Fix warnings, LinkedIn selector update, Telegram notify listing, usajobs title filter

- LinkedIn: new URL with geoId, updated selectors to li[data-occludable-job-id], scroll adapter, pagination disabled
- usajobs: add include_title_keywords post-filter (dentist/orthodontist/orthodontic)
- aroragroup: fix AJAX load via doloadJBSearchList(), fix container selector
- govtjobs/srpmic: clear result_list_wait_selector to avoid 15s timeout on 0-result pages
- orchestrator: wait selector timeout WARNING → DEBUG
- extractor: no-container WARNING → DEBUG
- notifier: provider-based dispatch (telegram/kakaotalk), build_listing_messages() for Telegram
- cli notify: send job listing with links instead of Ollama summary
- global.yaml: provider set to telegram

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

gimme_job/adapters/aroragroup.py

@@ -16,11 +16,19 @@ class AroraGroupAdapter(ManifestDrivenAdapter):
     _BASE_URL = "https://jobs.aroragroup.com"
 
     def collect_cards(self, page, manifest):
-        """Load initial results, click 'Load More' up to N times, then filter by keywords."""
+        """Trigger AJAX job load, click 'Load More' up to N times, then filter by keywords."""
+        # Trigger initial job list load via the page's JS function
         try:
-            page.wait_for_selector(".job-post-row", timeout=15000, state="attached")
+            page.evaluate("doloadJBSearchList(1)")
         except Exception:
-            logger.debug("[aroragroup] Initial results wait timed out")
+            pass
+
+        # Wait for at least one title to be populated
+        try:
+            page.wait_for_selector(".POST_TITLE:not(:empty)", timeout=15000, state="attached")
+        except Exception:
+            logger.debug("[aroragroup] Job titles did not populate — no results or load failed")
+            return []
 
         for i in range(_MAX_LOAD_MORE):
             btn = page.query_selector("button#loadMore")
@@ -28,12 +36,12 @@ class AroraGroupAdapter(ManifestDrivenAdapter):
                 logger.debug(f"[aroragroup] Load More button gone after {i} extra loads")
                 break
 
-            current_count = len(page.query_selector_all(".job-post-row[onclick]"))
+            current_count = len(page.query_selector_all(".POST_TITLE:not(:empty)"))
             btn.click()
 
             try:
                 page.wait_for_function(
-                    f"document.querySelectorAll('.job-post-row[onclick]').length > {current_count}",
+                    f"document.querySelectorAll('.POST_TITLE:not(:empty)').length > {current_count}",
                     timeout=15000,
                 )
                 logger.debug(f"[aroragroup] Load More {i + 1}/{_MAX_LOAD_MORE}: loaded more cards")

gimme_job/adapters/linkedin.py

@@ -0,0 +1,34 @@
+"""LinkedIn Jobs adapter — scroll to load all cards on first page only."""
+from __future__ import annotations
+
+from loguru import logger
+
+from gimme_job.adapters.base import ManifestDrivenAdapter
+from gimme_job.adapters.registry import register
+from gimme_job.models.dto import JobPostingCandidate, RawJobCard
+
+
+@register("linkedin")
+class LinkedInAdapter(ManifestDrivenAdapter):
+
+    def collect_cards(self, page, manifest) -> list[RawJobCard]:
+        """Scroll each card into view to trigger lazy-loading, then extract."""
+        self._scroll_to_load(page)
+        return super().collect_cards(page, manifest)
+
+    def normalize(self, raw: RawJobCard) -> JobPostingCandidate:
+        return super().normalize(raw)
+
+    @staticmethod
+    def _scroll_to_load(page) -> None:
+        """Scroll each job card into view to force LinkedIn's virtual list to render."""
+        try:
+            page.evaluate("""() => {
+                const cards = Array.from(document.querySelectorAll('li[data-occludable-job-id]'));
+                cards.forEach(card => card.scrollIntoView({ behavior: 'instant', block: 'center' }));
+            }""")
+            page.wait_for_timeout(800)
+            page.evaluate("window.scrollTo(0, 0)")
+            page.wait_for_timeout(400)
+        except Exception as e:
+            logger.debug(f"[linkedin] scroll_to_load failed: {e}")

gimme_job/cli.py

@@ -5,6 +5,9 @@ import sys
 from pathlib import Path
 from typing import Optional
 
