| 1 | """Public jobs/careers retrieval for Hiring Signals. |
| 2 | |
| 3 | Tiered strategy (each tier degrades gracefully; the artifact records which |
| 4 | tier produced results so synthesis knows the confidence level): |
| 5 | |
| 6 | - Tier 1 - direct ATS API. The company's own board is authoritative and |
| 7 | structured. Discovery is careers-page-first: fetch the careers page and read |
| 8 | the provider + exact slug straight off the embed/link, then call that API. |
| 9 | We only emit ATS results when a call actually returns a non-empty board, so a |
| 10 | bad slug guess never produces fabricated coverage. Slug-probing is a fallback, |
| 11 | never the entry point. |
| 12 | - Tier 2 - careers page found, no supported ATS API. Parse schema.org |
| 13 | ``JobPosting`` JSON-LD (emitted by most careers pages for Google Jobs SEO, |
| 14 | regardless of ATS). |
| 15 | - Tier 3 - generic web search. Last resort, noisy, clearly low-confidence. |
| 16 | """ |
| 17 | |
| 18 | from __future__ import annotations |
| 19 | |
| 20 | import json |
| 21 | import re |
| 22 | from html import unescape |
| 23 | from typing import Any, Callable |
| 24 | from urllib.parse import urlparse |
| 25 | |
| 26 | from . import dates, grounding, http |
| 27 | |
| 28 | |
| 29 | ATS_PROVIDER_GREENHOUSE = "greenhouse" |
| 30 | ATS_PROVIDER_ASHBY = "ashby" |
| 31 | ATS_PROVIDER_LEVER = "lever" |
| 32 | ATS_PROVIDER_WORKABLE = "workable" |
| 33 | ATS_PROVIDER_SMARTRECRUITERS = "smartrecruiters" |
| 34 | |
| 35 | # Detection patterns: map an ATS embed/link found on a careers page to |
| 36 | # (provider, slug). The published slug is authoritative - this is why discovery |
| 37 | # is careers-page-first rather than blind probing. |
| 38 | _ATS_LINK_PATTERNS: list[tuple[str, str]] = [ |
| 39 | (ATS_PROVIDER_ASHBY, r"(?:jobs|api)\.ashbyhq\.com/(?:posting-api/job-board/)?([A-Za-z0-9_.-]+)"), |
| 40 | (ATS_PROVIDER_GREENHOUSE, r"boards(?:-api)?\.greenhouse\.io/(?:v1/boards/|embed/job_board\?for=)?([A-Za-z0-9_.-]+)"), |
| 41 | (ATS_PROVIDER_GREENHOUSE, r"job-boards\.greenhouse\.io/([A-Za-z0-9_.-]+)"), |
| 42 | (ATS_PROVIDER_GREENHOUSE, r"greenhouse\.io/embed/job_board\?for=([A-Za-z0-9_.-]+)"), |
| 43 | (ATS_PROVIDER_LEVER, r"(?:jobs\.lever\.co|api\.lever\.co/v0/postings)/([A-Za-z0-9_.-]+)"), |
| 44 | (ATS_PROVIDER_WORKABLE, r"apply\.workable\.com/(?:api/v[0-9]+/accounts/)?([A-Za-z0-9_.-]+)"), |
| 45 | (ATS_PROVIDER_WORKABLE, r"([A-Za-z0-9_-]+)\.workable\.com"), |
| 46 | (ATS_PROVIDER_SMARTRECRUITERS, r"(?:careers|jobs)\.smartrecruiters\.com/([A-Za-z0-9_.-]+)"), |
| 47 | (ATS_PROVIDER_SMARTRECRUITERS, r"api\.smartrecruiters\.com/v1/companies/([A-Za-z0-9_.-]+)"), |
| 48 | ] |
| 49 | |
| 50 | # Tokens that show up in ATS URLs but are never real board slugs. |
| 51 | _SLUG_STOPWORDS = {"embed", "job_board", "v1", "v0", "api", "posting-api", "boards", "jobs", "job-boards", "www"} |
| 52 | |
| 53 | |
| 54 | def search_jobs( |
| 55 | company: str, |
| 56 | date_range: tuple[str, str], |
| 57 | config: dict[str, Any], |
| 58 | *, |
| 59 | depth: str = "default", |
| 60 | web_backend: str = "auto", |
| 61 | explicit: bool = False, |
| 62 | ) -> tuple[list[dict[str, Any]], dict[str, Any]]: |
| 63 | """Fetch public job postings for a company via the tiered strategy.""" |
| 64 | company = company.strip() |
| 65 | if not company: |
| 66 | return [], {} |
| 67 | |
| 68 | attempted: list[str] = [] |
| 69 | |
| 70 | # --- Discovery: fetch the careers page and read the ATS off it (Tier 1). --- |
| 71 | careers_html, careers_url = _resolve_careers_page( |
| 72 | company, date_range, config, backend=web_backend, |
| 73 | ) |
| 74 | provider, slug = (None, None) |
| 75 | if careers_html: |
| 76 | provider, slug = detect_ats(careers_html) |
