@pentatonic-ai/ai-agent-sdk 0.9.6 → 0.10.0

package/packages/memory-engine-v2/extractor-async/test_noise_filter.py

@@ -0,0 +1,177 @@
+"""Tests for noise_filter — entity-name junk drop list.
+
+Cases anchored to live-data audit of pentatonic-team (2026-05-25).
+Every reject case below corresponds to a real entity row we found
+in production org_model that should never have been there.
+
+Run: pytest packages/memory-engine-v2/extractor-async/test_noise_filter.py
+"""
+
+from __future__ import annotations
+
+from noise_filter import is_noise_entity_name
+
+
+# --------------------------------------------------------------------
+# Reject cases — these landed in prod and should never recur
+# --------------------------------------------------------------------
+
+
+class TestPronounsAreDropped:
+    """First-person + second-person pronouns the LLM treats as entities."""
+
+    def test_first_person_singular(self):
+        for name in ("i", "I", "me", "Me", "my", "myself"):
+            assert is_noise_entity_name("person", name), name
+
+    def test_second_person(self):
+        for name in ("you", "You", "your", "yourself"):
+            assert is_noise_entity_name("person", name), name
+
+    def test_first_person_plural(self):
+        for name in ("we", "We", "us", "ourselves"):
+            assert is_noise_entity_name("person", name), name
+
+
+class TestGenericReferentsAreDropped:
+    """LLM fallbacks when there's no real proper noun in scope."""
+
+    def test_agent_and_user_referents(self):
+        for name in ("agent", "the user", "User", "Speaker", "the speaker"):
+            assert is_noise_entity_name("person", name), name
+
+    def test_descriptive_phrases(self):
+        # Live-data examples: "the guy who wrote it", "both of you".
+        # Anything lowercased with >=4 tokens.
+        for name in ("the guy who wrote it", "both of you and me"):
+            assert is_noise_entity_name("person", name), name
+
+
+class TestTailscaleHostnamesAreDropped:
+    """Tailscale hosts that leak from agent scratchpads."""
+
+    def test_classic_pin_dev_office(self):
+        # Live-data example.
+        assert is_noise_entity_name(
+            "other", "pin-dev-office-01.tail4aaf0c.ts.net"
+        )
+
+    def test_other_ts_net_shapes(self):
+        for name in (
+            "laptop-mossop.tail123.ts.net",
+            "MyMachine.TS.net",  # case-insensitive
+        ):
+            assert is_noise_entity_name("other", name), name
+
+
+class TestFilesystemPathsAreDropped:
+    """File paths the LLM extracts from agent output."""
+
+    def test_absolute_unix_paths(self):
+        for name in (
+            "/Users/admin/cursor/Pip",
+            "/etc/passwd",
+            "/opt/engine-v2",
+        ):
+            assert is_noise_entity_name("other", name), name
+
+    def test_repo_relative_paths(self):
+        for name in (
+            "src/memory-engine/compat/server.py",
+            "packages/memory/src/engine.js",
+        ):
+            assert is_noise_entity_name("other", name), name
+
+
+class TestAgentWorktreeLabelsAreDropped:
+    """Parallel-agent orchestrator labels that leak into transcripts."""
+
+    def test_agent_letter_em_dash_label(self):
+        for name in (
+            "Agent A — L2 provenance taint",
+            "Agent B - cross-user dispatcher",
+            "Agent C – memory engine",
+        ):
+            assert is_noise_entity_name("person", name), name
+
+
+class TestSingleCharIsDropped:
+    """Single-char strings are never useful as entities."""
+
+    def test_single_char_letters(self):
+        for name in ("a", "I", "x", "."):
+            assert is_noise_entity_name("person", name), name
+
+    def test_empty_or_whitespace(self):
+        for name in ("", "   ", "\t\n"):
+            assert is_noise_entity_name("person", name), name
+
+
+class TestLowercasePersonsAreDropped:
+    """Person entities without any capital letter are almost always
+    junk descriptive phrases, not real names. Exception: anything
+    containing '@' (an email signal) is allowed."""
+
+    def test_lowercase_person_dropped(self):
+        for name in ("someone", "anyone", "the boss"):
+            assert is_noise_entity_name("person", name), name
+
+    def test_lowercase_email_kept(self):
+        # An email is a strong real-entity signal even when lower-case.
+        assert not is_noise_entity_name("person", "philip@pentatonic.com")
+
+
+# --------------------------------------------------------------------
+# Accept cases — real entities that must NOT be dropped
+# --------------------------------------------------------------------
+
+
+class TestRealPeopleAreKept:
+    def test_proper_names(self):
+        for name in (
+            "Philip Mossop",
+            "Phil H",
+            "Chris Wheeler",
+            "Yan Manka",
+            "Raakkel Sims",
+        ):
+            assert not is_noise_entity_name("person", name), name
+
+    def test_single_first_name(self):
+        # "Yan" is a real name — single-token title-cased entities
+        # are common in conversational extraction.
+        for name in ("Yan", "Jamie", "Philip"):
+            assert not is_noise_entity_name("person", name), name
+
+
+class TestRealCompaniesAreKept:
+    def test_brand_names(self):
+        for name in (
+            "Mastercard",
+            "Samsung",
+            "Investec",
+            "BGI Worldwide",
+            "Bethnal Green Ventures",
+        ):
+            assert not is_noise_entity_name("company", name), name
+
+
+class TestShortConceptsAreKept:
+    def test_two_char_acronyms(self):
+        # "AI", "ML", "QA" are legitimate concepts.
+        for name in ("AI", "ML", "QA", "VR", "AR"):
+            assert not is_noise_entity_name("concept", name), name
+
+    def test_capitalised_acronyms(self):
+        for name in ("OpenHands", "TES", "SaaS", "API", "JSON"):
+            assert not is_noise_entity_name("concept", name), name
+
+
+class TestNonStringInput:
+    """Defensive: the upstream might hand us None or a number."""
+
+    def test_none_drops(self):
+        assert is_noise_entity_name("person", None)  # type: ignore[arg-type]
+
+    def test_number_drops(self):
+        assert is_noise_entity_name("person", 42)  # type: ignore[arg-type]