Spaces:
Sleeping
Sleeping
First commit
Browse files- Dockerfile +30 -0
- Dockerfile.milvus +32 -0
- app/main.py +135 -0
- app/milvus_singleton.py +25 -0
- app/requirements.txt +8 -0
Dockerfile
ADDED
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
FROM python:3.10.8
|
2 |
+
|
3 |
+
WORKDIR /app
|
4 |
+
|
5 |
+
COPY ./app/requirements.txt /app/requirements.txt
|
6 |
+
|
7 |
+
# Create cache and milvus_data directories and set permissions
|
8 |
+
RUN mkdir -p /app/cache /app/milvus_data && chmod -R 777 /app/cache /app/milvus_data
|
9 |
+
|
10 |
+
# Install dependencies
|
11 |
+
RUN pip install --no-cache-dir --upgrade -r requirements.txt
|
12 |
+
|
13 |
+
# Create a non-root user
|
14 |
+
RUN useradd -m -u 1000 user
|
15 |
+
USER user
|
16 |
+
|
17 |
+
# Set environment variables for Hugging Face cache and Milvus data
|
18 |
+
ENV HF_HOME=/app/cache \
|
19 |
+
HF_MODULES_CACHE=/app/cache/hf_modules \
|
20 |
+
MILVUS_DATA_DIR=/app/milvus_data \
|
21 |
+
HF_WORKER_COUNT=1
|
22 |
+
|
23 |
+
# Copy the application code
|
24 |
+
COPY ./app /app
|
25 |
+
|
26 |
+
# Expose the port Uvicorn will run on
|
27 |
+
EXPOSE 7860
|
28 |
+
|
29 |
+
# Start Uvicorn
|
30 |
+
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]
|
Dockerfile.milvus
ADDED
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
FROM python:3.11
|
2 |
+
|
3 |
+
# Install Milvus dependencies
|
4 |
+
USER root
|
5 |
+
RUN apt-get update && apt-get install -y wget ffmpeg libsm6 libxext6 libaio1
|
6 |
+
|
7 |
+
# Download and install Milvus
|
8 |
+
RUN wget https://github.com/milvus-io/milvus/releases/download/v2.3.7/milvus_2.3.7-1_amd64.deb && \
|
9 |
+
dpkg -i milvus_2.3.7-1_amd64.deb && \
|
10 |
+
apt-get -f install && \
|
11 |
+
apt-get clean && \
|
12 |
+
rm milvus_2.3.7-1_amd64.deb
|
13 |
+
|
14 |
+
# Create a directory for Milvus data
|
15 |
+
RUN mkdir -p /milvus/data
|
16 |
+
|
17 |
+
# Set up Milvus user
|
18 |
+
RUN useradd -m -u 1000 milvus
|
19 |
+
USER milvus
|
20 |
+
|
21 |
+
# Set Milvus environment variables
|
22 |
+
ENV MILVUS_HOME=/home/milvus
|
23 |
+
ENV PATH=$MILVUS_HOME/bin:$PATH
|
24 |
+
|
25 |
+
# Set working directory
|
26 |
+
WORKDIR $MILVUS_HOME
|
27 |
+
|
28 |
+
# Expose Milvus ports
|
29 |
+
EXPOSE 19530
|
30 |
+
|
31 |
+
# Start Milvus server
|
32 |
+
CMD ["milvus", "run", "standalone", "-d", "/milvus/data"]
|
app/main.py
ADDED
@@ -0,0 +1,135 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from io import BytesIO
|
2 |
+
from fastapi import FastAPI, Form, Depends, Request, File, UploadFile
|
3 |
+
from fastapi.encoders import jsonable_encoder
|
4 |
+
from fastapi.responses import JSONResponse
|
5 |
+
from fastapi.middleware.cors import CORSMiddleware
|
6 |
+
from pydantic import BaseModel
|
7 |
+
from pymilvus import connections, utility, Collection, CollectionSchema, FieldSchema, DataType
|
8 |
+
import os
|
9 |
+
import pypdf
|
10 |
+
from uuid import uuid4
|
11 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
12 |
+
from sentence_transformers import SentenceTransformer
|
13 |
+
import torch
|
14 |
+