| 77 | if provider: |
| 78 | attempted.append(f"careers:{provider}:{slug}") |
| 79 | |
| 80 | # Fallback discovery: cheap deterministic slug probe (only after careers-page |
| 81 | # discovery fails - never the entry point). |
| 82 | if not provider: |
| 83 | provider, slug, probe_attempts = _probe_ats(company) |
| 84 | attempted.extend(probe_attempts) |
| 85 | |
| 86 | # --- Tier 1: call the resolved ATS API. Trust it only if it has jobs. --- |
| 87 | if provider and slug: |
| 88 | try: |
| 89 | items = _fetch_ats(provider, slug) |
| 90 | except http.HTTPError: |
| 91 | items = [] |
| 92 | if items: |
| 93 | return items, _artifact("jobs", company, attempted, items, explicit, |
| 94 | tier="ats", provider=provider, slug=slug) |
| 95 | |
| 96 | # --- Tier 2: parse JSON-LD JobPosting off the careers page. --- |
| 97 | if careers_html: |
| 98 | attempted.append("careers:jsonld") |
| 99 | jsonld_items = extract_jsonld_jobs(careers_html, careers_url or "") |
| 100 | if jsonld_items: |
| 101 | return jsonld_items, _artifact("jobs", company, attempted, jsonld_items, |
| 102 | explicit, tier="careers-jsonld") |
| 103 | |
| 104 | # --- Tier 3: generic web search (noisy, low-confidence). --- |
| 105 | fallback_items, artifact = search_jobs_web( |
| 106 | company, date_range, config, backend=web_backend, |
| 107 | ) |
| 108 | artifact = dict(artifact or {}) |
| 109 | artifact.setdefault("attempted", attempted) |
| 110 | artifact.update({ |
| 111 | "label": artifact.get("label", "jobs"), |
| 112 | "company": company, |
| 113 | "tier": "web", |
| 114 | "explicit": explicit, |
| 115 | "resultCount": len(fallback_items), |
| 116 | }) |
| 117 | return fallback_items, artifact |
| 118 | |
| 119 | |
| 120 | # --------------------------------------------------------------------------- # |
| 121 | # Careers-page discovery |
| 122 | # --------------------------------------------------------------------------- # |
| 123 | |
| 124 | def _resolve_careers_page( |
| 125 | company: str, |
| 126 | date_range: tuple[str, str], |
| 127 | config: dict[str, Any], |
| 128 | *, |
| 129 | backend: str = "auto", |
| 130 | ) -> tuple[str | None, str | None]: |
| 131 | """Find and fetch the company's careers page HTML. |
| 132 | |
| 133 | Tries conventional URLs on a guessed domain first (free, deterministic); |
| 134 | falls back to a single web search for the careers page when a backend is |
| 135 | configured. Returns (html, url) or (None, None). Never raises. |
| 136 | """ |
| 137 | slug = _company_slug(company) |
| 138 | candidates: list[str] = [] |
| 139 | if slug: |
| 140 | for host in (f"{slug}.com", f"{slug}.ai", f"{slug}.io"): |
| 141 | candidates.extend([f"https://{host}/careers", f"https://{host}/jobs"]) |
| 142 | |
| 143 | for url in candidates: |
| 144 | with http.expected_misses(403, 404): |
| 145 | html = http.get_text(url, accept="text/html", retries=1) |
| 146 | if html and _looks_like_careers_html(html): |
| 147 | return html, url |
| 148 | |
| 149 | if backend != "none": |
| 150 | careers_url = _search_for_careers_url(company, date_range, config, backend=backend) |
| 151 | if careers_url: |
| 152 | html = http.get_text(careers_url, accept="text/html", retries=1) |
| 153 | if html: |
| 154 | return html, careers_url |
| 155 | |
| 156 | return None, None |
| 157 | |
| 158 | |
| 159 | def _search_for_careers_url( |
| 160 | company: str, |
| 161 | date_range: tuple[str, str], |
| 162 | config: dict[str, Any], |
| 163 | *, |
| 164 | backend: str = "auto", |
| 165 | ) -> str | None: |
| 166 | """Use the configured web backend to locate the careers page URL.""" |
| 167 | try: |
| 168 | raw_items, _ = grounding.web_search( |
| 169 | f"{company} careers jobs", date_range, config, backend=backend, |
| 170 | ) |
| 171 | except Exception: |
| 172 | return None |
| 173 | for raw in raw_items or []: |
| 174 | if not isinstance(raw, dict): |
| 175 | continue |
| 176 | url = str(raw.get("url") or "").strip() |
| 177 | if not url: |
| 178 | continue |
| 179 | lowered = url.lower() |
| 180 | if any(token in lowered for token in ( |
| 181 | "career", "/jobs", "ashbyhq", "greenhouse", "lever.co", "workable", "smartrecruiters", |
| 182 | )): |
| 183 | return url |
| 184 | return None |
| 185 | |
| 186 | |
| 187 | def detect_ats(html: str) -> tuple[str | None, str | None]: |
| 188 | """Read the ATS provider + slug off a careers page's embed/links.""" |
| 189 | if not html: |
| 190 | return None, None |
| 191 | for provider, pattern in _ATS_LINK_PATTERNS: |
| 192 | for match in re.finditer(pattern, html): |
| 193 | slug = match.group(1).strip().strip("/.") |
| 194 | if slug and slug.lower() not in _SLUG_STOPWORDS: |
| 195 | return provider, slug |
| 196 | return None, None |
| 197 | |
| 198 | |
| 199 | def _probe_ats(company: str) -> tuple[str | None, str | None, list[str]]: |
| 200 | """Fallback: probe candidate slugs against ATS APIs (deterministic). |
| 201 | |
| 202 | Only reached when careers-page discovery fails. Returns the first provider |
| 203 | whose API returns a non-empty board. |
| 204 | """ |
| 205 | attempts: list[str] = [] |
| 206 | # Workable/SmartRecruiters are omitted here: their slugs rarely match the |
| 207 | # company name, so blind probing is unreliable - they're reached only via |
| 208 | # careers-page discovery, where the real slug is published. |
| 209 | for slug in _candidate_slugs(company): |
| 210 | for provider in (ATS_PROVIDER_GREENHOUSE, ATS_PROVIDER_ASHBY, ATS_PROVIDER_LEVER): |
| 211 | attempts.append(f"probe:{provider}:{slug}") |
| 212 | try: |
| 213 | with http.expected_misses(400, 401, 403, 404): |
| 214 | items = _fetch_ats(provider, slug) |
| 215 | except http.HTTPError as exc: |
| 216 | if exc.status_code in {400, 401, 403, 404}: |
| 217 | continue |
| 218 | raise |
| 219 | if items: |
| 220 | return provider, slug, attempts |
| 221 | return None, None, attempts |
| 222 | |
| 223 | |
| 224 | # --------------------------------------------------------------------------- # |
| 225 | # Tier 1: ATS API fetchers + parsers |
| 226 | # --------------------------------------------------------------------------- # |
| 227 | |
| 228 | def _fetch_ats(provider: str, slug: str) -> list[dict[str, Any]]: |
| 229 | fetchers: dict[str, Callable[[str], list[dict[str, Any]]]] = { |
| 230 | ATS_PROVIDER_GREENHOUSE: search_greenhouse_board, |
| 231 | ATS_PROVIDER_ASHBY: search_ashby_board, |
| 232 | ATS_PROVIDER_LEVER: search_lever_board, |
| 233 | ATS_PROVIDER_WORKABLE: search_workable_board, |
| 234 | ATS_PROVIDER_SMARTRECRUITERS: search_smartrecruiters_board, |
| 235 | } |
| 236 | fetcher = fetchers.get(provider) |
| 237 | return fetcher(slug) if fetcher else [] |
| 238 | |
| 239 | |
| 240 | def search_greenhouse_board(board_token: str) -> list[dict[str, Any]]: |
| 241 | """Return jobs from Greenhouse's public Job Board API.""" |
| 242 | url = f"https://boards-api.greenhouse.io/v1/boards/{board_token}/jobs" |
| 243 | data = http.get(url, params={"content": "true"}, timeout=15, retries=2) |
| 244 | jobs = data.get("jobs") if isinstance(data, dict) else [] |
| 245 | if not isinstance(jobs, list): |
| 246 | return [] |
| 247 | return [_greenhouse_job_to_item(job, board_token) for job in jobs if isinstance(job, dict)] |
| 248 | |
| 249 | |
| 250 | def search_ashby_board(slug: str) -> list[dict[str, Any]]: |
| 251 | """Return jobs from Ashby's public posting API (needs a browser UA).""" |
