Update app.py

app.py

@@ -1,6 +1,8 @@
 import gradio as gr
 import re
+import os
 from docx import Document
+from docx.shared import Inches
 from ragas import evaluate, EvaluationDataset
 from ragas.metrics import Faithfulness, FactualCorrectness, SemanticSimilarity, ResponseGroundedness, AnswerAccuracy
 from langchain_openai import ChatOpenAI, OpenAIEmbeddings
@@ -37,35 +39,44 @@ def format_ragas_results(ragas_results):
         for sample_scores in ragas_results.scores
     ]
 
-def plot_ragas_radar_chart(ragas_results):
-    # Extract first sample if multiple scores
-    if isinstance(ragas_results.scores, list):
-        scores_dict = ragas_results.scores[0]  # first sample
-    else:
-        scores_dict = ragas_results.scores
-
+def plot_radar_chart(ragas_results):
+    scores_dict = ragas_results.scores[0]
     labels = list(scores_dict.keys())
     values = list(scores_dict.values())
+    values.append(values[0])  # Close the loop
+    labels.append(labels[0])
+
+    fig, ax = plt.subplots(figsize=(6, 6), subplot_kw=dict(polar=True))
+    angles = [n / float(len(labels)) * 2 * 3.14159 for n in range(len(labels))]
+    ax.plot(angles, values, 'b-', linewidth=2)
+    ax.fill(angles, values, 'skyblue', alpha=0.4)
+    ax.set_yticklabels([])
+    ax.set_xticks(angles)
+    ax.set_xticklabels(labels, fontsize=9)
+    ax.set_title("RAGAS Evaluation Radar Chart", size=14)
 
-    # Close the loop for radar chart
-    values += values[:1]
-    angles = np.linspace(0, 2 * np.pi, len(labels), endpoint=False).tolist()
-    angles += angles[:1]
+    chart_path = "radar_chart.png"
+    fig.savefig(chart_path)
+    plt.close(fig)
+    return chart_path
 
-    fig, ax = plt.subplots(figsize=(6,6), subplot_kw=dict(polar=True))
+def generate_word_report(science_goal, ragas_results, radar_chart_path):
+    doc = Document()
+    doc.add_heading("SCDD Evaluation Report", 0)
 
-    ax.fill(angles, values, color='skyblue', alpha=0.4)
-    ax.plot(angles, values, color='blue', linewidth=2)
+    doc.add_heading("Science Goal", level=1)
+    doc.add_paragraph(science_goal)
 
-    ax.set_yticks([0.2, 0.4, 0.6, 0.8, 1.0])
-    ax.set_yticklabels(['0.2', '0.4', '0.6', '0.8', '1.0'])
+    doc.add_heading("RAGAS Evaluation Scores", level=1)
+    for metric, score in ragas_results.scores[0].items():
+        doc.add_paragraph(f"{metric}: {score*100:.2f}%")
 
-    ax.set_xticks(angles[:-1])
-    ax.set_xticklabels(labels)
+    doc.add_heading("Radar Chart", level=1)
+    doc.add_picture(radar_chart_path, width=Inches(5))
 
-    ax.set_title('RAGAS Metrics Radar Chart', size=14, y=1.05)
-    plt.tight_layout()
-    return fig
+    output_path = "SCDD_Evaluation_Report.docx"
+    doc.save(output_path)
+    return output_path
 
 def evaluate_scdd(ai_scdd_file, human_scdd_file, user_input):
     
@@ -105,9 +116,11 @@ def evaluate_scdd(ai_scdd_file, human_scdd_file, user_input):
     # RAGAS metrics outputs
     
     formatted_scores = format_ragas_results(ragas_result)
-    radar_chart = plot_ragas_radar_chart(ragas_result)
-    
-    return formatted_scores, radar_chart
+    radar_chart_path = plot_radar_chart(ragas_result)
+    word_report_path = generate_word_report(user_input, ragas_result, radar_chart_path)
+
+    score_text = "\n".join([f"{k}: {v}" for k, v in formatted_scores.items()])
+    return score_text, radar_chart_path, word_report_path
 
 
 # ----- Gradio Interface -----
@@ -117,11 +130,12 @@ interface = gr.Interface(
     inputs=[
         gr.File(label="Upload AI-Generated SCDD (Word .docx)", type='filepath'),
         gr.File(label="Upload Human-Generated SCDD (Word .docx)", type='filepath'),
-        gr.Textbox(label="Science Goal (Optional)", placeholder="Enter science goal here..."),
+        gr.Textbox(label="Science Goal", placeholder="Enter science goal here..."),
     ],
     outputs=[
         gr.JSON(label="RAGAS Scores"),
-        gr.Plot(label="Metrics Radar Chart")
+        gr.Image(label="Radar Chart"),
+        gr.File(label="Download Word Report")
     ],
     title="RAGAS Evaluation: AI vs Human SCDD",
     description="Compare AI-generated and human-generated science case documents using RAGAS LLM-powered metrics"