Update app.py

app.py

@@ -3,10 +3,12 @@ import os
 from typing import List, Dict
 from ragas import evaluate
 from ragas.metrics import (
-    ContextRecall,
-    ContextRelevancy,  
-    Faithfulness,
-    AnswerRelevancy
+    answer_relevancy,
+    faithfulness,
+    context_recall,
+    context_precision,
+    answer_correctness,
+    answer_similarity
 )
 from datasets import load_dataset
 from langchain.text_splitter import (
@@ -87,12 +89,11 @@ def load_evaluation_dataset():
     dataset = load_dataset("explodinggradients/fiqa", split="test")
     return dataset
 
-def evaluate_rag_pipeline(qa_chain, dataset):
+def prepare_ragas_dataset(qa_chain, dataset):
     # Sample a few examples for evaluation
     eval_samples = dataset.select(range(5))
     
-    # Prepare data for RAGAS evaluation
-    eval_data = []
+    ragas_dataset = []
     for sample in eval_samples:
         question = sample["question"]
         
@@ -102,33 +103,39 @@ def evaluate_rag_pipeline(qa_chain, dataset):
             "chat_history": []
         })
         
-        eval_data.append({
+        ragas_dataset.append({
             "question": question,
             "answer": response["answer"],
-            "ground_truth": sample["answer"],
-            "contexts": [doc.page_content for doc in response["source_documents"]]
+            "contexts": [doc.page_content for doc in response["source_documents"]],
+            "ground_truth": sample["answer"]
         })
     
-    # Initialize RAGAS metrics
-    metrics = [
-        ContextRecall(),
-        ContextRelevancy(),
-        Faithfulness(),
-        AnswerRelevancy()
-    ]
+    return ragas_dataset
+
+def evaluate_rag_pipeline(qa_chain, dataset):
+    ragas_dataset = prepare_ragas_dataset(qa_chain, dataset)
     
-    # Run evaluation
+    # Run RAGAS evaluation
     results = evaluate(
-        eval_data,
-        metrics=metrics
+        ragas_dataset,
+        metrics=[
+            context_precision,
+            faithfulness,
+            answer_relevancy,
+            context_recall,
+            answer_correctness,
+            answer_similarity
+        ]
     )
     
-    # Convert results to dictionary
+    # Convert results to a dictionary
     return {
-        "context_recall": float(results["context_recall"]),
-        "context_relevancy": float(results["context_relevancy"]),
+        "context_precision": float(results["context_precision"]),
         "faithfulness": float(results["faithfulness"]),
-        "answer_relevancy": float(results["answer_relevancy"])
+        "answer_relevancy": float(results["answer_relevancy"]),
+        "context_recall": float(results["context_recall"]),
+        "answer_correctness": float(results["answer_correctness"]),
+        "answer_similarity": float(results["answer_similarity"])
     }
 
 # Initialize langchain LLM chain