Update app.py

app.py

@@ -7,6 +7,8 @@ import gradio as gr
 from langchain.evaluation import load_evaluator
 from pprint import pprint as print
 import time
+from utils import *
+from beschreibungen import *
 
 #from langchain.chains import LLMChain, RetrievalQA
 #from langchain.retrievers.self_query.base import SelfQueryRetriever
@@ -258,14 +260,27 @@ def generate_prompt_with_history_openai(prompt, history):
 
     history_openai_format.append({"role": "user", "content": prompt})
     return history_openai_format
-        
+
+##############################################
+#History - die Frage oder das File eintragen...
+##############################################
+def add_text(history, text):
+    history = history + [(text, None)]
+    return history, "" #gr.Textbox(value="", interactive=False)
+    
+def add_file(history, file, prompt):
+    if (prompt == ""):
+        history = history + [((file.name,), None)]
+    else:
+        history = history + [((file.name,), None), (prompt, None)]
+    return history, ""
 
 ##############################################
 ##############################################
 ##############################################
 # generate function
 ##############################################
-def generate(text, history, rag_option, model_option,  k=3, temperature=0.5, max_new_tokens=4048, top_p=0.6, repetition_penalty=1.3,):
+def generate(history, text, rag_option, model_option,  k=3, top_p=0.6, temperature=0.5, max_new_tokens=4048, max_context_length_tokens=2048, repetition_penalty=1.3,):
     #mit RAG
     if (rag_option is None):
         raise gr.Error("Retrieval Augmented Generation ist erforderlich.")
@@ -323,11 +338,23 @@ def generate(text, history, rag_option, model_option,  k=3, temperature=0.5, max
     #return chatbot_message 
     
     #Antwort als Stream ausgeben... 
-    for i in range(len(chatbot_message)):
+    #for i in range(len(chatbot_message)):
+        #time.sleep(0.03)
+        #yield  chatbot_message[: i+1]
+
+    #Antwort als Stream ausgeben... 
+    history[-1][1] = ""
+    for character in chatbot_message:
+        history[-1][1] += character
         time.sleep(0.03)
-        yield  chatbot_message[: i+1]
-    
-    
+        yield history, "Generating"
+        if shared_state.interrupted:
+            shared_state.recover()
+            try:
+                yield history, "Stop: Success"
+                return
+            except:
+                pass
     
