Skip to content

Repository navigation

fix(openai): count base64 embedding vector dimensions #2077

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

Open
as992949791 wants to merge 2 commits into
base: main
Choose a base branch
from
Open

fix(openai): count base64 embedding vector dimensions #2077

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4d95016
fix(openai): count base64 embedding vector dimensions
as992949791 Oct 30, 2026
4f61fd0
fix(openai): preserve malformed embedding response passthrough
as992949791 Oct 30, 2026
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
12 changes: 11 additions & 1 deletion langfuse/openai.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,8 @@
See docs for more details: https://langfuse.com/docs/integrations/openai
"""

import base64
import binascii
import json
import os
import types
Expand Down Expand Up @@ -963,8 +965,16 @@ def _get_langfuse_data_from_default_response(
if hasattr(first_embedding, "embedding")
else first_embedding.get("embedding", [])
)
if isinstance(embedding_vector, str):
try:
# Base64 embeddings contain one float32 per dimension.
dimensions = len(base64.b64decode(embedding_vector)) // 4

@greptile-apps greptile-apps Bot Oct 30, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

P2 Malformed strings bypass the fallback

base64.b64decode silently discards invalid characters by default. A provider response of "!!!!" therefore exports 0 dimensions instead of the previous 4. Likewise, "AAAA" decodes to three bytes, which // 4 truncates to 0 despite not containing a complete float32.

Use strict decoding and check that the decoded length is divisible by four before counting dimensions. Add both cases to the fallback tests.

Prompt To Fix With AI 
This is a comment left during a code review.
Path: langfuse/openai.py
Line: 971

Comment:
**Malformed strings bypass the fallback**

`base64.b64decode` silently discards invalid characters by default. A provider response of `"!!!!"` therefore exports 0 dimensions instead of the previous 4. Likewise, `"AAAA"` decodes to three bytes, which `// 4` truncates to 0 despite not containing a complete float32.

Use strict decoding and check that the decoded length is divisible by four before counting dimensions. Add both cases to the fallback tests.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

except (binascii.Error, ValueError):
dimensions = len(embedding_vector)
else:
dimensions = len(embedding_vector) if embedding_vector else 0
completion = {
"dimensions": len(embedding_vector) if embedding_vector else 0,
"dimensions": dimensions,
"count": len(data),
}

Expand Down
48 changes: 48 additions & 0 deletions tests/unit/test_openai.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@
from types import SimpleNamespace
from unittest.mock import patch

import httpx
import pytest
from openai.types.chat import ChatCompletion
from openai.types.responses import ParsedResponseOutputMessage, ParsedResponseOutputText
Expand Down Expand Up @@ -1015,6 +1016,53 @@ def test_embedding_exports_dimensions_and_count(
}


@pytest.mark.parametrize(
"encoded_embedding, dimensions",
[
("AAAAAA==", 1),
("AAAAAAAAAAA=", 2),
("AAAAAAAAAAAAAAAA", 3),
pytest.param("!invalid!", 9, id="invalid-base64"),
pytest.param("\u2603", 1, id="non-ascii"),
],
)
def test_base64_embedding_exports_vector_dimensions(
langfuse_memory_client, get_span, json_attr, encoded_embedding, dimensions
):

def respond(request):
return httpx.Response(
200,
json={
"model": "text-embedding-3-small",
"object": "list",
"data": [
{"index": 0, "object": "embedding", "embedding": encoded_embedding}
],
"usage": {"prompt_tokens": 2, "total_tokens": 2},
},
)

with lf_openai.OpenAI(
api_key="test",
http_client=httpx.Client(transport=httpx.MockTransport(respond)),
) as openai_client:
result = openai_client.embeddings.create(
name="unit-openai-base64-embedding",
model="text-embedding-3-small",
input="hello world",
encoding_format="base64",
)

assert result.data[0].embedding == encoded_embedding
langfuse_memory_client.flush()
span = get_span("unit-openai-base64-embedding")
assert json_attr(span, LangfuseOtelSpanAttributes.OBSERVATION_OUTPUT) == {
"dimensions": dimensions,
"count": 1,
}


def _make_chat_response(**extra_response_fields):
return SimpleNamespace(
model="gpt-4o-mini",
Expand Down
Loading