File size: 3,717 Bytes
4beb7b0
 
 
 
d020550
 
4beb7b0
 
 
 
 
 
 
d020550
 
4beb7b0
 
 
 
d020550
 
4beb7b0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c066163
4beb7b0
 
 
 
 
 
 
 
 
 
 
d020550
 
c1c1c06
4beb7b0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
186fef7
4beb7b0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
cfb1119
d020550
4beb7b0
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
from fastapi import FastAPI, UploadFile, File
from fastapi.middleware.cors import CORSMiddleware
# Loading
import os
from os import makedirs,getcwd
from os.path import join,exists,dirname
from datasets import load_dataset
import torch
from tqdm import tqdm
from sentence_transformers import SentenceTransformer
import uuid
from qdrant_client import models, QdrantClient
from itertools import islice

app = FastAPI()


FILEPATH_PATTERN = "structured_data_doc.parquet"
NUM_PROC = os.cpu_count()
parent_path  = dirname(getcwd())

temp_path  = join(parent_path,'temp')
if not exists(temp_path ):
    makedirs(temp_path )

# Determine device based on GPU availability
device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Using device: {device}")
# Load the desired model
model = SentenceTransformer(
          'sentence-transformers/all-MiniLM-L6-v2',
          device=device
)

# Create function to upsert embeddings in batches
def batched(iterable, n):
    iterator = iter(iterable)
    while batch := list(islice(iterator, n)):
        yield batch

batch_size = 100
# Create an in-memory Qdrant instance
client2 = QdrantClient(path="database")

# Create a Qdrant collection for the embeddings
client2.create_collection(
    collection_name="law",
    vectors_config=models.VectorParams(
        size=model.get_sentence_embedding_dimension(),
        distance=models.Distance.COSINE,
    ),
)






# Create function to generate embeddings (in batches) for a given dataset split
def generate_embeddings(dataset, batch_size=32):
    embeddings = []

    with tqdm(total=len(dataset), desc=f"Generating embeddings for dataset") as pbar:
        for i in range(0, len(dataset), batch_size):
            batch_sentences = dataset['content'][i:i+batch_size]
            batch_embeddings = model.encode(batch_sentences)
            embeddings.extend(batch_embeddings)
            pbar.update(len(batch_sentences))

    return embeddings
    
@app.post("/uploadfile/")
async def create_upload_file(file: UploadFile = File(...)):
    # Here you can save the file and do other operations as needed
    full_dataset = load_dataset("parquet",
                data_files=FILEPATH_PATTERN,
                split="train",
                cache_dir=temp_path,
                keep_in_memory=True,
                num_proc=NUM_PROC*2)
    # Generate and append embeddings to the train split
    law_embeddings = generate_embeddings(full_dataset)
    full_dataset= full_dataset.add_column("embeddings", law_embeddings)
    
    if not 'uuid' in full_dataset.column_names:
      full_dataset = full_dataset.add_column('uuid', [str(uuid.uuid4()) for _ in range(len(full_dataset))])
    # Upsert the embeddings in batches
    for batch in batched(full_dataset, batch_size):
        ids = [point.pop("uuid") for point in batch]
        vectors = [point.pop("embeddings") for point in batch]
    
        client2.upsert(
            collection_name="law",
            points=models.Batch(
                ids=ids,
                vectors=vectors,
                payloads=batch,
            ),
        )
    return {"filename": file.filename, "message": "Done"}

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

@app.get("/search")
def search(prompt: str):
    # Let's see what senators are saying about immigration policy
    hits = client2.search(
        collection_name="law",
        query_vector=model.encode(prompt).tolist(),
        limit=5
    )
    for hit in hits:
      print(hit.payload, "score:", hit.score)
    return hits
    
@app.get("/")
def api_home():
    return {'detail': 'Welcome to FastAPI Qdrant importer!'}