remove unneccesary functions

app.py

@@ -1,42 +1,12 @@
 import spaces
 import gradio as gr
 from huggingface_hub import InferenceClient
-from qdrant_client import QdrantClient, models
 from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
 from sentence_transformers import SentenceTransformer
 from huggingface_hub import login
 import torch
-import json
-import bs4
 import os
 os.environ["USE_FLASH_ATTENTION"] = "0"
-
-product_strings = []
-recipe_strings = []
-with open('./Data/product_strings.json', 'r', encoding='utf-8') as f:
-    product_strings = [product for product in json.load(f)["product_strings"]]
-with open('./Data/recipe_strings.json', 'r', encoding='utf-8') as f:
-    recipe_strings = [recipe for recipe in json.load(f)["recipe_strings"]]
-
-client = QdrantClient(":memory:") #QdrantClient("localhost:6333")
-client.set_model("sentence-transformers/all-MiniLM-L6-v2")
-client.set_sparse_model("prithivida/Splade_PP_en_v1")
-client.delete_collection(collection_name="products")
-client.create_collection(
-    collection_name="products",
-    vectors_config=client.get_fastembed_vector_params(),
-    sparse_vectors_config=client.get_fastembed_sparse_vector_params(),
-)
-client.delete_collection(collection_name="recipes")
-client.create_collection(
-    collection_name="recipes",
-    vectors_config=client.get_fastembed_vector_params(),
-    sparse_vectors_config=client.get_fastembed_sparse_vector_params(),
-)
-client.add(collection_name="products",
-        documents=product_strings)
-client.add(collection_name="recipes",
-        documents=recipe_strings)
 model_name = "LeoLM/leo-hessianai-13b-chat"
 
 last_messages = []
@@ -84,7 +54,7 @@ def generate_response(query, context, prompts, max_tokens, temperature, top_p, g
         <|im_end|>
         {"".join(last_messages)}
         <|im_start|>user
-        Frage: {query}
+        {query}
         <|im_end|>
         <|im_start|>assistant"""
 
@@ -100,7 +70,7 @@ def generate_response(query, context, prompts, max_tokens, temperature, top_p, g
         <|im_end|>
         {"".join(last_messages)}
         <|im_start|>user
-        Frage: {query}
+        {query}
         <|im_end|>
         <|im_start|>assistant"""
 
@@ -116,98 +86,28 @@ def generate_response(query, context, prompts, max_tokens, temperature, top_p, g
         <|im_end|>
         {"".join(last_messages)}
         <|im_start|>user
-        Frage: {query}
+        {query}
         <|im_end|>
         <|im_start|>assistant"""
 
     system_message = system_message_products
-    
-    if collection_name =="recipes":
+
+    if collection_name == "recipes":
         system_message = system_message_recipes
-    elif collection_name =="service":
+    elif collection_name == "services":
         system_message = system_message_support
-        
+
 
     print("Prompt: ", system_message)
 
     response = generator(system_message, do_sample=True, top_p=top_p, max_new_tokens=max_tokens, temperature=temperature)[0]["generated_text"]
-
+    print(f"""-----Response: {response}-----""")
     # Extract only the assistant's response
     if "assistant" in response:
         response = response.split("assistant").pop().strip()
 
     return response
 
-def search_qdrant_with_context(query_text, collection_name, top_k=3):
-    """Search Qdrant using a GPT-2 generated embedding."""
-    print(collection_name)
-    # print(query_embedding)
-    search_results = client.query(
-        collection_name=collection_name,
-        query_text=query_text,
-        query_filter=None,
-        limit=top_k  # Number of top results to return
-    )
-    retrieved_texts = [result.metadata for result in search_results if result.score > 0.3]
-
-    if not retrieved_texts:
-        retrieved_texts = "Keinen passenden Kontext gefunden."
-    print("Retrieved Text ", retrieved_texts)
-
-    return retrieved_texts
-
-@spaces.GPU
-def interactive_chat(query):
-    generator = get_model()
-    collection_name = "products"
-    if "rezept" in query.lower() or "gericht" in query.lower():
-        collection_name = "recipes"
-    elif "bestellung" in query.lower() or "order" in query.lower():
-        collection_name = "products"
-    print(collection_name)
-    print(query)
-    if len(query.split()) < 3:
-        return generate_response(query, "Der Kunde muss womöglich detailliertere Angaben machen, entscheide, was du sagst.", last_messages, 512, 0.2, 0.95, generator[0])
-    context = [document["document"] for document in search_qdrant_with_context(query, collection_name)]
-
-    system_message = f"""<|im_start|>system Rolle: Du bist ein KI-Assistent der die Informationen in Relation zum Kontext bewertet.
-        Oberstes Ziel: Bewerte die die Ergebnisse und stufe sie nach Relevanz in Bezug auf die Konversation ein.
-        Meta-Anweisung: Analysiere die Konversation und mache Vorschläge für Suchbegriffe in Stichpunkten.
-        Suchergebnisse: {context}
-        <|im_end|>
-        {"".join(last_messages)}
-        <|im_start|>user
-         {query}
-        <|im_end|>
-        <|im_start|>assistant"""
-    refined_context = generator[1](system_message, do_sample=True, padding=True, truncation=True, top_p=0.95, max_new_tokens=100)
-    # Retrieve relevant context from Qdrant
-    print(f"""Refined context: {refined_context[0]["generated_text"].split("assistant").pop()}""")
-
-    context = [document["document"] for document in search_qdrant_with_context(query + " " + refined_context[0]["generated_text"].split("assistant\n").pop(), collection_name)]
-    answer = generate_response(query, context, last_messages, 512, 0.2, 0.95, generator)
-    full_conv = f"<|im_start|>user {query}<|im_end|><|im_start|>assistent {answer}<|im_end|>"
-    # if len(last_messages) > 5:
-    #     last_messages.pop(0)
-    # last_messages.append(full_conv)
-    print(f"last messages: {last_messages}")
-    print()
-    return answer
-@spaces.GPU(duration=1500)
-def get_answers():
-    answers = []
-    last_messages = []
-    with open("./Data/questions.json", "r", encoding="utf-8")as f:
-        json_data = json.load(f)["questions"]
-        for (index, question) in enumerate(json_data):
-            if index <= 5:
-                continue
-            answer = interactive_chat(question)
-            answers.append(answer)
-            with open("./Data/answers.json", "w", encoding="utf-8") as file:
-                json.dump({"answers": answers}, file, ensure_ascii=False, indent=4)
-
-@spaces.GPU
 def respond(
     query,
     history: list[tuple[str, str]],