fix(scrape): make vision OCR a per-bundle allowlist, not a corpus sweep (#6)

Co-authored-by: claude <[email protected]>

scrape/bundles.py

@@ -62,6 +62,11 @@ class BundleSpec:
     platform: str | None = None
     language: str = "en-US"
     source_url: str | None = None   # overrides the default support.hpe.com URL
+    # Vision-OCR this bundle's images? Default False. Image-only data lives in
+    # only a handful of docs (the release-schedule matrix); every other image
+    # in the corpus is a product UI screenshot that OCR would just add noise
+    # for. So OCR is an explicit per-bundle opt-in, not a corpus-wide sweep.
+    ocr: bool = False
 
 
 # Pinned baseline bundles. docIds confirmed by probing the portal for
@@ -101,7 +106,7 @@ BUNDLES: list[BundleSpec] = [
     # the matrix text comes from the VISION_OCR pass (see scrape/vision.py).
     # version=None — it spans all major streams, not one release.
     BundleSpec("morpheus_release_schedule",    "sf000111242en_us", "HPE Morpheus Software Release Schedule",
-               None, "Release Schedule", "single"),
+               None, "Release Schedule", "single", ocr=True),
 ]
 
 
@@ -180,6 +185,7 @@ def discover_bundle(s: requests.Session, spec: BundleSpec) -> dict[str, Any]:
             "dates": {},
             "landing_page": spec.doc_id,
             "source_url": spec.source_url or f"https://www.hpe.com/psnow/doc/{spec.doc_id}",
+            "ocr": spec.ocr,
         }
 
     abstract_html = _get(s, f"{API}/{spec.doc_id}", expect_json=False)
@@ -211,6 +217,7 @@ def discover_bundle(s: requests.Session, spec: BundleSpec) -> dict[str, Any]:
         "dates": {"Published": meta.get("published", "")},
         "landing_page": landing,
         "source_url": spec.source_url or DOC_URL.format(doc_id=spec.doc_id),
+        "ocr": spec.ocr,
     }