Update LLM evaluation dashboard

app.py

@@ -2,6 +2,11 @@ import streamlit as st
 import requests
 import pandas as pd
 
+
+# =========================================================
+# Page Configuration
+# =========================================================
+
 st.set_page_config(
     page_title="LLM Hallucination Detector",
     layout="wide"
@@ -11,80 +16,126 @@ st.title("🔍 LLM Evaluation & Hallucination Detection Framework")
 
 API_URL = "https://sha6th-llm-eval-ap.hf.space"
 
-# ---------------------------------------------------------
+
+# =========================================================
 # Tabs
-# ---------------------------------------------------------
+# =========================================================
 
-tab1, tab2 = st.tabs(["Evaluate New Response", "History"])
+tab1, tab2 = st.tabs(
+    ["Evaluate New Response", "History"]
+)
 
 
 # =========================================================
-# TAB 1: Evaluate
+# TAB 1: LLM Evaluation
 # =========================================================
 
 with tab1:
 
     st.subheader("Evaluate an LLM Response")
 
-    st.info("Enter each context chunk on a new line")
+    st.info(
+        "Enter each context chunk on a new line"
+    )
+
+    # -----------------------------------------------------
+    # Input: Context
+    # -----------------------------------------------------
 
     contexts_input = st.text_area(
         "Retrieved Contexts (one chunk per line)",
         height=150
     )
 
-    question = st.text_input("Question")
+    # -----------------------------------------------------
+    # Input: Question
+    # -----------------------------------------------------
+
+    question = st.text_input(
+        "Question"
+    )
+
+    # -----------------------------------------------------
+    # Input: LLM Response
+    # -----------------------------------------------------
 
     llm_response = st.text_area(
         "LLM Response",
         height=100
     )
 
+    # -----------------------------------------------------
+    # Evaluate Button
+    # -----------------------------------------------------
+
     if st.button("Evaluate"):
 
+        # Convert the context input into a list
         retrieved_contexts = [
             c.strip()
             for c in contexts_input.split("\n")
             if c.strip()
         ]
 
-        # Validate input
+        # -------------------------------------------------
+        # Validate Input
+        # -------------------------------------------------
+
         if (
             not retrieved_contexts
             or not question.strip()
             or not llm_response.strip()
         ):
-            st.error("All fields are required.")
+
+            st.error(
+                "All fields are required."
+            )
 
         else:
 
-            with st.spinner("Running evaluation..."):
+            # -------------------------------------------------
+            # Convert context chunks into ONE string
+            # -------------------------------------------------
+
+            context = "\n".join(
+                retrieved_contexts
+            )
+
+            with st.spinner(
+                "Running LLM evaluation..."
+            ):
 
                 try:
 
+                    # -------------------------------------------------
+                    # Call LLM-only evaluation endpoint
+                    # -------------------------------------------------
+
                     response = requests.post(
                         f"{API_URL}/evaluate-llm",
                         json={
-                            "retrieved_contexts": retrieved_contexts,
+                            "context": context,
                             "question": question,
                             "llm_response": llm_response
                         },
                         timeout=120
                     )
 
-                    # -------------------------------------------------
-                    # Successful response
-                    # -------------------------------------------------
+                    # =================================================
+                    # Successful Response
+                    # =================================================
 
                     if response.status_code == 200:
 
                         result = response.json()
 
-                        # -----------------------------
+                        # -------------------------------------------------
                         # Final Verdict
-                        # -----------------------------
+                        # -------------------------------------------------
 
-                        verdict = result["final_verdict"]
+                        verdict = result[
+                            "final_verdict"
+                        ]
 
                         if verdict == "Hallucinated":
 
@@ -104,59 +155,23 @@ with tab1:
                                 f"**Verdict: {verdict}**"
                             )
 
-                        # -----------------------------
-                        # Retrieval Evaluation
-                        # -----------------------------
-
-                        st.subheader("Retrieval Evaluation")
-
-                        col1, col2 = st.columns(2)
+                        # =================================================
+                        # Generation Evaluation
+                        # =================================================
 
-                        col1.metric(
-                            "Context Relevance",
-                            result[
-                                "retrieval_evaluation"
-                            ][
-                                "context_relevance"
-                            ][
-                                "average_score"
-                            ],
-                            result[
-                                "retrieval_evaluation"
-                            ][
-                                "context_relevance"
-                            ][
-                                "verdict"
-                            ]
+                        st.subheader(
+                            "Generation Evaluation"
                         )
 
-                        col2.metric(
-                            "Context Recall",
-                            result[
-                                "retrieval_evaluation"
-                            ][
-                                "context_recall"
-                            ][
-                                "score"
-                            ],
-                            result[
-                                "retrieval_evaluation"
-                            ][
-                                "context_recall"
-                            ][
-                                "verdict"
-                            ]
+                        col1, col2, col3, col4 = (
+                            st.columns(4)
                         )
 
-                        # -----------------------------
-                        # Generation Evaluation
-                        # -----------------------------
-
-                        st.subheader("Generation Evaluation")
-
-                        col3, col4, col5, col6 = st.columns(4)
+                        # -------------------------------------------------
+                        # Cosine
+                        # -------------------------------------------------
 
