Integrating Phoenix Trace with Streamlit for Agent Monitoring

·Feb 21, 2025 06:55 AM

Hello guys, is there any way to integrate phoenix trace with streamlit, so I can see if the agents are working as expected.

11 comments

· Sorted by Oldest

RunLLM
APP
·
Yes—you can integrate Phoenix traces into your Streamlit app. For example, you can use Phoenix’s Python client to either: • Fetch trace data (via methods like `px.Client().get_spans_dataframe()`) and present it directly in Streamlit (e.g. using `st.dataframe`), or • Launch the Phoenix UI (using `px.launch_app()`) and embed its URL (for example, in an iframe via Streamlit’s components API) so you can interact with the full Phoenix interface. Here's a simple example: ``` import streamlit as st import phoenix as px import streamlit.components.v1 as components # Option 1: Display trace data as a table client = px.Client() trace_df = client.get_spans_dataframe() st.subheader("Phoenix Trace Data") st.dataframe(trace_df) # Option 2: Embed the Phoenix UI in an iframe session = px.launch_app() # This launches Phoenix and provides a session URL st.subheader("Phoenix UI") components.iframe(session.url, height=600, width=1000) ``` According to the <https://docs.arize.com/phoenix/~/changes/v6Zhm276x8LlKmwqElIA/how-to/export-your-data|Phoenix documentation>, you can export traces as a pandas DataFrame, and the <https://docs.arize.com/phoenix/deployment/deploying-phoenix#03132250edec410c927a4cdc4fff90fe|Phoenix Deployment Documentation> explains how to deploy and expose the Phoenix UI. These approaches let you monitor whether your agents are working as expected right from Streamlit. Would you like more details or additional examples on one of these methods? --- Sources: - <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1723828797.781399|https://arize-ai.slack.com/archives/C04R3GXC8HK/p1723828797.781399> - <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1726632533.889129|https://arize-ai.slack.com/archives/C04R3GXC8HK/p1726632533.889129>
💡 Hint: Mention RunLLM in the thread for followups.
Arjit P.
·
RunLLM How can I connect to an ip address in the iframe option? Do I have to put the IP address in place of session url
RunLLM
APP
·
You don’t have to manually replace the session URL with an IP address—if you want the iframe to point to a specific IP, you simply need to launch Phoenix with that IP as the host. For example, by setting the host parameter in px.launch_app (or by setting the PHOENIX_HOST environment variable), the session.url property will automatically be generated with that IP: ``` import phoenix as px # Launch Phoenix with a specific IP address (e.g., 192.168.1.100) session = px.launch_app(host="192.168.1.100", port=6006) print(session.url) # This should show "http://192.168.1.100:6006/" ``` Then, when embedding the URL in your iframe (for instance, using Streamlit’s components.iframe), it will connect to the desired IP address without any manual substitution. Would you like more examples of configuring the host parameter or further details on embedding in an iframe? --- Sources: - <https://arize-phoenix.readthedocs.io/en/latest/api/session.html|Phoenix API Reference on launch_app>
💡 Hint: Mention RunLLM in the thread for followups.
Arjit P.
·
RunLLM The phoenix is already running on a server. How can I connect it to my streamlit and show the traces?
RunLLM
APP
·
Since your Phoenix server is already running on a remote machine, you don’t need to launch a new instance in your Streamlit code. Instead, you can connect to it using the client API by specifying its endpoint (IP address and port). For example: ``` import streamlit as st import phoenix as px import streamlit.components.v1 as components # Connect to the remote Phoenix instance by its endpoint endpoint = "http://<your-server-ip> :6006" client = px.Client(endpoint=endpoint) # Retrieve trace data from the default project (add project_name if needed) trace_df = client.get_spans_dataframe() st.subheader("Phoenix Trace Data") if trace_df is not None: st.dataframe(trace_df) else: st.write("No trace data available. Confirm that the correct project is being queried.") # Optionally, embed the full Phoenix UI in an iframe st.subheader("Phoenix UI") components.iframe(endpoint, height=600, width=1000) ``` In this example: - Replace `<your-server-ip> ` with your server’s IP address. - By using `px.Client(endpoint=...)