Native DeepSeek datasheet extraction in periscope/src (Fase C3).

Pipeline and LCSC resolve call datasheet_extract; inherited extraction.py
stays in dependency as fallback. No Anthropic Console skill ids.

tests/test_native_datasheet_extract.py

@@ -0,0 +1,31 @@
+"""Native datasheet_extract vs inherited extraction — prove before switching pipeline."""
+
+from __future__ import annotations
+
+import ast
+from pathlib import Path
+
+from backend.services.datasheet_extract import _coerce_abs_max
+from backend.services.extraction import _coerce_abs_max as inherited_coerce
+
+
+def test_coerce_abs_max_matches_inherited():
+    raw = [
+        {"parameter": "Vin", "min": 0, "max": 6, "unit": "V", "source_page": 4},
+        "not-a-dict",
+    ]
+    assert _coerce_abs_max(raw) == inherited_coerce(raw)
+    assert _coerce_abs_max(None) == inherited_coerce(None)
+
+
+def test_native_extract_has_no_anthropic_import():
+    path = Path(__import__("backend.services.datasheet_extract", fromlist=["x"]).__file__)
+    tree = ast.parse(path.read_text())
+    imported = []
+    for node in ast.walk(tree):
+        if isinstance(node, ast.ImportFrom) and node.module:
+            imported.append(node.module)
+        elif isinstance(node, ast.Import):
+            imported.extend(a.name for a in node.names)
+    assert not any("anthropic" in (m or "") for m in imported)
+    assert "backend.services.llm.factory" in imported

tests/test_passive_from_distributor.py

@@ -86,12 +86,12 @@ def test_auto_resolve_skips_llm_when_catalog_parses(monkeypatch):
     import asyncio
     from pathlib import Path
 
-    from backend.services.extraction import auto_resolve_specs
+    from backend.services.datasheet_extract import auto_resolve_specs
 
     async def boom(*_a, **_k):
         raise AssertionError("LLM must not run")
 
-    monkeypatch.setattr("backend.services.extraction.call_with_fallback", boom)
+    monkeypatch.setattr("backend.services.datasheet_extract.call_with_fallback", boom)
     tax = TAXONOMY
     model = asyncio.run(
         auto_resolve_specs(
@@ -110,12 +110,12 @@ def test_auto_resolve_use_llm_false_on_ferrite(monkeypatch):
     import asyncio
     from pathlib import Path
 
-    from backend.services.extraction import auto_resolve_specs
+    from backend.services.datasheet_extract import auto_resolve_specs
 
     async def boom(*_a, **_k):
         raise AssertionError("LLM must not run")
 
-    monkeypatch.setattr("backend.services.extraction.call_with_fallback", boom)
+    monkeypatch.setattr("backend.services.datasheet_extract.call_with_fallback", boom)
     tax = TAXONOMY
     model = asyncio.run(
         auto_resolve_specs(

tests/test_passive_from_value.py

@@ -52,12 +52,12 @@ def test_ambiguous_string_returns_none():
 
 
 def test_resolve_from_value_skips_llm_when_parseable(monkeypatch):
-    from backend.services.extraction import resolve_from_value
+    from backend.services.datasheet_extract import resolve_from_value
 
     async def boom(*_a, **_k):
         raise AssertionError("LLM must not run")
 
-    monkeypatch.setattr("backend.services.extraction.call_with_fallback", boom)
+    monkeypatch.setattr("backend.services.datasheet_extract.call_with_fallback", boom)
     tax = TAXONOMY
     model = asyncio.run(
         resolve_from_value(
@@ -71,12 +71,12 @@ def test_resolve_from_value_skips_llm_when_parseable(monkeypatch):
 
 
 def test_resolve_from_value_placeholder_no_llm(monkeypatch):
-    from backend.services.extraction import resolve_from_value
+    from backend.services.datasheet_extract import resolve_from_value
 
     async def boom(*_a, **_k):
         raise AssertionError("LLM must not run")
 
-    monkeypatch.setattr("backend.services.extraction.call_with_fallback", boom)
+    monkeypatch.setattr("backend.services.datasheet_extract.call_with_fallback", boom)
     tax = TAXONOMY
     with pytest.raises(ValueError, match="Placeholder"):
         asyncio.run(

tests/test_pdf_ingest.py

@@ -4,7 +4,7 @@ from __future__ import annotations
 
 from pathlib import Path
 
-from backend.services.extraction import _coerce_abs_max
+from backend.services.datasheet_extract import _coerce_abs_max
 from backend.services.llm.pdf_ingest import (
     extract_pdf_text,
     make_text_pdf,

tests/test_purple_parts_resolver.py

@@ -452,7 +452,7 @@ async def test_lcsc_resolve_passive_success_and_cached(tmp_path, monkeypatch):
     from backend.config import settings
     from backend.periscopex.models import CapacitorSpecs, ComponentModel
     from backend.services import purple_parts
-    from backend.services import extraction as extraction_svc
+    from backend.services import datasheet_extract as extraction_svc
 
     # Billing modules are absent in the open-source repo; the endpoint under
     # test only charges when billing is enabled, so skip there.