Update app.py

app.py

@@ -54,8 +54,8 @@ def hybrid_search(text):
         title = row['section']
         content = row['text']
 
-        document.append(f"**{hash_id}**\n{title}\n{content}")
-        document_html.append(f'<div class="source" id="{hash_id}"><p><b>{hash_id}</b> : {title}<br>{content}</div>')
+        document.append(f"<|source_id_start|>{hash_id}<|source_id_end|>\n{content}")
+        document_html.append(f'<div class="source" id="{hash_id}"><p><b>{hash_id}</b> : <br>{content}</div>')
 
     document = "\n\n".join(document)
     document_html = '<div id="source_listing">' + "".join(document_html) + "</div>"
@@ -68,7 +68,7 @@ class CassandreChatBot:
     def predict(self, user_message):
         fiches, fiches_html = hybrid_search(user_message)
         
-        detailed_prompt = f"""### Query ###\n{user_message}\n\n### Source ###\n{fiches}\n\n### Analysis ###\n"""
+        detailed_prompt = f"""<|query_start|>{user_message}<|query_end|>\n### Source ###\n{fiches}\n\n<|source_analysis_start|>\n"""
 
         # Convert inputs to tensor
         input_ids = tokenizer.encode(detailed_prompt, return_tensors="pt").to(device)
@@ -190,13 +190,13 @@ def gradio_interface(user_message):
 demo = gr.Blocks(css=css)
 
 with demo:
-    gr.HTML("""<h1 style="text-align:center">Cassandre</h1>""")
+    gr.HTML("""<h1 style="text-align:center">pleias-RAG 1.0</h1>""")
     with gr.Row():
         with gr.Column(scale=2):
             text_input = gr.Textbox(label="Votre question ou votre instruction", lines=3)
-            text_button = gr.Button("Interroger Cassandre")
+            text_button = gr.Button("Interroger pleias-RAG")
         with gr.Column(scale=3):
-            text_output = gr.HTML(label="La réponse de Cassandre")
+            text_output = gr.HTML(label="La réponse de pleias-RAG")
     with gr.Row():
         embedding_output = gr.HTML(label="Les sources utilisées")