`, your Streamlit app connects to the remote Phoenix instance without starting a new session. - You can display trace data as a table (using `st.dataframe`) or embed the full Phoenix UI (using `components.iframe`). Make sure your network/firewall settings allow your Streamlit app to access that IP and port. Would you like more details on customizing the client connection (such as specifying a project) or additional examples of integrating the Phoenix UI into Streamlit? --- *Sources:* - <https://docs.arize.com/phoenix/deployment/deploying-phoenix#03132250edec410c927a4cdc4fff90fe|Phoenix Deployment Documentation> - <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1715682199.787409|Phoenix Slack Discussion> Sources: - <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1715682199.787409|Phoenix Slack Discussion>
💡 Hint: Mention RunLLM in the thread for followups.
Arjit P.
·
RunLLM I am getting an error "ModuleNotFoundError: No module named 'phoenix'" while running the streamlit application. I can see phoenix installed in my virtual environment
RunLLM
APP
·
This error usually means your Streamlit app isn’t running with the Python interpreter that has Phoenix installed. Here are a few steps to troubleshoot: 1. *Activate Your Virtual Environment:* Make sure you’ve activated the same virtual environment where you installed the Phoenix package (i.e. arize‑phoenix). For example, in your terminal run: ``` source path/to/venv/bin/activate ``` Then run your Streamlit app (e.g. `streamlit run app.py`). 2. *Check the Correct Package is Installed:* Ensure you installed `arize-phoenix` and not a different package named “phoenix.” Although you import it as “phoenix” in your code, the pip package is named arize‑phoenix. You can confirm this by running: ``` pip show arize-phoenix ``` 3. *Verify Interpreter Consistency:* Sometimes if you have multiple Python versions or if your IDE/streamlit command isn’t using the activated environment, you might get a ModuleNotFoundError. Verify your Streamlit command uses your virtual environment’s Python (for example, check with `which python` or `which streamlit`). Would you like more guidance on verifying your environment settings or adjusting your IDE configuration? --- *Sources:* - According to a <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1715102389.037279|Slack discussion>, users should ensure they install `arize-phoenix` (not a package just named `phoenix`) and run their applications within the correct virtual environment. Sources: - <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1715102389.037279|Slack discussion>
💡 Hint: Mention RunLLM in the thread for followups.
Arjit P.
·
RunLLM Checked all of the points, still the issue is there. What could be the other issues
RunLLM
APP
·
I'm not able to find any data sources to answer your question. You might want to check if there are any conflicting packages or if the PYTHONPATH is set correctly. If this doesn't help, please reach out to the Phoenix support team for further assistance.
💡 Hint: Mention RunLLM in the thread for followups.
Arjit P.
·
RunLLM Phoenix serve is giving this error: "SyntaxError: multiple exception types must be parenthesized". What could be the issue?
RunLLM
APP
·
I'm not able to find any data sources to answer your question. You might want to check the Python version you are using, as this error can occur if the code is not compatible with your Python version. If this doesn't help, please reach out to the Phoenix support team for further assistance.
💡 Hint: Mention RunLLM in the thread for followups.

