Studio: require ASCII evidence for declared Latin-1/cp1252 web fetches

Latin-1 and cp1252 decode every byte to a printable character, so a high-byte
binary body declared as iso-8859-1/windows-1252 decoded cleanly and slipped
past the control-character binary check. Apply the existing ASCII-structure gate
to those declared decodes as well. Scoped to the Latin family so legitimate
non-Latin single-byte pages (Cyrillic, Greek) are not rejected.

studio/backend/core/inference/tools.py

@@ -3890,6 +3890,14 @@ def _fetch_page_text(
                 raw_html = alt
             else:
                 return f"(binary content, {len(raw_bytes)} bytes; not readable as text)"
+        elif declared_codec in ("iso8859-1", "cp1252") and not (
+            _has_single_byte_text_evidence(raw_bytes)
+        ):
+            # Latin-1/cp1252 map every byte to a printable character, so high-byte
+            # binary decodes cleanly and slips past _looks_binary; require ASCII text
+            # structure. Other single-byte charsets are left alone so legitimate
+            # non-Latin pages (Cyrillic, Greek, ...) are not rejected.
+            return f"(binary content, {len(raw_bytes)} bytes; not readable as text)"
     except _HTTPError as e:
         return f"Failed to fetch URL: HTTP {e.code} {getattr(e, 'reason', '')}"
     except Exception as e:

studio/backend/tests/test_web_fetch_binary_guard.py

@@ -264,3 +264,38 @@ def test_text_with_a_few_stray_replacement_chars_kept(monkeypatch):
     out = _fetch_with(monkeypatch, body, "text/html")
     assert "Real article text." in out
     assert "binary content" not in out
+
+
+@pytest.mark.parametrize("charset", ["iso-8859-1", "latin-1", "windows-1252", "cp1252"])
+def test_declared_latin1_high_byte_binary_rejected(monkeypatch, charset):
+    # Latin-1/cp1252 decode high bytes to printable chars, so binary evades the
+    # control-char check; the ASCII-structure gate must still reject it.
+    body = bytes(range(0xA0, 0x100)) * 40
+    out = _fetch_with(monkeypatch, body, f"text/plain; charset={charset}")
+    assert "binary content" in out
+
+
+@pytest.mark.parametrize("charset", ["iso-8859-1", "windows-1252"])
+def test_declared_latin1_real_text_still_kept(monkeypatch, charset):
+    # Genuine accented Western text is ASCII-dominated and must not be rejected.
+    body = ("Cafe un tres bon eleve a l ecole. MARKERWORD ".replace("e", "é") + "voila ").encode(
+        "cp1252"
+    ) * 40
+    out = _fetch_with(monkeypatch, body, f"text/plain; charset={charset}")
+    assert "MARKERWORD" in out
+    assert "binary content" not in out
+
+
+@pytest.mark.parametrize(
+    "charset,codec,sample",
+    [
+        ("windows-1251", "cp1251", "Это настоящая русская статья. Привет. "),
+        ("koi8-r", "koi8-r", "Это настоящая русская статья. Привет. "),
+        ("iso-8859-7", "iso-8859-7", "Αυτό είναι ένα ελληνικό άρθρο. "),
+    ],
+)
+def test_declared_non_latin_single_byte_text_kept(monkeypatch, charset, codec, sample):
+    # Cyrillic/Greek single-byte pages are high-byte dense; the Latin-1 ASCII
+    # gate must not touch them or legitimate non-Western text would be dropped.
+    out = _fetch_with(monkeypatch, (sample * 40).encode(codec), f"text/plain; charset={charset}")
+    assert "binary content" not in out