Fix research pipeline budget tracking

tests/test_crawler.py

@@ -11,6 +11,7 @@ import pytest
 
 from nsct.crawler.extraction import extract_main_content
 from nsct.crawler.fetcher import AsyncFetcher, FetchResult, FetchStatus
+from nsct.crawler.manager import CrawlerManager
 from nsct.crawler.normalize import NormalizedDocument
 from nsct.crawler.policy import (
     CrawlerPolicyError,
@@ -209,6 +210,15 @@ async def test_fetcher_unreachable_domain() -> None:
     await fetcher.close()
 
 
+def test_error_document_has_hash_for_empty_content() -> None:
+    """A failed fetch must still yield a valid normalized document."""
+    doc = CrawlerManager._error_doc("https://example.com/unavailable", "connection refused")
+
+    assert doc.text == ""
+    assert doc.metadata["error"] == "connection refused"
+    assert doc.content_hash == hashlib.sha256(b"").hexdigest()
+
+
 # ---------------------------------------------------------------------------
 # NormalizedDocument has all required fields
 # ---------------------------------------------------------------------------