-                        col3.metric(
+                        col1.metric(
                             "Cosine",
                             result[
                                 "generation_evaluation"
@@ -174,7 +189,11 @@ with tab1:
                             ]
                         )
 
-                        col4.metric(
+                        # -------------------------------------------------
+                        # BERTScore
+                        # -------------------------------------------------
+
+                        col2.metric(
                             "BERTScore",
                             result[
                                 "generation_evaluation"
@@ -192,7 +211,11 @@ with tab1:
                             ]
                         )
 
-                        col5.metric(
+                        # -------------------------------------------------
+                        # NLI
+                        # -------------------------------------------------
+
+                        col3.metric(
                             "NLI",
                             result[
                                 "generation_evaluation"
@@ -210,7 +233,11 @@ with tab1:
                             ]
                         )
 
-                        col6.metric(
+                        # -------------------------------------------------
+                        # Fluency
+                        # -------------------------------------------------
+
+                        col4.metric(
                             "Fluency",
                             "-",
                             result[
@@ -222,17 +249,19 @@ with tab1:
                             ]
                         )
 
-                        # -----------------------------
-                        # Detailed Result
-                        # -----------------------------
+                        # =================================================
+                        # Detailed Evaluation
+                        # =================================================
 
-                        with st.expander("View Detailed Evaluation"):
+                        with st.expander(
+                            "View Detailed Evaluation"
+                        ):
 
                             st.json(result)
 
-                    # -------------------------------------------------
-                    # API error
-                    # -------------------------------------------------
+                    # =================================================
+                    # API Error
+                    # =================================================
 
                     else:
 
@@ -242,17 +271,26 @@ with tab1:
                         )
 
                         try:
-                            error_data = response.json()
 
-                            with st.expander("View Error Details"):
-                                st.json(error_data)
+                            error_data = (
+                                response.json()
+                            )
+
+                            with st.expander(
+                                "View Error Details"
+                            ):
+
+                                st.json(
+                                    error_data
+                                )
 
                         except Exception:
+
                             pass
 
-                # -----------------------------------------------------
-                # Connection / timeout error
-                # -----------------------------------------------------
+                # =====================================================
+                # Timeout Error
+                # =====================================================
 
                 except requests.exceptions.Timeout:
 
@@ -261,10 +299,15 @@ with tab1:
                         "Please try again."
                     )
 
+                # =====================================================
+                # Connection Error
+                # =====================================================
+
                 except requests.exceptions.RequestException as e:
 
                     st.error(
-                        f"Could not connect to the evaluation API: {e}"
+                        f"Could not connect to the "
+                        f"evaluation API: {e}"
                     )
 
 
@@ -274,9 +317,17 @@ with tab1:
 
 with tab2:
 
-    st.subheader("Past Evaluations")
+    st.subheader(
+        "Past Evaluations"
+    )
+
+    # -----------------------------------------------------
+    # Refresh History
+    # -----------------------------------------------------
 
-    if st.button("Refresh History"):
+    if st.button(
+        "Refresh History"
+    ):
 
         st.rerun()
 
@@ -287,20 +338,38 @@ with tab2:
             timeout=30
         )
 
+        # =================================================
+        # Successful History Response
+        # =================================================
+
         if response.status_code == 200:
 
             data = response.json()
 
+            # -------------------------------------------------
+            # No evaluations
+            # -------------------------------------------------
+
             if data["total"] == 0:
 
-                st.info("No evaluations yet.")
+                st.info(
+                    "No evaluations yet."
+                )
 
             else:
 
+                # -------------------------------------------------
+                # Create DataFrame
+                # -------------------------------------------------
+
                 df = pd.DataFrame(
                     data["evaluations"]
                 )
 
+                # -------------------------------------------------
+                # Select columns
+                # -------------------------------------------------
+
                 df = df[
                     [
                         "id",
@@ -311,23 +380,37 @@ with tab2:
                     ]
                 ]
 
+                # -------------------------------------------------
+                # Display History
+                # -------------------------------------------------
+
                 st.dataframe(
                     df,
                     use_container_width=True
                 )
 
-                # -----------------------------
+                # =================================================
                 # Verdict Distribution
-                # -----------------------------
+                # =================================================
 
-                st.subheader("Verdict Distribution")
+                st.subheader(
+                    "Verdict Distribution"
+                )
 
                 verdict_counts = (
-                    df["final_verdict"]
+                    df[
+                        "final_verdict"
+                    ]
                     .value_counts()
                 )
 
-                st.bar_chart(verdict_counts)
+                st.bar_chart(
+                    verdict_counts
+                )
+
+        # =================================================
+        # History API Error
+        # =================================================
 
         else:
 
@@ -339,5 +422,6 @@ with tab2:
     except requests.exceptions.RequestException as e:
 
         st.error(
-            f"Could not connect to the evaluation API: {e}"
+            f"Could not connect to the "
+            f"evaluation API: {e}"
         )