Skip to content

fix: improve embedding diagnostics and handle invalid db #423

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
sng-asyncfunc merged 1 commit into from
Dec 27, 2025
Merged

fix: improve embedding diagnostics and handle invalid db #423

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
35 changes: 33 additions & 2 deletions api/data_pipeline.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -847,6 +847,21 @@ def prepare_db_index(self, embedder_type: str = None, is_ollama_embedder: bool =
Returns:
List[Document]: List of Document objects
"""
def _embedding_vector_length(doc: Document) -> int:
vector = getattr(doc, "vector", None)
if vector is None:
return 0
try:
if hasattr(vector, "shape"):
if len(vector.shape) == 0:
return 0
return int(vector.shape[-1])
if hasattr(vector, "__len__"):
return int(len(vector))
except Exception:
return 0
return 0

# Handle backward compatibility
if embedder_type is None and is_ollama_embedder is not None:
embedder_type = 'ollama' if is_ollama_embedder else None
Expand All @@ -857,8 +872,24 @@ def prepare_db_index(self, embedder_type: str = None, is_ollama_embedder: bool =
self.db = LocalDB.load_state(self.repo_paths["save_db_file"])
documents = self.db.get_transformed_data(key="split_and_embed")
if documents:
logger.info(f"Loaded {len(documents)} documents from existing database")
return documents
lengths = [_embedding_vector_length(doc) for doc in documents]
non_empty = sum(1 for n in lengths if n > 0)
empty = len(lengths) - non_empty
sample_sizes = sorted({n for n in lengths if n > 0})[:3]
logger.info(
"Loaded %s documents from existing database (embeddings: %s non-empty, %s empty; sample_dims=%s)",
len(documents),
non_empty,
empty,
sample_sizes,
)

if non_empty == 0:
logger.warning(
"Existing database contains no usable embeddings. Rebuilding embeddings..."
)
else:
return documents
except Exception as e:
logger.error(f"Error loading existing database: {e}")
# Continue to create a new database
Expand Down
91 changes: 60 additions & 31 deletions api/google_embedder_client.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -88,42 +88,62 @@ def parse_embedding_response(self, response) -> EmbedderOutput:
from adalflow.core.types import Embedding

embedding_data = []

def _extract_embedding_value(obj):
if obj is None:
return None
if isinstance(obj, dict):
if "embedding" in obj:
return obj.get("embedding")
if "embeddings" in obj:
return obj.get("embeddings")
if hasattr(obj, "embedding"):
return getattr(obj, "embedding")
if hasattr(obj, "embeddings"):
return getattr(obj, "embeddings")
for method_name in ("model_dump", "to_dict", "dict"):
if hasattr(obj, method_name):
try:
dumped = getattr(obj, method_name)()
if isinstance(dumped, dict):
if "embedding" in dumped:
return dumped.get("embedding")
if "embeddings" in dumped:
return dumped.get("embeddings")
except Exception:
pass
Comment on lines +113 to +114
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot Dec 23, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

The broad except Exception can hide underlying issues that might be useful for debugging why embedding parsing is failing. While the goal is robustness, completely silencing exceptions can make it difficult to diagnose problems with new or unexpected response object structures. Consider adding a debug level log to record the exception. This will keep the logs clean in normal operation but provide valuable information when debugging.

Suggested change
except Exception:
pass
except Exception as e:
log.debug("Failed to get embedding via '%s': %s", method_name, e)
pass

return None

if isinstance(response, dict):
if 'embedding' in response:
embedding_value = response['embedding']
if isinstance(embedding_value, list) and len(embedding_value) > 0:
# Check if it's a single embedding (list of floats) or batch (list of lists)
if isinstance(embedding_value[0], (int, float)):
# Single embedding response: {'embedding': [float, ...]}
embedding_data = [Embedding(embedding=embedding_value, index=0)]
else:
# Batch embeddings response: {'embedding': [[float, ...], [float, ...], ...]}
embedding_data = [
Embedding(embedding=emb_list, index=i)
for i, emb_list in enumerate(embedding_value)
]
else:
log.warning(f"Empty or invalid embedding data: {embedding_value}")
embedding_data = []
elif 'embeddings' in response:
# Alternative batch format: {'embeddings': [{'embedding': [float, ...]}, ...]}
embedding_value = _extract_embedding_value(response)
if embedding_value is None:
log.warning("Unexpected embedding response type/structure: %s", type(response))
embedding_data = []
elif isinstance(embedding_value, list) and len(embedding_value) > 0:
if isinstance(embedding_value[0], (int, float)):
embedding_data = [Embedding(embedding=embedding_value, index=0)]
elif isinstance(embedding_value[0], list):
embedding_data = [
Embedding(embedding=item['embedding'], index=i)
for i, item in enumerate(response['embeddings'])
Embedding(embedding=emb_list, index=i)
for i, emb_list in enumerate(embedding_value)
if isinstance(emb_list, list) and len(emb_list) > 0
]
else:
log.warning(f"Unexpected response structure: {response.keys()}")
embedding_data = []
elif hasattr(response, 'embeddings'):
# Custom batch response object from our implementation
embedding_data = [
Embedding(embedding=emb, index=i)
for i, emb in enumerate(response.embeddings)
]
extracted = []
for item in embedding_value:
item_emb = _extract_embedding_value(item)
if isinstance(item_emb, list) and len(item_emb) > 0:
extracted.append(item_emb)
embedding_data = [
Embedding(embedding=emb_list, index=i)
for i, emb_list in enumerate(extracted)
]
else:
log.warning(f"Unexpected response type: {type(response)}")
log.warning("Empty or invalid embedding data parsed from response")
embedding_data = []

if embedding_data:
first_dim = len(embedding_data[0].embedding) if embedding_data[0].embedding is not None else 0
log.info("Parsed %s embedding(s) (dim=%s)", len(embedding_data), first_dim)

return EmbedderOutput(
data=embedding_data,
Expand Down Expand Up @@ -201,7 +221,16 @@ def call(self, api_kwargs: Dict = {}, model_type: ModelType = ModelType.UNDEFINE
if model_type != ModelType.EMBEDDER:
raise ValueError(f"GoogleEmbedderClient only supports EMBEDDER model type")

log.info(f"Google AI Embeddings API kwargs: {api_kwargs}")
safe_log_kwargs = {k: v for k, v in api_kwargs.items() if k not in {"content", "contents"}}
if "content" in api_kwargs:
safe_log_kwargs["content_chars"] = len(str(api_kwargs.get("content", "")))
if "contents" in api_kwargs:
try:
contents = api_kwargs.get("contents")
safe_log_kwargs["contents_count"] = len(contents) if hasattr(contents, "__len__") else None
except Exception:
safe_log_kwargs["contents_count"] = None
log.info("Google AI Embeddings call kwargs (sanitized): %s", safe_log_kwargs)

try:
# Use embed_content for single text or batch embedding
Expand Down