Update rag_pipeline.py

rag_pipeline.py

@@ -1,8 +1,90 @@
+# rag_pipeline.py
+
+import os
+import pickle
+import numpy as np
+import faiss
+from sentence_transformers import SentenceTransformer
+from langchain_community.embeddings import HuggingFaceEmbeddings
+from langchain.text_splitter import RecursiveCharacterTextSplitter, SentenceTransformersTokenTextSplitter
+import umap.umap_ as umap
+from dotenv import load_dotenv
+from groq import Groq
+from openai import OpenAI
+import tqdm
+
+# === Load environment variables (handled by Hugging Face Secrets in deployment) ===
+openai_api_key = os.getenv("OPENAI_API_KEY")
+groq_api_key = os.getenv("GROQ_API_KEY")
+
+groq_client = Groq(api_key=groq_api_key) if groq_api_key else None
+openai_client = OpenAI(api_key=openai_api_key) if openai_api_key else None
+
+# === Load FAISS Index and Chunks ===
+index = faiss.read_index("faiss/faiss_index.index")
+with open("faiss/chunks_mapping.pkl", "rb") as f:
+    token_split_texts = pickle.load(f)
+
+# === Load SentenceTransformer model ===
+model = SentenceTransformer("Sahajtomar/German-semantic")
+chunk_embeddings = model.encode(token_split_texts, convert_to_numpy=True)
+
+# === Fit UMAP ===
+umap_transform = umap.UMAP(random_state=0, transform_seed=0).fit(chunk_embeddings)
+
+def project_embeddings(embeddings, umap_transform):
+    umap_embeddings = np.empty((len(embeddings), 2))
+    for i, embedding in enumerate(tqdm.tqdm(embeddings, desc="Projecting Embeddings")):
+        umap_embeddings[i] = umap_transform.transform([embedding])
+    return umap_embeddings
+
+def retrieve(query, k=5):
+    query_embedding = model.encode([query], convert_to_numpy=True)
+    distances, indices = index.search(query_embedding, k)
+    retrieved_texts = [token_split_texts[i] for i in indices[0]]
+    retrieved_embeddings = np.array([chunk_embeddings[i] for i in indices[0]])
+    return retrieved_texts, retrieved_embeddings, distances[0]
+
+def build_prompt(query, retrieved_texts):
+    context_block = "\n\n".join(retrieved_texts)
+    prompt = f"""Beantworte die folgende Frage basierend auf dem gegebenen Kontext.
+
+Kontext:
+{context_block}
+
+Frage:
+{query}
+"""
+    return prompt
+
+def ask_groq_llm(prompt):
+    if not groq_client:
+        return "[Fehler] Kein Groq API Key vorhanden."
+    response = groq_client.chat.completions.create(
+        model="llama3-70b-8192",
+        messages=[
+            {"role": "system", "content": "Du bist ein hilfreicher Assistent."},
+            {"role": "user", "content": prompt}
+        ]
+    )
+    return response.choices[0].message.content.strip()
+
+def ask_openai_llm(prompt):
+    if not openai_client:
+        return "[Fehler] Kein OpenAI API Key vorhanden."
+    response = openai_client.chat.completions.create(
+        model="gpt-4",
+        messages=[
+            {"role": "system", "content": "You are a helpful assistant."},
+            {"role": "user", "content": prompt}
+        ]
+    )
+    return response.choices[0].message.content.strip()
+
 def run_qa_pipeline(user_query, k=5):
     retrieved_texts, _, _ = retrieve(user_query, k=k)
     prompt = build_prompt(user_query, retrieved_texts)
-    
     answer_groq = ask_groq_llm(prompt)
     answer_openai = ask_openai_llm(prompt)
 
-    return f" Groq LLaMA 3:\n{answer_groq}\n\n OpenAI GPT-4:\n{answer_openai}"
+    return f"\U0001f999 Groq LLaMA 3 Antwort:\n{answer_groq}\n\n\U0001f52e OpenAI GPT-4 Antwort:\n{answer_openai}"