retrieved_context: drop document completeness concept

surfsense_backend/app/agents/chat/multi_agent_chat/shared/retrieved_context/models.py

@@ -21,17 +21,11 @@ class RetrievedPassage:
 
 @dataclass(frozen=True)
 class RetrievedDocument:
-    """A source document and the passages retrieved from it, in order.
-
-    ``is_complete`` is ``True`` when every chunk of the document is present in
-    this block, so the model knows whether it is seeing the whole source or
-    only excerpts.
-    """
+    """A source document and the passages retrieved from it, in order."""
 
     document_id: int
     title: str
     source_label: str | None = None
-    is_complete: bool = False
     passages: list[RetrievedPassage] = field(default_factory=list)
 
 

surfsense_backend/app/agents/chat/multi_agent_chat/shared/retrieved_context/renderer.py

@@ -18,9 +18,8 @@ from .models import RetrievedDocument, RetrievedPassage
 _HEADER = (
     "These are excerpts from the user's knowledge base, selected for this query.\n"
     "A document is a full source (a file, a Slack thread, a Notion page); a chunk\n"
-    "is one ordered fragment of it. Each document is tagged (partial) when only\n"
-    "some of its chunks were retrieved or (complete) when all of them are shown\n"
-    "here, so you know whether you have the whole source or only parts of it.\n"
+    "is one ordered fragment of it. You are seeing only the chunks that matched\n"
+    "this query, not the whole source.\n"
     "Cite a chunk with [n]."
 )
 
@@ -61,10 +60,9 @@ def _render_document(
 
 
 def _render_header(document: RetrievedDocument) -> str:
-    """``Document: "Title"  (source)  (partial|complete)``."""
+    """``Document: "Title"  (source)``."""
     source = f"  ({document.source_label})" if document.source_label else ""
-    completeness = "(complete)" if document.is_complete else "(partial)"
-    return f'Document: "{_clean(document.title)}"{source}  {completeness}'
+    return f'Document: "{_clean(document.title)}"{source}'
 
 
 def _render_passage(

surfsense_backend/tests/unit/agents/multi_agent_chat/shared/retrieved_context/test_renderer.py

@@ -23,13 +23,11 @@ def _document(
     chunk_ids: list[int],
     *,
     source_label: str | None = None,
-    is_complete: bool = False,
 ) -> RetrievedDocument:
     return RetrievedDocument(
         document_id=document_id,
         title=title,
         source_label=source_label,
-        is_complete=is_complete,
         passages=[
             RetrievedPassage(document_id=document_id, chunk_id=cid, content=f"text {cid}")
             for cid in chunk_ids
@@ -73,20 +71,20 @@ def test_registers_passages_with_chunk_locators() -> None:
     assert entry.display["title"] == "Doc"
 
 
-def test_header_shows_source_and_completeness() -> None:
+def test_header_shows_source_when_present() -> None:
     registry = CitationRegistry()
 
     block = render_retrieved_context(
         [
-            _document(1, "Q3", [1], source_label="Slack · #launch", is_complete=False),
-            _document(2, "Plan", [2], is_complete=True),
+            _document(1, "Q3", [1], source_label="Slack · #launch"),
+            _document(2, "Plan", [2]),
         ],
         registry,
     )
 
     assert block is not None
-    assert 'Document: "Q3"  (Slack · #launch)  (partial)' in block
-    assert 'Document: "Plan"  (complete)' in block
+    assert 'Document: "Q3"  (Slack · #launch)' in block
+    assert 'Document: "Plan"' in block
 
 
 def test_wraps_block_and_explains_chunk_vs_document() -> None: