Troubleshooting Python Client Annotations Logging with Phoenix Server

GGhiasi · 2025-09-18T11:34:05.124Z

I have Phoenix server 11.32.1 running in Docker but my Python client is 10.13.2. I'm getting 'Client' object has no attribute 'annotations' when trying to log evaluation results. from examples I should be able to use" await px_client.annotations.log_span_annotations_dataframe(...)" What's the correct way to log evaluation annotations

10 comments

· Sorted by Oldest

RunLLM
APP
·
·
The error occurs because in Phoenix client versions ≥1.17.0, the annotations API is deprecated and replaced by resource-based APIs. You should use px_client.spans.log_span_annotations_dataframe(...) instead of px_client.annotations.log_span_annotations_dataframe(...). The correct method is: LBKSfrom phoenix.client import Client px_client = Client() px_client.spans.log_span_annotations_dataframe( dataframe=your_eval_df, annotation_name="Your Eval Name", annotator_kind="LLM", )LBKSSee client migration guide and official docs for details.Would you like more detail or async usage examples?Sources: - client migration guide - official docs: LLM evaluations
💡 Hint: Mention RunLLM in the thread for followups.
GGhiasi
·
·
The method px_client.spans.log_span_annotations_dataframe() doesn't exist in Phoenix 11.34.0.
GGhiasi
·
·
I'm on Phoenix client 11.34.0 and server 11.32.1. You told me to use px_client.spans.log_span_annotations_dataframe() but this method doesn't exist. The actual available methods on px_client.spans are:
get_span_annotations
get_span_annotations_dataframe
get_spans
get_spans_dataframe
log_spans
log_spans_dataframe
There is no method for logging annotations that I an see here. How do I actually log evaluation results/annotations to Phoenix in version 11.34.0
Roger Y.
·
·
Can you try updating the client to the latest version? I can see the new method in colab as shown in screenshot below.
pip install -Uqqq arize-phoenix-client
➕1
Mikyo
·
·
Hey GGhiasi anything under phoenix.client is actually managed by a sub-package called arize-phoenix-client and you should be able to bump the version independently. It could be a case where that dependency is just a bit lagging. Sorry for the confusion.
GGhiasi
·
·
Getting in to another problem, trying to log llmevaluation results using client.log_evaluations() but getting "422 Unprocessable Entity". My setup:
Phoenix server: 11.32.1 (Docker)
Phoenix client: 11.34.0
Using: px.Client(endpoint="http://localhost:6006").log_evaluations(df)
DataFrame format I'm trying: python
pd.DataFrame({ 'span_id': ['946bd4893035eb92'], 'name': ['Tool Usage Evaluation'], 'label': ['correct'], 'score': [1.0], 'explanation': ['Test evaluation'] })
What's the correct DataFrame schema for log_evaluations() in Phoenix 11.32.1?
GGhiasi
·
·
The tutorials dont seem to be updated with the versioning and how things should wrok
Mikyo
·
·
Hey GGhiasi - sorry about the version issues in our tutorials there. We'll switch over those tutorials ASAP. cc Sanjana Y.
🙌1
Mikyo
·
·
In general GGhiasi everything is moving to this method https://arize-phoenix.readthedocs.io/projects/client/en/latest/api/spans.html#client.resources.spans.Spans.log_span_annotations
🙌1
Mikyo
·
·
in the new client, there is no log_evaluations as it's just an annotation that's added via an LLM judge.

RunLLM
APP
·
·
The error occurs because in Phoenix client versions ≥1.17.0, the annotations API is deprecated and replaced by resource-based APIs. You should use px_client.spans.log_span_annotations_dataframe(...) instead of px_client.annotations.log_span_annotations_dataframe(...). The correct method is: LBKSfrom phoenix.client import Client px_client = Client() px_client.spans.log_span_annotations_dataframe( dataframe=your_eval_df, annotation_name="Your Eval Name", annotator_kind="LLM", )LBKSSee client migration guide and official docs for details.Would you like more detail or async usage examples?Sources: - client migration guide - official docs: LLM evaluations
💡 Hint: Mention RunLLM in the thread for followups.
GGhiasi
·
·
The method px_client.spans.log_span_annotations_dataframe() doesn't exist in Phoenix 11.34.0.
GGhiasi
·
·
I'm on Phoenix client 11.34.0 and server 11.32.1. You told me to use px_client.spans.log_span_annotations_dataframe() but this method doesn't exist. The actual available methods on px_client.spans are:
get_span_annotations
get_span_annotations_dataframe
get_spans
get_spans_dataframe
log_spans
log_spans_dataframe
There is no method for logging annotations that I an see here. How do I actually log evaluation results/annotations to Phoenix in version 11.34.0
Roger Y.
·
·
Can you try updating the client to the latest version? I can see the new method in colab as shown in screenshot below.
pip install -Uqqq arize-phoenix-client
➕1
Mikyo
·
·
Hey GGhiasi anything under phoenix.client is actually managed by a sub-package called arize-phoenix-client and you should be able to bump the version independently. It could be a case where that dependency is just a bit lagging. Sorry for the confusion.
GGhiasi
·
·
Getting in to another problem, trying to log llmevaluation results using client.log_evaluations() but getting "422 Unprocessable Entity". My setup:
Phoenix server: 11.32.1 (Docker)
Phoenix client: 11.34.0
Using: px.Client(endpoint="http://localhost:6006").log_evaluations(df)
DataFrame format I'm trying: python
pd.DataFrame({ 'span_id': ['946bd4893035eb92'], 'name': ['Tool Usage Evaluation'], 'label': ['correct'], 'score': [1.0], 'explanation': ['Test evaluation'] })
What's the correct DataFrame schema for log_evaluations() in Phoenix 11.32.1?
GGhiasi
·
·
The tutorials dont seem to be updated with the versioning and how things should wrok
Mikyo
·
·
Hey GGhiasi - sorry about the version issues in our tutorials there. We'll switch over those tutorials ASAP. cc Sanjana Y.
🙌1
Mikyo
·
·
In general GGhiasi everything is moving to this method https://arize-phoenix.readthedocs.io/projects/client/en/latest/api/spans.html#client.resources.spans.Spans.log_span_annotations
🙌1
Mikyo
·
·
in the new client, there is no log_evaluations as it's just an annotation that's added via an LLM judge.