mirror of
https://github.com/langbot-app/LangBot.git
synced 2026-11-22 00:09:48 +08:00
fix(LongTermMemory): clear derived memory across document and KB deletion
This commit is contained in:
@@ -24,6 +24,7 @@ from langbot_plugin.api.entities.builtin.rag.enums import DocumentStatus
|
||||
from store.recall import matches_filters, select_memories
|
||||
from store.serialization import serialized_write
|
||||
from store.episode_provenance import normalize_provenance
|
||||
from store.consolidation import invalidate_derived
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
@@ -504,6 +505,33 @@ class LongTermMemoryEngine(KnowledgeEngine):
|
||||
|
||||
@serialized_write
|
||||
async def _delete_document_vectors(self, kb_id: str, document_id: str) -> int:
|
||||
# Resolve all sources before deleting summaries can change pagination.
|
||||
sources: dict[str, set[str]] = {}
|
||||
offset = 0
|
||||
while True:
|
||||
page = await self.plugin.vector_list(
|
||||
collection_id=kb_id,
|
||||
filters={"document_id": document_id},
|
||||
limit=100,
|
||||
offset=offset,
|
||||
)
|
||||
items = page.get("items", [])
|
||||
for item in items:
|
||||
metadata = item.get("metadata", {})
|
||||
episode_id = item.get("id")
|
||||
user_key = metadata.get("user_key")
|
||||
if (
|
||||
metadata.get("document_id") != document_id
|
||||
or not isinstance(episode_id, str) or not episode_id
|
||||
or not isinstance(user_key, str) or not user_key
|
||||
):
|
||||
raise ValueError("Document memory source identity is missing or outside the requested document")
|
||||
sources.setdefault(user_key, set()).add(episode_id)
|
||||
offset += len(items)
|
||||
if len(items) < 100 or (page.get("total", -1) >= 0 and offset >= page["total"]):
|
||||
break
|
||||
for user_key, episode_ids in sorted(sources.items()):
|
||||
await invalidate_derived(self.plugin.memory_store, kb_id, episode_ids, user_key)
|
||||
return await self._writes.mutate(
|
||||
self.plugin,
|
||||
kb_id,
|
||||
|
||||
Reference in New Issue
Block a user