| 252 | url = f"https://api.ashbyhq.com/posting-api/job-board/{slug}" |
| 253 | data = http.get( |
| 254 | url, |
| 255 | params={"includeCompensation": "true"}, |
| 256 | headers={"User-Agent": http.BROWSER_USER_AGENT}, |
| 257 | timeout=15, |
| 258 | retries=2, |
| 259 | ) |
| 260 | return parse_ashby_response(data, slug) |
| 261 | |
| 262 | |
| 263 | def search_lever_board(slug: str) -> list[dict[str, Any]]: |
| 264 | """Return jobs from Lever's public postings API (returns a JSON list).""" |
| 265 | url = f"https://api.lever.co/v0/postings/{slug}" |
| 266 | data = http.get(url, params={"mode": "json"}, timeout=15, retries=2) |
| 267 | return parse_lever_response(data, slug) |
| 268 | |
| 269 | |
| 270 | def search_workable_board(slug: str) -> list[dict[str, Any]]: |
| 271 | """Return jobs from Workable's public widget API.""" |
| 272 | url = f"https://apply.workable.com/api/v3/accounts/{slug}/jobs" |
| 273 | data = http.post(url, json_data={}, timeout=15, retries=2) |
| 274 | return parse_workable_response(data, slug) |
| 275 | |
| 276 | |
| 277 | def search_smartrecruiters_board(slug: str) -> list[dict[str, Any]]: |
| 278 | """Return jobs from SmartRecruiters' public postings API.""" |
| 279 | url = f"https://api.smartrecruiters.com/v1/companies/{slug}/postings" |
| 280 | data = http.get(url, params={"limit": "100"}, timeout=15, retries=2) |
| 281 | return parse_smartrecruiters_response(data, slug) |
| 282 | |
| 283 | |
| 284 | def parse_greenhouse_response(payload: dict[str, Any], board_token: str = "") -> list[dict[str, Any]]: |
| 285 | """Parse a Greenhouse jobs payload for tests and callers with cached data.""" |
| 286 | jobs = payload.get("jobs") if isinstance(payload, dict) else [] |
| 287 | if not isinstance(jobs, list): |
| 288 | return [] |
| 289 | return [_greenhouse_job_to_item(job, board_token) for job in jobs if isinstance(job, dict)] |
| 290 | |
| 291 | |
| 292 | def parse_ashby_response(payload: dict[str, Any], slug: str = "") -> list[dict[str, Any]]: |
| 293 | jobs = payload.get("jobs") if isinstance(payload, dict) else [] |
| 294 | if not isinstance(jobs, list): |
| 295 | return [] |
| 296 | items: list[dict[str, Any]] = [] |
| 297 | for job in jobs: |
| 298 | if not isinstance(job, dict): |
| 299 | continue |
| 300 | department = str(job.get("departmentName") or job.get("teamName") or "").strip() |
| 301 | location = str(job.get("locationName") or job.get("location") or "").strip() |
| 302 | description = _clean_html(str(job.get("descriptionHtml") or job.get("descriptionPlain") or "")) |
| 303 | items.append(_ats_item( |
| 304 | provider=ATS_PROVIDER_ASHBY, |
| 305 | slug=slug, |
| 306 | ident=str(job.get("id") or job.get("jobId") or ""), |
| 307 | title=str(job.get("title") or "").strip(), |
| 308 | url=str(job.get("jobUrl") or job.get("applyUrl") or "").strip(), |
| 309 | description=description, |
| 310 | date=_date_part(job.get("publishedDate") or job.get("publishedAt")), |
| 311 | department=department, |
| 312 | location=location, |
| 313 | )) |
| 314 | return items |
| 315 | |
| 316 | |
| 317 | def parse_lever_response(payload: Any, slug: str = "") -> list[dict[str, Any]]: |
| 318 | postings = payload if isinstance(payload, list) else [] |
| 319 | items: list[dict[str, Any]] = [] |
| 320 | for job in postings: |
| 321 | if not isinstance(job, dict): |
| 322 | continue |
| 323 | categories = job.get("categories") if isinstance(job.get("categories"), dict) else {} |
| 324 | department = str(categories.get("department") or categories.get("team") or "").strip() |
| 325 | location = str(categories.get("location") or "").strip() |
| 326 | description = _clean_html(str(job.get("descriptionPlain") or job.get("description") or "")) |
| 327 | items.append(_ats_item( |