+import dotenv
+dotenv.load_dotenv()
+
 import typer
 from rich.console import Console
 from rich.panel import Panel
@@ -327,15 +330,16 @@ def test(
 
 @app.command()
 def notify(
-    today: bool = typer.Option(True, "--today/--no-today", help="Re-send today's digest"),
+    today: bool = typer.Option(True, "--today/--no-today", help="Re-send today's job listing"),
 ) -> None:
-    """Re-send today's job summary via Telegram."""
+    """Send today's new job listings via Telegram (one message per site group)."""
     from datetime import date
 
     from gimme_job.config import load_global_config
     from gimme_job.db.engine import get_engine, get_session_factory
-    from gimme_job.db.repo import JobPostingRepo, SummaryRepo
+    from gimme_job.db.repo import JobPostingRepo
-    from gimme_job.runtime.notifier import NotificationDispatcher
+    from gimme_job.runtime.notifier import NotificationDispatcher, build_listing_messages
+    from gimme_job.runtime.telegram import TelegramClient
 
     cfg = load_global_config()
     engine = get_engine()
@@ -343,31 +347,33 @@ def notify(
     run_date = date.today()
 
     with factory() as session:
-        summary_repo = SummaryRepo()
-        summary = summary_repo.get_latest(session, run_date)
-
-        if summary:
-            text = summary.content
-        else:
-            # Build summary from today's postings
         postings = JobPostingRepo().get_today_new(session, run_date)
+
     if not postings:
         console.print("[yellow]No new postings found for today.[/yellow]")
         raise typer.Exit(0)
-            from gimme_job.runtime.summarizer import OllamaSummarizer
-            summarizer = OllamaSummarizer(
-                base_url=cfg.summarization.ollama_base_url,
-                model=cfg.summarization.model,
-                temperature=cfg.summarization.temperature,
-            )
-            text = summarizer.summarize(postings)
 
-    dispatcher = NotificationDispatcher(global_config=cfg, session_factory=factory)
+    chunks = build_listing_messages(postings)
-    ok = dispatcher.send(text, run_date)
+    console.print(f"Sending {len(chunks)} message(s) for {len(postings)} postings...")
-    if ok:
+
-        console.print("[green]✓[/green] Notification sent")
+    client = TelegramClient()
+    sent = 0
+    if client.is_configured():
+        for chunk in chunks:
+            if client.send_message(chunk):
+                sent += 1
+        if sent == len(chunks):
+            console.print(f"[green]✓[/green] All {sent} message(s) sent")
+        else:
+            console.print(f"[yellow]![/yellow] Sent {sent}/{len(chunks)} messages")
     else:
-        console.print("[yellow]![/yellow] Notification failed — saved as markdown fallback")
+        console.print("[yellow]Telegram not configured[/yellow]")
+
+    # Always write markdown fallback
+    dispatcher = NotificationDispatcher(global_config=cfg, session_factory=factory)
+    full_text = "\n\n".join(chunks)
+    path = dispatcher.send_markdown_fallback(full_text, run_date)
+    console.print(f"Markdown saved: {path}")
 
 
 # ── list ──────────────────────────────────────────────────────────────────────

gimme_job/models/manifest.py

@@ -53,6 +53,7 @@ class ExtractConfig(BaseModel):
 
 class PostFilterConfig(BaseModel):
     include_posted_text: list[str] = Field(default_factory=list)
+    include_title_keywords: list[str] = Field(default_factory=list)
 
 
 class HealthcheckConfig(BaseModel):

gimme_job/runtime/extractor.py

@@ -60,7 +60,7 @@ def extract_cards_from_page(page: "Page", extract_config: ExtractConfig) -> list
             continue
 
     if not container_elements:
-        logger.warning("No container elements found with any selector")
+        logger.debug("No container elements found with any selector")
         return cards
 
     for element in container_elements:

gimme_job/runtime/notifier.py

@@ -1,12 +1,73 @@
-"""Notification dispatcher: Telegram with Markdown fallback."""
+"""Notification dispatcher: Telegram or KakaoTalk with Markdown fallback."""
 from __future__ import annotations
 
