Update app.py

app.py

@@ -17,6 +17,7 @@ llm_models = [
     "meta-llama/Meta-Llama-3-8B-Instruct",
     "mistralai/Mistral-7B-Instruct-v0.2",
     "tiiuae/falcon-7b-instruct",
+    # "mistralai/Mixtral-8x22B-Instruct-v0.1",    ## 281GB>10GB
     # "NousResearch/Yarn-Mistral-7b-64k",  ## 14GB>10GB
     # "impira/layoutlm-document-qa",  ## ERR
     # "Qwen/Qwen1.5-7B",      ## 15GB
@@ -39,7 +40,6 @@ llm_models = [
 embed_models = [
     "BAAI/bge-small-en-v1.5",  # 33.4M
     "NeuML/pubmedbert-base-embeddings",
-    "sentence-transformers/all-mpnet-base-v2", # 109M
     "BAAI/llm-embedder", # 109M
     "BAAI/bge-large-en" # 335M
 ]
@@ -108,7 +108,7 @@ def respond(message, history):
         llm = HuggingFaceInferenceAPI(
             model_name=selected_llm_model_name,
             contextWindow=8192,  # Context window size (typically max length of the model)
-            maxTokens=2048,  # Tokens per response generation (512-1024 works well for detailed answers)
+            maxTokens=1024,  # Tokens per response generation (512-1024 works well for detailed answers)
             temperature=0.3,  # Lower temperature for more focused answers (0.2-0.4 for factual info)
             topP=0.9,  # Top-p sampling to control diversity while retaining quality
             frequencyPenalty=0.5,  # Slight penalty to avoid repetition
@@ -120,8 +120,8 @@ def respond(message, history):
         query_engine = vector_index.as_query_engine(llm=llm)
         bot_message = query_engine.query(message)
 
-        print(f"\n{datetime.now()}:{selected_llm_model_name} :: {message} --> {str(bot_message)}\n")
-        return f"{selected_llm_model_name}:\n\n{str(bot_message)}"
+        print(f"\n{datetime.now()}:{selected_llm_model_name}:: {message} --> {str(bot_message)}\n")
+        return f"{selected_llm_model_name}:\n{str(bot_message)}"
     except Exception as e:
         if str(e) == "'NoneType' object has no attribute 'as_query_engine'":
             return "Please upload a file."
@@ -144,14 +144,16 @@ with gr.Blocks(theme=gr.themes.Soft(font=[gr.themes.GoogleFont("Roboto Mono")]),
             gr.Markdown(md.description)
 
         with gr.TabItem("DocBot"):
+            with gr.Accordion("=== IMPORTANT: READ ME FIRST ===", open=False):
+                guid = gr.Markdown(md.guide)
             with gr.Row():
                 with gr.Column(scale=1):
                     file_input = gr.File(file_count="single", type='filepath', label="Step-1: Upload document")
-                    gr.Markdown("Dont know what to select check out in Intro tab")
+                    # gr.Markdown("Dont know what to select check out in Intro tab")
                     embed_model_dropdown = gr.Dropdown(embed_models, label="Step-2: Select Embedding", interactive=True)
                     with gr.Row():
-                        clear = gr.ClearButton()
                         btn = gr.Button("Submit", variant='primary')
+                        clear = gr.ClearButton()
                     output = gr.Text(label='Vector Index')
                     llm_model_dropdown = gr.Dropdown(llm_models, label="Step-3: Select LLM", interactive=True)
                 with gr.Column(scale=3):