| 328 | provider=ATS_PROVIDER_LEVER, |
| 329 | slug=slug, |
| 330 | ident=str(job.get("id") or ""), |
| 331 | title=str(job.get("text") or "").strip(), |
| 332 | url=str(job.get("hostedUrl") or job.get("applyUrl") or "").strip(), |
| 333 | description=description, |
| 334 | date=_epoch_ms_to_date(job.get("createdAt")), |
| 335 | department=department, |
| 336 | location=location, |
| 337 | )) |
| 338 | return items |
| 339 | |
| 340 | |
| 341 | def parse_workable_response(payload: dict[str, Any], slug: str = "") -> list[dict[str, Any]]: |
| 342 | results = [] |
| 343 | if isinstance(payload, dict): |
| 344 | results = payload.get("results") or payload.get("jobs") or [] |
| 345 | if not isinstance(results, list): |
| 346 | return [] |
| 347 | items: list[dict[str, Any]] = [] |
| 348 | for job in results: |
| 349 | if not isinstance(job, dict): |
| 350 | continue |
| 351 | loc = job.get("location") if isinstance(job.get("location"), dict) else {} |
| 352 | location = _join_parts(str(loc.get("city") or ""), str(loc.get("country") or "")) |
| 353 | shortcode = str(job.get("shortcode") or "").strip() |
| 354 | url = str(job.get("url") or job.get("application_url") or "").strip() |
| 355 | if not url and shortcode and slug: |
| 356 | url = f"https://apply.workable.com/{slug}/j/{shortcode}/" |
| 357 | items.append(_ats_item( |
| 358 | provider=ATS_PROVIDER_WORKABLE, |
| 359 | slug=slug, |
| 360 | ident=shortcode or str(job.get("id") or ""), |
| 361 | title=str(job.get("title") or "").strip(), |
| 362 | url=url, |
| 363 | description=_clean_html(str(job.get("description") or "")), |
| 364 | date=_date_part(job.get("published_on") or job.get("created_at")), |
| 365 | department=str(job.get("department") or "").strip(), |
| 366 | location=location, |
| 367 | )) |
| 368 | return items |
| 369 | |
| 370 | |
| 371 | def parse_smartrecruiters_response(payload: dict[str, Any], slug: str = "") -> list[dict[str, Any]]: |
| 372 | content = payload.get("content") if isinstance(payload, dict) else [] |
| 373 | if not isinstance(content, list): |
| 374 | return [] |
| 375 | items: list[dict[str, Any]] = [] |
| 376 | for job in content: |
| 377 | if not isinstance(job, dict): |
| 378 | continue |
| 379 | department = "" |
| 380 | if isinstance(job.get("department"), dict): |
| 381 | department = str(job["department"].get("label") or "").strip() |
| 382 | location = "" |
| 383 | if isinstance(job.get("location"), dict): |
| 384 | location = _join_parts( |
| 385 | str(job["location"].get("city") or ""), |
| 386 | str(job["location"].get("country") or ""), |
| 387 | ) |
| 388 | ident = str(job.get("id") or job.get("uuid") or "").strip() |
| 389 | url = "" |
| 390 | if isinstance(job.get("ref"), str): |
| 391 | url = job["ref"] |
| 392 | if not url and slug and ident: |
| 393 | url = f"https://jobs.smartrecruiters.com/{slug}/{ident}" |
| 394 | items.append(_ats_item( |
| 395 | provider=ATS_PROVIDER_SMARTRECRUITERS, |
| 396 | slug=slug, |
| 397 | ident=ident, |
| 398 | title=str(job.get("name") or "").strip(), |
| 399 | url=url, |
| 400 | description="", |
| 401 | date=_date_part(job.get("releasedDate") or job.get("createdOn")), |
| 402 | department=department, |
| 403 | location=location, |
| 404 | )) |
| 405 | return items |
| 406 | |
| 407 | |
| 408 | # --------------------------------------------------------------------------- # |
| 409 | # Tier 2: JSON-LD JobPosting crawler |
| 410 | # --------------------------------------------------------------------------- # |
| 411 | |
| 412 | _JSONLD_RE = re.compile( |
| 413 | r'<script[^>]+type=["\']application/ld\+json["\'][^>]*>(.*?)</script>', |
| 414 | re.IGNORECASE | re.DOTALL, |
| 415 | ) |
| 416 | |
| 417 | |