 from datetime import date
+from typing import TYPE_CHECKING
 
 from loguru import logger
 
 from gimme_job.config import GlobalConfig
 
+if TYPE_CHECKING:
+    from gimme_job.models.db import JobPosting
+
+_TELEGRAM_MAX_LEN = 4000  # leave room for safety margin
+
+
+def build_listing_messages(postings: list["JobPosting"]) -> list[str]:
+    """Format job postings as Telegram HTML chunks (≤4000 chars each)."""
+    from collections import defaultdict
+
+    by_site: dict[str, list] = defaultdict(list)
+    for p in postings:
+        by_site[p.site_id].append(p)
+
+    total = len(postings)
+    header = f"📋 <b>오늘의 신규 채용 공고 ({total}개)</b>\n"
+
+    chunks: list[str] = []
+    current = header
+
+    for site_id, site_postings in sorted(by_site.items()):
+        site_block = f"\n<b>── {site_id.upper()} ({len(site_postings)}) ──</b>\n"
+        for p in site_postings:
+            # Title line
+            title = _escape_html(p.title)
+            line = f"• <b>{title}</b>"
+
+            # Details line
+            details = []
+            if p.company:
+                details.append(_escape_html(p.company))
+            if p.location:
+                details.append(_escape_html(p.location))
+            if p.employment_type:
+                details.append(_escape_html(p.employment_type))
+            if details:
+                line += f"\n  {' | '.join(details)}"
+
+            # Link
+            if p.job_url:
+                line += f'\n  <a href="{p.job_url}">링크</a>'
+
+            site_block += line + "\n"
+
+        # Flush chunk if adding this site would exceed limit
+        if len(current) + len(site_block) > _TELEGRAM_MAX_LEN:
+            chunks.append(current.rstrip())
+            current = site_block
+        else:
+            current += site_block
+
+    if current.strip():
+        chunks.append(current.rstrip())
+
+    return chunks or ["신규 채용 공고가 없습니다."]
+
+
+def _escape_html(text: str) -> str:
+    return text.replace("&", "&amp;").replace("<", "&lt;").replace(">", "&gt;")
+
 
 class NotificationDispatcher:
     def __init__(self, global_config: GlobalConfig, session_factory):
@@ -15,11 +76,27 @@ class NotificationDispatcher:
 
     def send(self, summary: str, run_date: date) -> bool:
         """Send the summary. Returns True if any method succeeded."""
-        from gimme_job.runtime.telegram import TelegramClient
+        provider = self.cfg.notification.provider
-
-        client = TelegramClient()
         success = False
 
+        if provider == "kakaotalk":
+            from gimme_job.runtime.kakao import KakaoTalkClient
+            client = KakaoTalkClient()
+            if client.is_configured():
+                ok = client.send_self_memo(summary)
+                if ok:
+                    success = True
+                    self._log_notification(run_date, "kakaotalk", "success")
+                else:
+                    logger.warning("KakaoTalk failed — writing markdown fallback")
+                    self._log_notification(run_date, "kakaotalk", "failed", "send_self_memo returned False")
+            else:
+                logger.warning("KakaoTalk not configured — writing markdown fallback only")
+                self._log_notification(run_date, "kakaotalk", "skipped", "not configured")
+
+        else:  # telegram (default)
+            from gimme_job.runtime.telegram import TelegramClient
+            client = TelegramClient()
             if client.is_configured():
                 ok = client.send_message(summary)
                 if ok:

gimme_job/runtime/orchestrator.py

@@ -196,7 +196,7 @@ class RunOrchestrator:
                         state="attached",
                     )
                 except Exception:
-                    logger.warning(f"[{site_id}] Wait selector timed out — continuing anyway")
+                    logger.debug(f"[{site_id}] Wait selector timed out — continuing anyway")
 
             # Check if site has existing data (for early stop on update runs)
             site_has_data = False
@@ -292,14 +292,24 @@ class RunOrchestrator:
 
         # Apply post-filters
         if manifest.post_filters.include_posted_text:
-            from gimme_job.models.dto import RawJobCard
-            # Re-filter candidates based on posted_text
             include_lower = [t.lower() for t in manifest.post_filters.include_posted_text]
             candidates = [
                 c for c in candidates
                 if not c.posted_text or any(t in (c.posted_text or "").lower() for t in include_lower)
             ]
 
+        if manifest.post_filters.include_title_keywords:
+            kws = [t.lower() for t in manifest.post_filters.include_title_keywords]
+            before = len(candidates)
+            candidates = [
+                c for c in candidates
+                if any(kw in (c.title or "").lower() for kw in kws)
+            ]
+            logger.info(
+                f"[{site_id}] title keyword filter: {len(candidates)}/{before} kept "
+                f"(keywords: {manifest.post_filters.include_title_keywords})"
+            )
+
         # Ensure fingerprints and deduplicate within batch
         candidates = ensure_fingerprints(candidates)
         candidates = deduplicate_in_batch(candidates)

sites/aroragroup.yaml

@@ -27,7 +27,7 @@ pagination:
 
 extract:
   container_selectors:
-    - ".job-post-row[onclick]"
+    - ".job-post-row"
   fields:
     title:
       text:

sites/global.yaml

@@ -25,5 +25,5 @@ summarization:
   max_input_items: 200
 
 notification:
-  provider: kakaotalk
+  provider: telegram
   fallback_markdown: true

sites/govtjobs.yaml

@@ -19,7 +19,7 @@ search:
   location_mode: none
   sort_mode: none
   date_mode: none
-  result_list_wait_selector: "h3 a[href*='/jobs/']"
+  result_list_wait_selector: ""
 
 search_override:
   keywords:

sites/linkedin.yaml

@@ -6,7 +6,7 @@ identity:
   label: LinkedIn Jobs
   category: aggregator
   base_url: https://www.linkedin.com
-  start_url_template: "https://www.linkedin.com/jobs/search/?keywords={keywords_urlencoded}&f_TPR=r86400&sortBy=DD"
+  start_url_template: "https://www.linkedin.com/jobs/search/?keywords={keywords_urlencoded}&geoId=103644278&origin=JOB_SEARCH_PAGE_LOCATION_AUTOCOMPLETE"
 
 browser:
   profile_mode: persistent_chrome_profile
@@ -19,45 +19,28 @@ search:
   location_mode: url_param
   sort_mode: url_param
   date_mode: url_param
-  result_list_wait_selector: ".jobs-search__results-list li, .scaffold-layout__list-container li"
+  result_list_wait_selector: "li[data-occludable-job-id]"
 
 pagination:
-  mode: next_button
+  mode: none
-  next_button_selectors:
+  max_pages: 1
-    - "button[aria-label='View next page']"
-    - "button[aria-label*='next']"
-    - "li.artdeco-pagination__indicator--number.selected + li button"
-  max_pages: 3
 
 extract:
   container_selectors:
-    - ".jobs-search__results-list li"
+    - "li[data-occludable-job-id]"
-    - ".scaffold-layout__list-container li"
-    - "li:has(.base-card)"
   fields:
     title:
       text:
-        - "h3.base-search-card__title"
+        - "a.job-card-list__title--link"
-        - ".job-card-list__title"
-        - "h3"
-        - "a[data-control-name='job_card_title']"
     company:
       text:
-        - "h4.base-search-card__subtitle"
+        - ".artdeco-entity-lockup__subtitle span"
-        - ".job-card-container__primary-description"
-        - "a.job-card-container__company-name"
     location:
       text:
-        - "span.job-search-card__location"
+        - ".job-card-container__metadata-wrapper li span"
-        - ".job-card-container__metadata-item"
-    posted_text:
-      text:
-        - "time"
-        - ".job-search-card__listdate"
-        - "span[class*='listdate']"
     url:
       attr:
-        selector: "a.base-card__full-link, a[data-control-name='job_card_title'], a[href*='/jobs/view/']"
+        selector: "a.job-card-list__title--link"
         name: href
 
 post_filters:

sites/srpmic.yaml

@@ -19,7 +19,7 @@ search:
   location_mode: none
   sort_mode: none
   date_mode: none
-  result_list_wait_selector: "a[href*='/careers/srpmic/jobs/']"
+  result_list_wait_selector: ""
 
 search_override:
   keywords:

sites/usajobs.yaml

@@ -64,6 +64,10 @@ search_override:
 
 post_filters:
   include_posted_text: []
+  include_title_keywords:
+    - dentist
+    - orthodontist
+    - orthodontic
 
 healthcheck:
   min_items_expected: 0