Spaces:
Sleeping
Sleeping
Delete tools.py
Browse files
tools.py
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
from langchain_core.tools import tool
|
| 2 |
-
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
| 3 |
-
from langchain_community.vectorstores import FAISS
|
| 4 |
-
from langchain_community.document_loaders import PyPDFLoader
|
| 5 |
-
from langchain_openai import OpenAIEmbeddings
|
| 6 |
-
from langchain_community.tools import WikipediaQueryRun, ArxivQueryRun
|
| 7 |
-
from langchain_community.utilities import WikipediaAPIWrapper, ArxivAPIWrapper
|
| 8 |
-
from langchain_community.tools.tavily_search import TavilySearchResults
|
| 9 |
-
from dotenv import load_dotenv
|
| 10 |
-
import os
|
| 11 |
-
|
| 12 |
-
load_dotenv()
|
| 13 |
-
|
| 14 |
-
# ==============================
|
| 15 |
-
# CONFIG
|
| 16 |
-
# ==============================
|
| 17 |
-
VECTORSTORE_DIR = "data/vectorstore"
|
| 18 |
-
os.makedirs(VECTORSTORE_DIR, exist_ok=True)
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
# ==============================
|
| 22 |
-
# VECTOR STORE CREATION
|
| 23 |
-
# ==============================
|
| 24 |
-
def build_vectorstore(file_path: str):
|
| 25 |
-
loader = PyPDFLoader(file_path)
|
| 26 |
-
documents = loader.load()
|
| 27 |
-
|
| 28 |
-
splitter = RecursiveCharacterTextSplitter(
|
| 29 |
-
chunk_size=500,
|
| 30 |
-
chunk_overlap=100
|
| 31 |
-
)
|
| 32 |
-
|
| 33 |
-
chunks = splitter.split_documents(documents)
|
| 34 |
-
|
| 35 |
-
embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
|
| 36 |
-
vectorstore = FAISS.from_documents(chunks, embeddings)
|
| 37 |
-
|
| 38 |
-
vectorstore.save_local(VECTORSTORE_DIR)
|
| 39 |
-
return vectorstore
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
def update_retriever(file_path: str):
|
| 43 |
-
"""Rebuild vectorstore when a new document is uploaded."""
|
| 44 |
-
build_vectorstore(file_path)
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
# ==============================
|
| 48 |
-
# RAG TOOL (HF SAFE)
|
| 49 |
-
# ==============================
|
| 50 |
-
def create_rag_tool():
|
| 51 |
-
|
| 52 |
-
@tool
|
| 53 |
-
def rag_search(query: str) -> str:
|
| 54 |
-
"""
|
| 55 |
-
Retrieve relevant information from uploaded documents.
|
| 56 |
-
"""
|
| 57 |
-
|
| 58 |
-
if not os.path.exists(os.path.join(VECTORSTORE_DIR, "index.faiss")):
|
| 59 |
-
return "No document has been uploaded yet."
|
| 60 |
-
|
| 61 |
-
embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
|
| 62 |
-
|
| 63 |
-
vectorstore = FAISS.load_local(
|
| 64 |
-
VECTORSTORE_DIR,
|
| 65 |
-
embeddings,
|
| 66 |
-
allow_dangerous_deserialization=True
|
| 67 |
-
)
|
| 68 |
-
|
| 69 |
-
retriever = vectorstore.as_retriever(search_kwargs={"k": 4})
|
| 70 |
-
docs = retriever.invoke(query)
|
| 71 |
-
|
| 72 |
-
if not docs:
|
| 73 |
-
return "No relevant information found in the document."
|
| 74 |
-
|
| 75 |
-
return "\n\n".join(d.page_content for d in docs)
|
| 76 |
-
|
| 77 |
-
return rag_search
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
# ==============================
|
| 81 |
-
# EXTRA TOOLS
|
| 82 |
-
# ==============================
|
| 83 |
-
|
| 84 |
-
@tool
|
| 85 |
-
def wikipedia_search(query: str) -> dict:
|
| 86 |
-
"""Search Wikipedia."""
|
| 87 |
-
try:
|
| 88 |
-
return {"results": WikipediaQueryRun(api_wrapper=WikipediaAPIWrapper()).run(query)}
|
| 89 |
-
except Exception as e:
|
| 90 |
-
return {"error": str(e)}
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
@tool
|
| 94 |
-
def arxiv_search(query: str) -> dict:
|
| 95 |
-
"""Search academic papers on arXiv."""
|
| 96 |
-
try:
|
| 97 |
-
return {"results": ArxivQueryRun(api_wrapper=ArxivAPIWrapper()).run(query)}
|
| 98 |
-
except Exception as e:
|
| 99 |
-
return {"error": str(e)}
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
@tool
|
| 103 |
-
def tavily_search(query: str) -> dict:
|
| 104 |
-
"""Search the web using Tavily."""
|
| 105 |
-
try:
|
| 106 |
-
return {"results": TavilySearchResults(max_results=5).run(query)}
|
| 107 |
-
except Exception as e:
|
| 108 |
-
return {"error": str(e)}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|