| 418 | def extract_jsonld_jobs(html: str, base_url: str = "") -> list[dict[str, Any]]: |
| 419 | """Extract schema.org JobPosting objects embedded as JSON-LD in a page.""" |
| 420 | if not html: |
| 421 | return [] |
| 422 | domain = _domain(base_url) |
| 423 | items: list[dict[str, Any]] = [] |
| 424 | seen: set[str] = set() |
| 425 | for block in _JSONLD_RE.findall(html): |
| 426 | for obj in _walk_jsonld(block): |
| 427 | if not isinstance(obj, dict): |
| 428 | continue |
| 429 | if not _is_job_posting(obj): |
| 430 | continue |
| 431 | title = str(obj.get("title") or obj.get("name") or "").strip() |
| 432 | if not title or title in seen: |
| 433 | continue |
| 434 | seen.add(title) |
| 435 | posting_url = str(obj.get("url") or "").strip() |
| 436 | items.append({ |
| 437 | "id": f"JL{len(items) + 1}", |
| 438 | "title": title, |
| 439 | "url": posting_url, |
| 440 | "description": _clean_html(str(obj.get("description") or ""))[:2000], |
| 441 | "date": _date_part(obj.get("datePosted")), |
| 442 | "date_confidence": "high" if obj.get("datePosted") else "low", |
| 443 | "provider": "careers-jsonld", |
| 444 | "department": str(obj.get("occupationalCategory") or "").strip(), |
| 445 | "location": _jsonld_location(obj), |
| 446 | "source_url": base_url, |
| 447 | "source_domain": domain, |
| 448 | "relevance": 0.6, |
| 449 | "why_relevant": "JobPosting structured data on careers page", |
| 450 | }) |
| 451 | return items |
| 452 | |
| 453 | |
| 454 | def _walk_jsonld(block: str) -> list[Any]: |
| 455 | try: |
| 456 | data = json.loads(block.strip()) |
| 457 | except (json.JSONDecodeError, ValueError): |
| 458 | return [] |
| 459 | out: list[Any] = [] |
| 460 | stack = [data] |
| 461 | while stack: |
| 462 | node = stack.pop() |
| 463 | if isinstance(node, list): |
| 464 | stack.extend(node) |
| 465 | elif isinstance(node, dict): |
| 466 | out.append(node) |
| 467 | graph = node.get("@graph") |
| 468 | if isinstance(graph, list): |
| 469 | stack.extend(graph) |
| 470 | return out |
| 471 | |
| 472 | |
| 473 | def _is_job_posting(obj: dict[str, Any]) -> bool: |
| 474 | type_field = obj.get("@type") |
| 475 | if isinstance(type_field, str): |
| 476 | return type_field.lower() == "jobposting" |
| 477 | if isinstance(type_field, list): |
| 478 | return any(isinstance(t, str) and t.lower() == "jobposting" for t in type_field) |
| 479 | return False |
| 480 | |
| 481 | |
| 482 | def _jsonld_location(obj: dict[str, Any]) -> str: |
| 483 | loc = obj.get("jobLocation") |
| 484 | if isinstance(loc, list): |
| 485 | loc = loc[0] if loc else None |
| 486 | if isinstance(loc, dict): |
| 487 | address = loc.get("address") |
| 488 | if isinstance(address, dict): |
| 489 | return _join_parts( |
| 490 | str(address.get("addressLocality") or ""), |
| 491 | str(address.get("addressRegion") or ""), |
| 492 | str(address.get("addressCountry") or ""), |
| 493 | ) |
| 494 | if obj.get("jobLocationType"): |
| 495 | return str(obj.get("jobLocationType")).strip() |
| 496 | return "" |
| 497 | |
| 498 | |
| 499 | # --------------------------------------------------------------------------- # |
| 500 | # Tier 3: generic web search fallback |
| 501 | # --------------------------------------------------------------------------- # |
| 502 | |
| 503 | def search_jobs_web( |
| 504 | company: str, |
| 505 | date_range: tuple[str, str], |
| 506 | config: dict[str, Any], |
| 507 | *, |
| 508 | backend: str = "auto", |
| 509 | ) -> tuple[list[dict[str, Any]], dict[str, Any]]: |
| 510 | """Fallback to configured web search for public careers/jobs pages.""" |
| 511 | if backend == "none": |
| 512 | return [], {} |
| 513 | query = f'{company} careers jobs hiring' |
| 514 | raw_items, artifact = grounding.web_search(query, date_range, config, backend=backend) |