from app.milvus_singleton import MilvusClientSingleton
|
15 |
+
|
16 |
+
# Set environment variables for Hugging Face cache
|
17 |
+
os.environ['HF_HOME'] = '/app/cache'
|
18 |
+
os.environ['HF_MODULES_CACHE'] = '/app/cache/hf_modules'
|
19 |
+
|
20 |
+
# Embedding model
|
21 |
+
embedding_model = SentenceTransformer('Alibaba-NLP/gte-large-en-v1.5',
|
22 |
+
trust_remote_code=True,
|
23 |
+
device='cuda' if torch.cuda.is_available() else 'cpu',
|
24 |
+
cache_folder='/app/cache')
|
25 |
+
|
26 |
+
# Milvus connection details
|
27 |
+
collection_name="rag"
|
28 |
+
milvus_uri = os.getenv("MILVUS_URI", "sqlite:///$MILVUS_DATA_DIR/milvus_demo.db")
|
29 |
+
|
30 |
+
# Initialize Milvus client using singleton
|
31 |
+
milvus_client = MilvusClientSingleton.get_instance(uri=milvus_uri)
|
32 |
+
|
33 |
+
def document_to_embeddings(content:str) -> list:
|
34 |
+
return embedding_model.encode(content, show_progress_bar=True)
|
35 |
+
|
36 |
+
app = FastAPI()
|
37 |
+
|
38 |
+
# Add CORS middleware
|
39 |
+
app.add_middleware(
|
40 |
+
CORSMiddleware,
|
41 |
+
allow_origins=["*"], # Replace with allowed origins for production
|
42 |
+
allow_credentials=True,
|
43 |
+
allow_methods=["*"],
|
44 |
+
allow_headers=["*"],
|
45 |
+
)
|
46 |
+
|
47 |
+
def split_documents(document_data):
|
48 |
+
splitter = RecursiveCharacterTextSplitter(chunk_size=512, chunk_overlap=10)
|
49 |
+
return splitter.split_text(document_data)
|
50 |
+
|
51 |
+
def create_a_collection(milvus_client, collection_name):
|
52 |
+
# Define the fields for the collection
|
53 |
+
id_field = FieldSchema(name="id", dtype=DataType.VARCHAR, max_length=40, is_primary=True)
|
54 |
+
content_field = FieldSchema(name="content", dtype=DataType.VARCHAR, max_length=4096)
|
55 |
+
vector_field = FieldSchema(name="vector", dtype=DataType.FLOAT_VECTOR, dim=1024)
|
56 |
+
# Define the schema for the collection
|
57 |
+
schema = CollectionSchema(fields=[id_field, content_field, vector_field])
|
58 |
+
# Create the collection
|
59 |
+
milvus_client.create_collection(
|
60 |
+
collection_name=collection_name,
|
61 |
+
schema=schema
|
62 |
+
)
|
63 |
+
connections.connect(uri=milvus_uri)
|
64 |
+
collection = Collection(name=collection_name)
|
65 |
+
# Create an index for the collection
|
66 |
+
# IVF_FLAT index is used here, with metric_type COSINE
|
67 |
+
index_params = {
|
68 |
+
"index_type": "FLAT",
|
69 |
+
"metric_type": "COSINE",
|
70 |
+
"params": {
|
71 |
+
"nlist": 128
|
72 |
+
}
|
73 |
+
}
|
74 |
+
# Create the index on the vector field
|
75 |
+
collection.create_index(
|
76 |
+
field_name="vector",
|
77 |
+
index_params=index_params
|
78 |
+
)
|
79 |
+
|
80 |
+
@app.get("/")
|
81 |
+
async def root():
|
82 |
+
return {"message": "Hello World"}
|
83 |
+
|
84 |
+
@app.post("/insert")
|
85 |
+
async def insert(file: UploadFile = File(...)):
|
86 |
+
contents = await file.read()
|
87 |
+
if not milvus_client.has_collection(collection_name):
|
88 |
+
create_a_collection(milvus_client, collection_name)
|
89 |
+
contents = pypdf.PdfReader(BytesIO(contents))
|
90 |
+
extracted_text = ""
|
91 |
+
