Upload 2 files

pages/trace.py

@@ -0,0 +1,24 @@
+
+import streamlit as st
+
+#from .streamapp import trace_df
+
+print("trace_df  ", st.session_state['trace_df'])
+
+trace_df = st.session_state['trace_df']
+print(list(trace_df))
+
+trace_df = trace_df.loc[:,['name', 'span_kind', 'start_time', 'end_time', 'attributes.__computed__.latency_ms', 'status_code', 'status_message',  'attributes.llm.invocation_parameters', 'attributes.llm.prompts', 'attributes.input.value',  'attributes.output.value',  'attributes.llm.prompt_template.template', 'attributes.llm.prompt_template.variables', 'attributes.llm.prompt_template.version', 'attributes.retrieval.documents']]
+trace_df = trace_df.sort_values(by='start_time', ascending = False)
+
+
+st.dataframe(trace_df)
+
+# if px.active_session():  
+#     df0 = px.active_session().get_spans_dataframe()
+#     if not df0.empty:
+#         df= df0.fillna('')
+#         st.dataframe(df)
+
+
+#'name', 'span_kind', 'start_time', 'end_time', 'status_code', 'status_message',  'attributes.llm.invocation_parameters', 'attributes.llm.prompts', 'attributes.input.value',  'attributes.output.value', 'attributes.__computed__.latency_ms', 'attributes.llm.prompt_template.template', 'attributes.llm.prompt_template.variables', 'attributes.llm.prompt_template.version', 'attributes.retrieval.documents'

streamapp.py

@@ -32,17 +32,21 @@ import pandas as pd
 # from sklearn import datasets
 # from sklearn.ensemble import RandomForestClassifier
 
-#phoenix setup
-session = px.launch_app()
-# If no exporter is specified, the tracer will export to the locally running Phoenix server
-tracer = OpenInferenceTracer()
-# If no tracer is specified, a tracer is constructed for you
-LangChainInstrumentor(tracer).instrument()
-print(session.url)
+global trace_df
 
+@st.cache_resource
+def tracer_config():
+    #phoenix setup
+    session = px.launch_app()
+    # If no exporter is specified, the tracer will export to the locally running Phoenix server
+    tracer = OpenInferenceTracer()
+    # If no tracer is specified, a tracer is constructed for you
+    LangChainInstrumentor(tracer).instrument()
+    print(session.url)
 
+tracer_config()
 
-tab1, tab2, tab3 = st.tabs(["📈 RAG", "🗃 FactVsHallucinate" , "trace"])
+tab1, tab2 = st.tabs(["📈 RAG", "🗃 FactVsHallucinate" ])
 
 
 
@@ -263,19 +267,23 @@ with tab2:
     if st.form_submit_button("Evaluate"):
         hallu_eval(question, answer, context)
       
-  
 
-with tab3:
+print("activ session:  ", px.active_session().get_spans_dataframe())  
+trace_df = px.active_session().get_spans_dataframe()
+
+st.session_state['trace_df'] = trace_df
+
+# with tab3:
   
   
   
-  with st.form(" trace"):
+#   with st.form(" trace"):
 
-    if px.active_session():  
-        df0 = px.active_session().get_spans_dataframe()
-        if not df0.empty:
-            df= df0.fillna('')
-            st.dataframe(df)
+#     if px.active_session():  
+#         df0 = px.active_session().get_spans_dataframe()
+#         if not df0.empty:
+#             df= df0.fillna('')
+#             st.dataframe(df)