| 515 | items: list[dict[str, Any]] = [] |
| 516 | for index, raw in enumerate(raw_items): |
| 517 | if not isinstance(raw, dict): |
| 518 | continue |
| 519 | title = str(raw.get("title") or "").strip() |
| 520 | url = str(raw.get("url") or "").strip() |
| 521 | snippet = str(raw.get("snippet") or "").strip() |
| 522 | if not _looks_like_jobs_page(title, url, snippet): |
| 523 | continue |
| 524 | items.append({ |
| 525 | "id": raw.get("id") or f"JW{index + 1}", |
| 526 | "title": title, |
| 527 | "url": url, |
| 528 | "description": snippet, |
| 529 | "date": raw.get("date"), |
| 530 | "date_confidence": raw.get("date_confidence") or "low", |
| 531 | "provider": "web", |
| 532 | "source_domain": raw.get("source_domain"), |
| 533 | "relevance": raw.get("relevance", 0.45), |
| 534 | "why_relevant": "Public careers/jobs web result", |
| 535 | }) |
| 536 | artifact = dict(artifact or {}) |
| 537 | artifact.update({"label": "jobs-web", "company": company, "resultCount": len(items)}) |
| 538 | return items, artifact |
| 539 | |
| 540 | |
| 541 | # --------------------------------------------------------------------------- # |
| 542 | # Item builders + helpers |
| 543 | # --------------------------------------------------------------------------- # |
| 544 | |
| 545 | def _ats_item( |
| 546 | *, |
| 547 | provider: str, |
| 548 | slug: str, |
| 549 | ident: str, |
| 550 | title: str, |
| 551 | url: str, |
| 552 | description: str, |
| 553 | date: str | None, |
| 554 | department: str, |
| 555 | location: str, |
| 556 | ) -> dict[str, Any]: |
| 557 | prefix = { |
| 558 | ATS_PROVIDER_ASHBY: "AB", |
| 559 | ATS_PROVIDER_LEVER: "LV", |
| 560 | ATS_PROVIDER_WORKABLE: "WK", |
| 561 | ATS_PROVIDER_SMARTRECRUITERS: "SR", |
| 562 | }.get(provider, "AT") |
| 563 | return { |
| 564 | "id": f"{prefix}{ident or slug}", |
| 565 | "title": title, |
| 566 | "url": url, |
| 567 | "description": description, |
| 568 | "date": date, |
| 569 | "date_confidence": "high" if date else "low", |
| 570 | "provider": provider, |
| 571 | "board_token": slug, |
| 572 | "department": department, |
| 573 | "departments": [department] if department else [], |
| 574 | "location": location, |
| 575 | "offices": [location] if location else [], |
| 576 | "relevance": 0.75, |
| 577 | "why_relevant": f"Public {provider} job posting", |
| 578 | } |
| 579 | |
| 580 | |
| 581 | def _greenhouse_job_to_item(job: dict[str, Any], board_token: str) -> dict[str, Any]: |
| 582 | departments = [ |
| 583 | str(dept.get("name") or "").strip() |
| 584 | for dept in (job.get("departments") or []) |
| 585 | if isinstance(dept, dict) and str(dept.get("name") or "").strip() |
| 586 | ] |
| 587 | offices = [ |
| 588 | str(office.get("name") or office.get("location") or "").strip() |
| 589 | for office in (job.get("offices") or []) |
| 590 | if isinstance(office, dict) and str(office.get("name") or office.get("location") or "").strip() |
| 591 | ] |
| 592 | location = "" |
| 593 | if isinstance(job.get("location"), dict): |
| 594 | location = str(job["location"].get("name") or "").strip() |
| 595 | description = _clean_html(str(job.get("content") or "")) |
| 596 | return { |
| 597 | "id": f"GH{job.get('id') or job.get('internal_job_id') or board_token}", |
| 598 | "title": str(job.get("title") or "").strip(), |
| 599 | "url": str(job.get("absolute_url") or "").strip(), |
| 600 | "description": description, |
| 601 | "date": _date_part(job.get("updated_at")), |
| 602 | "date_confidence": "high" if job.get("updated_at") else "low", |
| 603 | "provider": ATS_PROVIDER_GREENHOUSE, |
| 604 | "board_token": board_token, |
| 605 | "department": departments[0] if departments else "", |
| 606 | "departments": departments, |
