405 | from urllib.parse import quote |
| 406 | return quote(text, safe="") |
| 407 |