dashboard: clickable news — detail dialogs + on-demand AI summary endpoint

- news_engine: RSS parser captures link/description/publisher (three new optional NewsSentiment fields); _clean_html_text/_safe_link helpers - news.json: headlines carry id = sha1(headline) + enrichment fields - static/news_summary.php (NEW): on-demand DeepSeek gold-impact summary, whitelist-gated against news.json ids, cache-forever in ai/, fail-closed file-based daily cap (150/day), key at /etc/gold-dashboard/deepseek.key; installed to the docroot by install_static - dashboard.html: newswire + calendar detail dialogs (score tiles, source description, article link, AI summary with loading/error states); delegated on stable containers, dialog keyed by id across polls - tests: +6 (357 total); docs: README-dashboard §6 one-time setup (key file, apache-owned ai/, SELinux httpd_can_network_connect), CLAUDE.md Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Aq1jZ6jBsBdEC2xT3u46SA
2026-08-07 09:08:17 -04:00
parent a42ef16113
commit 15a4ed7831
7 changed files with 745 additions and 33 deletions
@@ -5295,6 +5295,97 @@ def _():
        NewsEngine._source_health = saved


+@test("NewsEngine: RSS parser captures link/description/publisher")
+def _():
+    cfg = NewsConfig()
+    cfg.rss_feeds = ["https://news.google.com/rss/search?q=gold"]
+    ne = NewsEngine(cfg)
+    saved = NewsEngine._source_health
+    NewsEngine._source_health = {}
+    # Google-News-shaped item: redirect <link>, HTML-escaped <description>,
+    # <source> carrying the real publisher.
+    rss = (b"<rss><channel><item>"
+           b"<title>Gold surges as Fed signals rate cuts - Reuters</title>"
+           b"<link>https://news.google.com/rss/articles/CBMiAbC?oc=5</link>"
+           b"<description>&lt;a href=\"https://x\"&gt;Gold &amp;amp; dollar "
+           b"diverge as the Fed turns dovish.&lt;/a&gt;</description>"
+           b"<source url=\"https://www.reuters.com\">Reuters</source>"
+           b"</item></channel></rss>")
+
+    class _Sess:
+        def get(self, url, timeout=None):
+            return type("R", (), {"status_code": 200, "content": rss})()
+
+    ne._session = _Sess()
+    try:
+        heads = ne._fetch_news_sentiment_raw()
+        assert len(heads) == 1, heads
+        h = heads[0]
+        assert h.link == "https://news.google.com/rss/articles/CBMiAbC?oc=5"
+        # tags stripped, entities (incl. double-encoded &amp;amp;) decoded
+        assert h.description == "Gold & dollar diverge as the Fed turns dovish."
+        assert h.publisher == "Reuters"
+    finally:
+        NewsEngine._source_health = saved
+
+
+@test("NewsEngine: enrichment fields default None / reject unsafe values")
+def _():
+    cfg = NewsConfig()
+    cfg.rss_feeds = ["https://feed.example/rss"]
+    ne = NewsEngine(cfg)
+    saved = NewsEngine._source_health
+    NewsEngine._source_health = {}
+    long_link = b"https://a.example/" + b"x" * 2100
+    rss = (b"<rss><channel>"
+           # bare item: no link/description/source at all
+           b"<item><title>Dollar steady before Fed minutes</title></item>"
+           # javascript: link + description that merely echoes the title
+           b"<item><title>Gold slips on strong dollar data</title>"
+           b"<link>javascript:alert(1)</link>"
+           b"<description>Gold slips on strong dollar data</description></item>"
+           # over-long link
+           b"<item><title>Fed officials split on rate cuts</title>"
+           b"<link>" + long_link + b"</link></item>"
+           b"</channel></rss>")
+
+    class _Sess:
+        def get(self, url, timeout=None):
+            return type("R", (), {"status_code": 200, "content": rss})()
