open-webui/backend/apps/rag/main.py

185 lines
4.9 KiB
Python
Raw Normal View History

2024-01-07 08:40:51 +01:00
from fastapi import (
FastAPI,
Request,
Depends,
HTTPException,
status,
UploadFile,
File,
Form,
)
2024-01-07 07:07:20 +01:00
from fastapi.middleware.cors import CORSMiddleware
2024-01-07 10:40:36 +01:00
import os, shutil
2024-01-07 07:07:20 +01:00
2024-01-07 07:59:22 +01:00
from chromadb.utils import embedding_functions
2024-01-07 07:07:20 +01:00
2024-01-07 08:40:51 +01:00
from langchain.document_loaders import WebBaseLoader, TextLoader, PyPDFLoader
2024-01-07 07:59:22 +01:00
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import Chroma
from langchain.chains import RetrievalQA
from pydantic import BaseModel
from typing import Optional
import uuid
2024-01-07 10:40:36 +01:00
from config import UPLOAD_DIR, EMBED_MODEL, CHROMA_CLIENT, CHUNK_SIZE, CHUNK_OVERLAP
2024-01-07 07:59:22 +01:00
from constants import ERROR_MESSAGES
EMBEDDING_FUNC = embedding_functions.SentenceTransformerEmbeddingFunction(
model_name=EMBED_MODEL
)
2024-01-07 07:07:20 +01:00
app = FastAPI()
origins = ["*"]
app.add_middleware(
CORSMiddleware,
allow_origins=origins,
allow_credentials=True,
allow_methods=["*"],
allow_headers=["*"],
)
2024-01-07 08:40:51 +01:00
class CollectionNameForm(BaseModel):
2024-01-07 07:59:22 +01:00
collection_name: Optional[str] = "test"
2024-01-07 08:40:51 +01:00
class StoreWebForm(CollectionNameForm):
url: str
2024-01-07 10:40:36 +01:00
def store_data_in_vector_db(data, collection_name) -> bool:
2024-01-07 07:59:22 +01:00
text_splitter = RecursiveCharacterTextSplitter(
chunk_size=CHUNK_SIZE, chunk_overlap=CHUNK_OVERLAP
)
docs = text_splitter.split_documents(data)
texts = [doc.page_content for doc in docs]
metadatas = [doc.metadata for doc in docs]
2024-01-07 10:40:36 +01:00
try:
collection = CHROMA_CLIENT.create_collection(
name=collection_name, embedding_function=EMBEDDING_FUNC
)
2024-01-07 07:59:22 +01:00
2024-01-07 10:40:36 +01:00
collection.add(
documents=texts, metadatas=metadatas, ids=[str(uuid.uuid1()) for _ in texts]
)
return True
except Exception as e:
print(e)
if e.__class__.__name__ == "UniqueConstraintError":
return True
return False
2024-01-07 07:59:22 +01:00
2024-01-07 07:07:20 +01:00
@app.get("/")
async def get_status():
return {"status": True}
2024-01-07 07:59:22 +01:00
@app.get("/query/{collection_name}")
def query_collection(collection_name: str, query: str, k: Optional[int] = 4):
2024-01-07 10:59:00 +01:00
try:
collection = CHROMA_CLIENT.get_collection(
name=collection_name,
)
result = collection.query(query_texts=[query], n_results=k)
2024-01-07 07:59:22 +01:00
2024-01-07 10:59:00 +01:00
return result
except Exception as e:
print(e)
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT(e),
)
2024-01-07 07:59:22 +01:00
@app.post("/web")
def store_web(form_data: StoreWebForm):
# "https://www.gutenberg.org/files/1727/1727-h/1727-h.htm"
try:
loader = WebBaseLoader(form_data.url)
data = loader.load()
store_data_in_vector_db(data, form_data.collection_name)
2024-01-07 09:57:10 +01:00
return {"status": True, "collection_name": form_data.collection_name}
2024-01-07 07:59:22 +01:00
except Exception as e:
print(e)
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT(e),
)
@app.post("/doc")
2024-01-07 08:40:51 +01:00
def store_doc(collection_name: str = Form(...), file: UploadFile = File(...)):
2024-01-07 07:59:22 +01:00
# "https://www.gutenberg.org/files/1727/1727-h/1727-h.htm"
2024-01-07 10:54:58 +01:00
file.filename = f"{collection_name}-{file.filename}"
2024-01-07 08:40:51 +01:00
if file.content_type not in ["application/pdf", "text/plain"]:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.FILE_NOT_SUPPORTED,
)
2024-01-07 07:59:22 +01:00
try:
2024-01-07 08:40:51 +01:00
filename = file.filename
2024-01-07 10:40:36 +01:00
file_path = f"{UPLOAD_DIR}/{filename}"
2024-01-07 07:59:22 +01:00
contents = file.file.read()
2024-01-07 08:40:51 +01:00
with open(file_path, "wb") as f:
2024-01-07 07:59:22 +01:00
f.write(contents)
f.close()
2024-01-07 08:40:51 +01:00
if file.content_type == "application/pdf":
loader = PyPDFLoader(file_path)
elif file.content_type == "text/plain":
loader = TextLoader(file_path)
data = loader.load()
2024-01-07 10:40:36 +01:00
result = store_data_in_vector_db(data, collection_name)
if result:
return {"status": True, "collection_name": collection_name}
else:
raise HTTPException(
status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
detail=ERROR_MESSAGES.DEFAULT(),
)
2024-01-07 07:59:22 +01:00
except Exception as e:
print(e)
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.DEFAULT(e),
)
2024-01-07 10:40:36 +01:00
@app.get("/reset/db")
2024-01-07 07:59:22 +01:00
def reset_vector_db():
CHROMA_CLIENT.reset()
2024-01-07 10:40:36 +01:00
@app.get("/reset")
def reset():
folder = f"{UPLOAD_DIR}"
for filename in os.listdir(folder):
file_path = os.path.join(folder, filename)
try:
if os.path.isfile(file_path) or os.path.islink(file_path):
os.unlink(file_path)
elif os.path.isdir(file_path):
shutil.rmtree(file_path)
except Exception as e:
print("Failed to delete %s. Reason: %s" % (file_path, e))
try:
CHROMA_CLIENT.reset()
except Exception as e:
print(e)
2024-01-07 07:59:22 +01:00
return {"status": True}