Update app.py

app.py

@@ -1,6 +1,15 @@
+import json
 from huggingface_hub import InferenceClient
 import gradio as gr
-import os 
+import os
+
+# Laden der Prompts aus der JSON-Datei
+def load_prompts_from_json(file_path):
+    with open(file_path, 'r') as file:
+        return json.load(file)
+
+# Angenommen, Sie haben eine JSON-Datei namens 'prompts.json'
+prompts = load_prompts_from_json('prompts.json')
 
 # Klient für die Inferenz
 client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
@@ -8,21 +17,20 @@ client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
 # Geheime Eingabeaufforderung aus Umgebungsvariablen
 secret_prompt = os.getenv("SECRET_PROMPT")
 
-def format_prompt(new_message, history):
-    prompt = secret_prompt
+def format_prompt(new_message, history, prompt_type='default'):
+    prompt = prompts.get(prompt_type, secret_prompt)
     for user_msg, bot_msg in history:
         prompt += f"[INST] {user_msg} [/INST]"
         prompt += f" {bot_msg}</s> "
     prompt += f"[INST] {new_message} [/INST]"
     return prompt
 
-def generate(prompt, history, temperature=0.25, max_new_tokens=512, top_p=0.95, repetition_penalty=1.0):
+def generate(prompt, history, temperature=0.25, max_new_tokens=512, top_p=0.95, repetition_penalty=1.0, prompt_type='default'):
     # Konfiguration der Parameter
     temperature = float(temperature)
     if temperature < 1e-2:
         temperature = 1e-2
     top_p = float(top_p)
-
     generate_kwargs = dict(
         temperature=temperature,
         max_new_tokens=max_new_tokens,
@@ -31,11 +39,9 @@ def generate(prompt, history, temperature=0.25, max_new_tokens=512, top_p=0.95,
         do_sample=True,
         seed=727,
     )
-
-    formatted_prompt = format_prompt(prompt, history)
+    formatted_prompt = format_prompt(prompt, history, prompt_type)
     stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
     output = ""
-
     for response in stream:
         output += response.token.text
         yield output
@@ -44,8 +50,25 @@ def generate(prompt, history, temperature=0.25, max_new_tokens=512, top_p=0.95,
 # Chatbot ohne Avatare und mit transparentem Design
 samir_chatbot = gr.Chatbot(bubble_full_width=True, show_label=False, show_copy_button=False, likeable=False)
 
+# Dropdown für Prompt-Typen
+prompt_type_dropdown = gr.Dropdown(choices=list(prompts.keys()), label="Prompt-Typ", value='default')
+
 # Minimalistisches Theme und Konfiguration der Gradio-Demo
 theme = 'syddharth/gray-minimal'
-demo = gr.ChatInterface(fn=generate, chatbot=samir_chatbot, title="Tutorial Master", theme=theme)
+demo = gr.Interface(
+    fn=generate,
+    inputs=[
+        gr.Textbox(lines=2, label="Eingabe"),
+        "state",
+        gr.Slider(0, 1, value=0.25, label="Temperature"),
+        gr.Slider(1, 2048, value=512, step=1, label="Max Tokens"),
+        gr.Slider(0, 1, value=0.95, label="Top P"),
+        gr.Slider(1, 2, value=1.0, label="Repetition Penalty"),
+        prompt_type_dropdown
+    ],
+    outputs=[samir_chatbot],
+    title="Tutorial Master",
+    theme=theme
+)
 
 demo.queue().launch(show_api=False)