fix(import): page import previews like execution

Preview plans read the vault manifest through the same keyset-paged reader as execution, so manifests larger than one list page no longer make previews silently drop beyond-boundary rows; they are reported as missing like any other unseen source. Regression test seeds a real 10k-row manifest and fails against the unpaged reader (stash-verified).

Author: Coding-Dev-Tools

CHANGELOG.md

@@ -108,6 +108,9 @@ All notable changes to Engraphis are documented here. Format loosely follows
 
 ### Fixed
 
+- Import previews now page the source manifest exactly like execution, so vaults whose manifest
+  outgrew one list page (10k identities) no longer show manifest-only files as silently absent
+  from the preview plan; beyond-boundary rows are reported as `missing` instead of dropped.
 - Importing more than 1,000 files through the dashboard no longer fails with "Internal Server
   Error": wizard upload routes parse multipart forms under the advertised 1,500-file ceiling
   instead of Starlette's hidden 1,000-part parser default, oversized batches return a clear 413,

engraphis/obsidian_import.py

@@ -511,7 +511,9 @@ def _preview_manifest(
                 scope=scope, memory_type=memory_type, strict_root=strict_root,
             )
             if manifest is None:
-                items = self.store.list_source_import_items(vault_id=str(vault["id"]))
+                # Page the manifest like import_scan does so previews on manifests
+                # larger than one list page plan against the full item set.
+                items, _ = self._all_source_items(vault_id=str(vault["id"]))
             else:
                 items = [row for row in items if row.get("vault_id") == vault.get("id")]
         else:

tests/test_obsidian_service.py

@@ -1,6 +1,7 @@
 """Real-service coverage for the owner-only Obsidian import facade."""
 from __future__ import annotations
 
+import hashlib
 import time
 
 import pytest
@@ -43,6 +44,52 @@ def _import(service: MemoryService, files: list[tuple[str, bytes]], **kwargs) ->
     return started, _await_job(service, started)
 
 
+def test_preview_pages_the_full_manifest_like_execution():
+    service = _service()
+    try:
+        started, imported = _import(
+            service,
+            [
+                ("A.md", b"# A\n"),
+                ("B.md", b"# B\n"),
+                ("C.md", b"# C\n"),
+            ],
+        )
+        vault_id = started["vault_id"]
+        assert imported["state"] == "completed"
+
+        # Push the vault manifest past the default 10k list-page boundary with
+        # filler identity rows that sort before the scan set, so an unpaged
+        # preview loses exactly the rows a real oversized vault would lose.
+        for i in range(10_001):
+            service.store.upsert_source_import_item(
+                vault_id=vault_id,
+                source_key=hashlib.sha256(f"seed-{i}".encode()).hexdigest(),
+                relative_path=f"0000-seed-{i:05d}.md",
+            )
+
+        preview = service.preview_obsidian_upload(
+            files=[("A.md", b"# A\n"), ("D.md", b"# D\n")],
+            attachment_manifest=[],
+            workspace="alpha",
+            vault_label="Team notes",
+            vault_id=vault_id,
+        )
+
+        # An unpaged preview reads only the first list page, so manifest rows
+        # beyond the boundary vanish from the report entirely. The paged reader
+        # must surface every manifest row: A plans as unchanged and B/C — not
+        # part of this preview's scan — are reported as missing, not dropped.
+        statuses = {
+            row["relative_path"]: row["status"] for row in preview["files"]
+        }
+        assert statuses["A.md"] != "missing"
+        assert "B.md" in statuses
+        assert "C.md" in statuses
+    finally:
+        service.close()
+
+
 def test_preview_is_write_free_and_service_enforces_confirmation_and_upload_guards(monkeypatch):
     service = _service()
     try: