Fimeg/letta-server
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

fix: re-add embedding dimension padding for archival memories (#6041)

Browse Source 
fix embedding dimension padding
This commit is contained in:
Kian Jones
2025-11-06 19:17:26 -08:00
committed by Caren Thomas
parent db6d4bb124
commit a217c8f1b6
1 changed files with 22 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
letta/schemas/passage.py
View File

@@ -56,20 +56,33 @@ class Passage(PassageBase):
created_at: datetime = Field(default_factory=get_utc_time, description="The creation date of the passage.")
@field_validator("embedding")
@field_validator("embedding", mode="before")
@classmethod
def pad_embeddings(cls, embedding: List[float]) -> List[float]:
def pad_embeddings(cls, embedding: List[float], info) -> List[float]:
"""Pad embeddings to `MAX_EMBEDDING_SIZE`. This is necessary to ensure all stored embeddings are the same size."""
# Only do this if using pgvector
from letta.helpers.tpuf_client import should_use_tpuf
if embedding is None:
return embedding
if not should_use_tpuf() or settings.environment != "PRODUCTION":
import numpy as np
# Check if this is an archival memory passage (has archive_id) or file passage (has file_id)
data = info.data if hasattr(info, "data") else {}
is_archival = data.get("archive_id") is not None
is_file = data.get("file_id") is not None
# Pad if using pgvector
if settings.letta_pg_uri_no_default:
# For archival memory: always pad
# For file passages: only pad if NOT using turbopuffer
from letta.helpers.tpuf_client import should_use_tpuf
should_pad = is_archival or (is_file and not should_use_tpuf())
if should_pad:
import numpy as np
if embedding and len(embedding) != MAX_EMBEDDING_DIM:
np_embedding = np.array(embedding)
padded_embedding = np.pad(np_embedding, (0, MAX_EMBEDDING_DIM - np_embedding.shape[0]), mode="constant")
return padded_embedding.tolist()
if np_embedding.shape[0] != MAX_EMBEDDING_DIM:
padded_embedding = np.pad(np_embedding, (0, MAX_EMBEDDING_DIM - np_embedding.shape[0]), mode="constant")
return padded_embedding.tolist()
return embedding