+
+    ne._session = _Sess()
+    try:
+        heads = ne._fetch_news_sentiment_raw()
+        assert len(heads) == 3, heads
+        by_title = {h.headline: h for h in heads}
+        bare = by_title["Dollar steady before Fed minutes"]
+        assert bare.link is None and bare.description is None \
+            and bare.publisher is None
+        js = by_title["Gold slips on strong dollar data"]
+        assert js.link is None          # non-http(s) scheme rejected
+        assert js.description is None   # identical to title → dropped
+        assert by_title["Fed officials split on rate cuts"].link is None
+    finally:
+        NewsEngine._source_health = saved
+
+
+@test("NewsEngine: _clean_html_text strips/unescapes/truncates")
+def _():
+    from news_engine import _clean_html_text, _safe_link
+    assert _clean_html_text(None) is None
+    assert _clean_html_text("") is None
+    assert _clean_html_text("<p></p>") is None
+    assert _clean_html_text("<p>Fed <b>holds</b>\n rates</p>") == "Fed holds rates"
+    assert _clean_html_text("&amp;lt;b&amp;gt;x&amp;lt;/b&amp;gt; &amp; y") == "x & y"
+    long = _clean_html_text("a" * 600)
+    assert len(long) == 501 and long.endswith("…")
+    assert _safe_link("  https://a.example/x ") == "https://a.example/x"
+    assert _safe_link("ftp://a.example/x") is None
+    assert _safe_link("javascript:alert(1)") is None
+    assert _safe_link("https://" + "x" * 2100) is None
+    assert _safe_link(None) is None
+
+
@test("/status renders news/calendar/AI health with ⚠️ degradations")
 def _():
    import telegram_bot as tb
@@ -5540,6 +5631,45 @@ def _():
    assert payload["headlines"][1]["score"] is None  # NaN → None
    assert payload["events"][0]["event"] == "ISM"
    assert payload["sentiment"] == {"XAU": 0.6, "USD": -0.1}
+    # Every headline carries the 40-hex id the PHP endpoint whitelists on.
+    import re as _re
+    assert all(_re.fullmatch(r"[0-9a-f]{40}", h["id"])
+               for h in payload["headlines"])
+
+
+@test("dashboard: headline_to_dict id matches LLM cache scheme + enrichment")
+def _():
+    import hashlib
+    h = NewsSentiment(
+        headline="Gold surges", source="http://feed/b",
+        timestamp=datetime(2026, 8, 7, 10, 0, tzinfo=timezone.utc),
+        currencies_mentioned=["XAU"], sentiment_score=0.6,
+        keywords=["surge"], sentiment_by_currency={"XAU": 0.6},
+        link="https://x.example/a", description="Body.", publisher="Reuters")
+    d = dpub.headline_to_dict(h)
+    # Same sha1(headline) scheme as NewsEngine._llm_sent_cache — the id is
+    # the join key news_summary.php validates against.
+    assert d["id"] == hashlib.sha1(b"Gold surges").hexdigest()
+    assert d["link"] == "https://x.example/a"
+    assert d["description"] == "Body."
+    assert d["publisher"] == "Reuters"
+    # Pre-enrichment NewsSentiment (defaults) still serializes cleanly.
+    bare = NewsSentiment(
+        headline="Old", source="s", timestamp=None,
+        currencies_mentioned=[], sentiment_score=0.0, keywords=[])
+    d2 = dpub.headline_to_dict(bare)
+    assert d2["link"] is None and d2["description"] is None \
+        and d2["publisher"] is None
+    json.dumps([d, d2], allow_nan=False)
+
+
+@test("dashboard: install_static installs page + PHP endpoint atomically")
+def _():
+    with tempfile.TemporaryDirectory() as tmp:
+        pub = dpub.DashboardPublisher(out_dir=tmp)
+        pub.install_static()
+        assert sorted(os.listdir(tmp)) == \
+            ["dashboard.html", "news_summary.php"]  # both installed, no .tmp


@test("dashboard: publish is atomic and leaves no temp file")