Integrating Phoenix Trace with Streamlit for Agent Monitoring | Arize AI Community

RunLLM
APP
·
Yes—you can integrate Phoenix traces into your Streamlit app. For example, you can use Phoenix’s Python client to either: • Fetch trace data (via methods like `px.Client().get_spans_dataframe()`) and present it directly in Streamlit (e.g. using `st.dataframe`), or • Launch the Phoenix UI (using `px.launch_app()`) and embed its URL (for example, in an iframe via Streamlit’s components API) so you can interact with the full Phoenix interface. Here's a simple example: ``` import streamlit as st import phoenix as px import streamlit.components.v1 as components # Option 1: Display trace data as a table client = px.Client() trace_df = client.get_spans_dataframe() st.subheader("Phoenix Trace Data") st.dataframe(trace_df) # Option 2: Embed the Phoenix UI in an iframe session = px.launch_app() # This launches Phoenix and provides a session URL st.subheader("Phoenix UI") components.iframe(session.url, height=600, width=1000) ``` According to the <https://docs.arize.com/phoenix/~/changes/v6Zhm276x8LlKmwqElIA/how-to/export-your-data|Phoenix documentation>, you can export traces as a pandas DataFrame, and the <https://docs.arize.com/phoenix/deployment/deploying-phoenix#03132250edec410c927a4cdc4fff90fe|Phoenix Deployment Documentation> explains how to deploy and expose the Phoenix UI. These approaches let you monitor whether your agents are working as expected right from Streamlit. Would you like more details or additional examples on one of these methods? --- Sources: - <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1723828797.781399|https://arize-ai.slack.com/archives/C04R3GXC8HK/p1723828797.781399> - <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1726632533.889129|https://arize-ai.slack.com/archives/C04R3GXC8HK/p1726632533.889129>
💡 Hint: Mention RunLLM in the thread for followups.
Arjit P.
·
RunLLM How can I connect to an ip address in the iframe option? Do I have to put the IP address in place of session url
RunLLM
APP
·
You don’t have to manually replace the session URL with an IP address—if you want the iframe to point to a specific IP, you simply need to launch Phoenix with that IP as the host. For example, by setting the host parameter in px.launch_app (or by setting the PHOENIX_HOST environment variable), the session.url property will automatically be generated with that IP: ``` import phoenix as px # Launch Phoenix with a specific IP address (e.g., 192.168.1.100) session = px.launch_app(host="192.168.1.100", port=6006) print(session.url) # This should show "http://192.168.1.100:6006/" ``` Then, when embedding the URL in your iframe (for instance, using Streamlit’s components.iframe), it will connect to the desired IP address without any manual substitution. Would you like more examples of configuring the host parameter or further details on embedding in an iframe? --- Sources: - <https://arize-phoenix.readthedocs.io/en/latest/api/session.html|Phoenix API Reference on launch_app>
💡 Hint: Mention RunLLM in the thread for followups.
Arjit P.
·
RunLLM The phoenix is already running on a server. How can I connect it to my streamlit and show the traces?
RunLLM
APP
·
Since your Phoenix server is already running on a remote machine, you don’t need to launch a new instance in your Streamlit code. Instead, you can connect to it using the client API by specifying its endpoint (IP address and port). For example: ``` import streamlit as st import phoenix as px import streamlit.components.v1 as components # Connect to the remote Phoenix instance by its endpoint endpoint = "http://<your-server-ip> :6006" client = px.Client(endpoint=endpoint) # Retrieve trace data from the default project (add project_name if needed) trace_df = client.get_spans_dataframe() st.subheader("Phoenix Trace Data") if trace_df is not None: st.dataframe(trace_df) else: st.write("No trace data available. Confirm that the correct project is being queried.") # Optionally, embed the full Phoenix UI in an iframe st.subheader("Phoenix UI") components.iframe(endpoint, height=600, width=1000) ``` In this example: - Replace `<your-server-ip> ` with your server’s IP address. - By using `px.Client(endpoint=...)`, your Streamlit app connects to the remote Phoenix instance without starting a new session. - You can display trace data as a table (using `st.dataframe`) or embed the full Phoenix UI (using `components.iframe`). Make sure your network/firewall settings allow your Streamlit app to access that IP and port. Would you like more details on customizing the client connection (such as specifying a project) or additional examples of integrating the Phoenix UI into Streamlit? --- *Sources:* - <https://docs.arize.com/phoenix/deployment/deploying-phoenix#03132250edec410c927a4cdc4fff90fe|Phoenix Deployment Documentation> - <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1715682199.787409|Phoenix Slack Discussion> Sources: - <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1715682199.787409|Phoenix Slack Discussion>
💡 Hint: Mention RunLLM in the thread for followups.
Arjit P.
·
RunLLM I am getting an error "ModuleNotFoundError: No module named 'phoenix'" while running the streamlit application. I can see phoenix installed in my virtual environment
RunLLM
APP
·
This error usually means your Streamlit app isn’t running with the Python interpreter that has Phoenix installed. Here are a few steps to troubleshoot: 1. *Activate Your Virtual Environment:* Make sure you’ve activated the same virtual environment where you installed the Phoenix package (i.e. arize‑phoenix). For example, in your terminal run: ``` source path/to/venv/bin/activate ``` Then run your Streamlit app (e.g. `streamlit run app.py`). 2. *Check the Correct Package is Installed:* Ensure you installed `arize-phoenix` and not a different package named “phoenix.” Although you import it as “phoenix” in your code, the pip package is named arize‑phoenix. You can confirm this by running: ``` pip show arize-phoenix ``` 3. *Verify Interpreter Consistency:* Sometimes if you have multiple Python versions or if your IDE/streamlit command isn’t using the activated environment, you might get a ModuleNotFoundError. Verify your Streamlit command uses your virtual environment’s Python (for example, check with `which python` or `which streamlit`). Would you like more guidance on verifying your environment settings or adjusting your IDE configuration? --- *Sources:* - According to a <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1715102389.037279|Slack discussion>, users should ensure they install `arize-phoenix` (not a package just named `phoenix`) and run their applications within the correct virtual environment. Sources: - <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1715102389.037279|Slack discussion>
💡 Hint: Mention RunLLM in the thread for followups.
Arjit P.
·
RunLLM Checked all of the points, still the issue is there. What could be the other issues
RunLLM
APP
·
I'm not able to find any data sources to answer your question. You might want to check if there are any conflicting packages or if the PYTHONPATH is set correctly. If this doesn't help, please reach out to the Phoenix support team for further assistance.
💡 Hint: Mention RunLLM in the thread for followups.
Arjit P.
·
RunLLM Phoenix serve is giving this error: "SyntaxError: multiple exception types must be parenthesized". What could be the issue?
RunLLM
APP
·
I'm not able to find any data sources to answer your question. You might want to check the Python version you are using, as this error can occur if the code is not compatible with your Python version. If this doesn't help, please reach out to the Phoenix support team for further assistance.
💡 Hint: Mention RunLLM in the thread for followups.