for page_num in range(len(contents.pages)):
|
92 |
+
page = contents.pages[page_num]
|
93 |
+
extracted_text += page.extract_text()
|
94 |
+
splitted_document_data = split_documents(extracted_text)
|
95 |
+
print(splitted_document_data)
|
96 |
+
data_objects = []
|
97 |
+
for doc in splitted_document_data:
|
98 |
+
data = {
|
99 |
+
"id": str(uuid4()),
|
100 |
+
"vector": document_to_embeddings(doc),
|
101 |
+
"content": doc,
|
102 |
+
}
|
103 |
+
data_objects.append(data)
|
104 |
+
print(data_objects)
|
105 |
+
try:
|
106 |
+
milvus_client.insert(collection_name=collection_name, data=data_objects)
|
107 |
+
except Exception as e:
|
108 |
+
raise JSONResponse(status_code=500, content={"error": str(e)})
|
109 |
+
else:
|
110 |
+
return JSONResponse(status_code=200, content={"result": 'good'})
|
111 |
+
|
112 |
+
class RAGRequest(BaseModel):
|
113 |
+
question: str
|
114 |
+
|
115 |
+
@app.post("/rag")
|
116 |
+
async def rag(request: RAGRequest):
|
117 |
+
question = request.question
|
118 |
+
if not question:
|
119 |
+
return JSONResponse(status_code=400, content={"message": "Please a question!"})
|
120 |
+
try:
|
121 |
+
search_res = milvus_client.search(
|
122 |
+
collection_name=collection_name,
|
123 |
+
data=[
|
124 |
+
document_to_embeddings(question)
|
125 |
+
],
|
126 |
+
limit=5, # Return top 3 results
|
127 |
+
# search_params={"metric_type": "COSINE"}, # Inner product distance
|
128 |
+
output_fields=["content"], # Return the text field
|
129 |
+
)
|
130 |
+
retrieved_lines_with_distances = [
|
131 |
+
(res["entity"]["content"]) for res in search_res[0]
|
132 |
+
]
|
133 |
+
return JSONResponse(status_code=200, content={"result": retrieved_lines_with_distances})
|
134 |
+
except Exception as e:
|
135 |
+
return JSONResponse(status_code=400, content={"error": str(e)})
|
app/milvus_singleton.py
ADDED
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from pymilvus import Milvus, connections
|
2 |
+
from pymilvus.exceptions import ConnectionConfigException
|
3 |
+
|
4 |
+
class MilvusClientSingleton:
|
5 |
+
_instance = None
|
6 |
+
|
7 |
+
@staticmethod
|
8 |
+
def get_instance(uri):
|
9 |
+
if MilvusClientSingleton._instance is None:
|
10 |
+
MilvusClientSingleton(uri)
|
11 |
+
return MilvusClientSingleton._instance
|
12 |
+
|
13 |
+
def __init__(self, uri):
|
14 |
+
if MilvusClientSingleton._instance is not None:
|
15 |
+
raise Exception("This class is a singleton!")
|
16 |
+
try:
|
17 |
+
# Use the regular Milvus client (not MilvusClient)
|
18 |
+
self._instance = Milvus(uri=uri)
|
19 |
+
print(f"Successfully connected to Milvus at {uri}")
|
20 |
+
except ConnectionConfigException as e:
|
21 |
+
print(f"Error connecting to Milvus: {e}")
|
22 |
+
raise # Re-raise the exception to stop initialization
|
23 |
+
|
24 |
+
def __getattr__(self, name):
|
25 |
+
return getattr(self._instance, name)
|
app/requirements.txt
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
fastapi
|
2 |
+
uvicorn
|
3 |
+
pypdf
|
4 |
+
python-multipart
|
5 |
+
langchain
|
6 |
+
sentence-transformers
|
7 |
+
torch
|
8 |
+
pymilvu
|