| 607 | "location": location, |
| 608 | "offices": offices, |
| 609 | "relevance": 0.75, |
| 610 | "why_relevant": "Public Greenhouse job posting", |
| 611 | } |
| 612 | |
| 613 | |
| 614 | def _artifact( |
| 615 | label: str, |
| 616 | company: str, |
| 617 | attempted: list[str], |
| 618 | items: list[dict[str, Any]], |
| 619 | explicit: bool, |
| 620 | *, |
| 621 | tier: str, |
| 622 | provider: str = "", |
| 623 | slug: str = "", |
| 624 | ) -> dict[str, Any]: |
| 625 | return { |
| 626 | "label": label, |
| 627 | "company": company, |
| 628 | "attempted": attempted, |
| 629 | "resultCount": len(items), |
| 630 | "explicit": explicit, |
| 631 | "tier": tier, |
| 632 | "provider": provider, |
| 633 | "board_token": slug, |
| 634 | } |
| 635 | |
| 636 | |
| 637 | def _candidate_slugs(company: str) -> list[str]: |
| 638 | base = _company_slug(company) |
| 639 | if not base: |
| 640 | return [] |
| 641 | candidates = [base] |
| 642 | compact = re.sub(r"(inc|labs|ai|hq|app|tech)$", "", base) |
| 643 | if compact and compact != base: |
| 644 | candidates.append(compact) |
| 645 | # hyphenated form (e.g. "listen labs" -> "listen-labs") |
| 646 | hyphen = re.sub(r"[^a-z0-9]+", "-", company.lower()).strip("-") |
| 647 | if hyphen and hyphen not in candidates: |
| 648 | candidates.append(hyphen) |
| 649 | deduped: list[str] = [] |
| 650 | for token in candidates: |
| 651 | if token and token not in deduped: |
| 652 | deduped.append(token) |
| 653 | return deduped[:4] |
| 654 | |
| 655 | |
| 656 | def _join_parts(*parts: str) -> str: |
| 657 | """Join non-empty, stripped location parts as 'City, Country'.""" |
| 658 | return ", ".join(part.strip() for part in parts if part and part.strip()) |
| 659 | |
| 660 | |
| 661 | def _company_slug(company: str) -> str: |
| 662 | text = company.lower() |
| 663 | text = re.sub(r"\b(inc|inc\.|llc|ltd|corp|corporation|company|co\.)\b", "", text) |
| 664 | return re.sub(r"[^a-z0-9]+", "", text).strip() |
| 665 | |
| 666 | |
| 667 | def _domain(url: str) -> str: |
| 668 | try: |
| 669 | return urlparse(url).netloc.lower().lstrip("www.") |
| 670 | except ValueError: |
| 671 | return "" |
| 672 | |
| 673 | |
| 674 | def _date_part(value: Any) -> str | None: |
| 675 | text = str(value or "") |
| 676 | if not text: |
| 677 | return None |
| 678 | match = re.search(r"\d{4}-\d{2}-\d{2}", text) |
| 679 | return match.group(0) if match else None |
| 680 | |
| 681 | |
| 682 | def _epoch_ms_to_date(value: Any) -> str | None: |
| 683 | try: |
| 684 | ms = int(value) |
| 685 | except (TypeError, ValueError): |
| 686 | return None |
| 687 | if ms <= 0: |
| 688 | return None |
| 689 | return dates.timestamp_to_date(ms / 1000) |
| 690 | |
| 691 | |
| 692 | def _clean_html(value: str) -> str: |
| 693 | value = unescape(value) |
| 694 | value = re.sub(r"<br\s*/?>", "\n", value, flags=re.I) |
| 695 | value = re.sub(r"</p\s*>", "\n", value, flags=re.I) |
| 696 | value = re.sub(r"<[^>]+>", " ", value) |
| 697 | value = re.sub(r"\s+", " ", value) |
| 698 | return value.strip() |
| 699 | |
| 700 | |
| 701 | def _looks_like_careers_html(html: str) -> bool: |
| 702 | lowered = html.lower() |
| 703 | if any(token in lowered for token in ("ashbyhq.com", "greenhouse.io", "lever.co", "workable.com", "smartrecruiters.com")): |
| 704 | return True |
| 705 | return bool(re.search(r"\b(open roles|open positions|join (our|the) team|current openings|jobposting)\b", lowered)) |
| 706 | |
| 707 | |
| 708 | def _looks_like_jobs_page(title: str, url: str, snippet: str) -> bool: |
| 709 | haystack = " ".join([title, url, snippet]).lower() |
| 710 | return bool(re.search(r"\b(careers?|jobs?|job openings?|hiring|greenhouse|lever|ashby|workable)\b", haystack)) |
| 711 |