     
     #zum Evaluieren:
@@ -354,11 +381,134 @@ evaluator = load_evaluator("criteria", criteria="conciseness", llm=evaluation_ll
 #Beschreibung oben in GUI
 ################################################
 print ("Start GUI")
+with open("custom.css", "r", encoding="utf-8") as f:
+    customCSS = f.read()
+
+with gr.Blocks(css=customCSS, theme=small_and_beautiful_theme) as demo:
+    history = gr.State([])
+    with gr.Row():
+        gr.HTML("LI Chatot")
+        status_display = gr.Markdown("Success", elem_id="status_display")
+    gr.Markdown(description_top)
+    with gr.Row():
+        with gr.Column(scale=5):
+            with gr.Row():
+                chatbot = gr.Chatbot(elem_id="chuanhu_chatbot")
+            with gr.Row():
+                with gr.Column(scale=12):
+                    user_input = gr.Textbox(
+                        show_label=False, placeholder="Gib hier deinen Prompt ein...",
+                        container=False
+                    )
+                with gr.Column(min_width=70, scale=1):
+                    submitBtn = gr.Button("Senden")
+                with gr.Column(min_width=70, scale=1):
+                    cancelBtn = gr.Button("Stop")
+            with gr.Row():
+                emptyBtn = gr.ClearButton( [user_input, chatbot], value="🧹 Neue Session")  
+                btn = gr.UploadButton("📁", file_types=["image", "video", "audio"])
+
+        with gr.Column():
+            with gr.Column(min_width=50, scale=1):
+                with gr.Tab(label="Parameter Einstellung"):
+                    gr.Markdown("# Parameters")
+                    rag_option = gr.Radio(["Aus", "An"], label="RAG - LI Erweiterungen", value = "Aus")
+                    model_option = gr.Radio(["HuggingFace1", "HuggingFace2"], label="Modellauswahl", value = "HuggingFace1")
+                    
+                    top_p = gr.Slider(
+                        minimum=-0,
+                        maximum=1.0,
+                        value=0.95,
+                        step=0.05,
+                        interactive=True,
+                        label="Top-p",
+                    )
+                    temperature = gr.Slider(
+                        minimum=0.1,
+                        maximum=2.0,
+                        value=1,
+                        step=0.1,
+                        interactive=True,
+                        label="Temperature",
+                    )
+                    max_length_tokens = gr.Slider(
+                        minimum=0,
+                        maximum=512,
+                        value=512,
+                        step=8,
+                        interactive=True,
+                        label="Max Generation Tokens",
+                    )
+                    max_context_length_tokens = gr.Slider(
+                        minimum=0,
+                        maximum=4096,
+                        value=2048,
+                        step=128,
+                        interactive=True,
+                        label="Max History Tokens",
+                    )
+                    repetition_penalty=gr.Slider(label="Repetition penalty", value=1.2, minimum=1.0, maximum=2.0, step=0.05, interactive=True, info="Strafe für wiederholte Tokens", visible=True)
+                    anzahl_docs = gr.Slider(label="Anzahl Vergleichsdokumente", value=3, minimum=1, maximum=10, step=1, interactive=True, info="wie viele Dokumententeile aus dem Vektorstore an den prompt gehängt werden", visible=True),
+
+    gr.Markdown(description)
+
+    #Argumente für generate Funktion als Input
+    predict_args = dict(
+        fn=generate,
+        inputs=[
+            user_input,
+            chatbot,
+            #history,
+            rag_option,
+            model_option,
+            anzahl_docs,
+            top_p,
+            temperature,
+            max_length_tokens,
+            max_context_length_tokens,
+        ],
+        outputs=[ chatbot,   status_display], #[ chatbot, history, status_display],
+        show_progress=True,
+    )
+
+
+    reset_args = dict(
+        fn=reset_textbox, inputs=[], outputs=[user_input, status_display]
+    )
+
+    # Chatbot
+    transfer_input_args = dict(
+        fn=add_text, inputs=[ chatbot, user_input], outputs=[chatbot,  user_input], show_progress=True
+    )
+
+    predict_event1 = user_input.submit(**transfer_input_args ,queue=False,).then(**predict_args)
+    predict_event3 = btn.upload(add_file, [chatbot, btn, user_input], [chatbot, user_input],queue=False, ).then(**predict_args)
+    predict_event2 = submitBtn.click(**transfer_input_args,queue=False,).then(**predict_args)
+
+    cancelBtn.click(
+        cancel_outputing, [], [status_display],
+        cancels=[
+            predict_event1,predict_event2, predict_event3
+        ]
+    )
+demo.title = "LI-ChatBot"
+
+demo.queue().launch(debug=True)
+
+
+
+
+
+
+
+
+
+"""
 
-description = """<strong>Information:</strong> Hier wird ein <strong>Large Language Model (LLM)</strong> mit 
+description = <strong>Information:</strong> Hier wird ein <strong>Large Language Model (LLM)</strong> mit 
                  <strong>Retrieval Augmented Generation (RAG)</strong> auf <strong>externen Daten</strong> verwendet.\n\n
-                 """
-css = """.toast-wrap { display: none !important } """
+                 
+css = .toast-wrap { display: none !important } 
 examples=[['Was ist ChtGPT-4?'],['schreibe ein Python Programm, dass die GPT-4 API aufruft.']]
 
 additional_inputs = [
@@ -390,3 +540,4 @@ with gr.Blocks() as demo:
         #chatbot_stream.like(vote, None, None)
         chat_interface_stream.queue().launch()   
         
+"""