Final version: multi-agent stats assistant using RAG + OpenAI router

app.py

@@ -3,15 +3,14 @@ import gradio as gr
 import torch
 from langchain_community.document_loaders import PyPDFLoader
 from langchain.text_splitter import RecursiveCharacterTextSplitter
-from langchain_community.embeddings import HuggingFaceEmbeddings
+from langchain_huggingface import HuggingFacePipeline, HuggingFaceEmbeddings
 from langchain_community.vectorstores import FAISS
 from langchain.chains import RetrievalQA
-from langchain_community.llms import HuggingFacePipeline
 from langchain_openai import ChatOpenAI
 from transformers.pipelines import pipeline
 
 # ✅ Load writer model and wrap it for LangChain
-writer_model = pipeline("text2text-generation", model="BivinSadler/llama3-finetuned-Statistics")
+writer_model = pipeline("text-generation", model="BivinSadler/llama3-finetuned-Statistics", return_full_text=False)
 writer_llm = HuggingFacePipeline(pipeline=writer_model)
 
 # ✅ RAG Agent Builder
@@ -65,7 +64,7 @@ Answer:
 
 Write your response in 2–3 sentences. Avoid technical jargon.
 """
-    result = writer_model(prompt, max_new_tokens=200, do_sample=False)
+    result = writer_model(prompt, max_new_tokens=200)
     return result[0]['generated_text']
 
 # ✅ Main logic

requirements.txt

@@ -9,4 +9,6 @@ sentence-transformers
 gradio
 torch
 tiktoken
+huggingface-hub
+accelerate