def main():
milvus = Milvus()
# Connect to Milvus server
# You may need to change _HOST and _PORT accordingly
param = {'host': _HOST, 'port': _PORT}
status = milvus.connect(**param)
if status.OK():
print("Server connected.")
else:
print("Server connect fail.")
sys.exit(1)
# Create collection demo_collection if it dosen't exist.
collection_name = 'example_collection'
status, ok = milvus.has_collection(collection_name)
if not ok:
param = {
'collection_name': collection_name,
'dimension': _DIM,
'index_file_size': _INDEX_FILE_SIZE, # optional
'metric_type': MetricType.L2 # optional
}
milvus.create_collection(param)
# Show collections in Milvus server
_, collections = milvus.show_collections()
# present collection info
_, info = milvus.collection_info(collection_name)
print(info)
# Describe demo_collection
_, collection = milvus.describe_collection(collection_name)
print(collection)
# 10000 vectors with 16 dimension
# element per dimension is float32 type
# vectors should be a 2-D array
vectors = [[random.random() for _ in range(_DIM)] for _ in range(10000)]
# You can also use numpy to generate random vectors:
# `vectors = np.random.rand(10000, 16).astype(np.float32)`
# Insert vectors into demo_collection, return status and vectors id list
status, ids = milvus.insert(collection_name=collection_name, records=vectors)
# Flush collection inserted data to disk.
milvus.flush([collection_name])
# Get demo_collection row count
status, result = milvus.count_collection(collection_name)
# create index of vectors, search more rapidly
index_param = {
'nlist': 2048
}
# Create ivflat index in demo_collection
# You can search vectors without creating index. however, Creating index help to
# search faster
print("Creating index: {}".format(index_param))
status = milvus.create_index(collection_name, IndexType.IVF_FLAT, index_param)
# describe index, get information of index
status, index = milvus.describe_index(collection_name)
print(index)
# Use the top 10 vectors for similarity search
query_vectors = vectors[0:10]
# execute vector similarity search
search_param = {
"nprobe": 16
}
param = {
'collection_name': collection_name,
'query_records': query_vectors,
'top_k': 1,
'params': search_param
}
print("Searching ... ")
status, results = milvus.search(**param)
if status.OK():
# indicate search result
# also use by:
# `results.distance_array[0][0] == 0.0 or results.id_array[0][0] == ids[0]`
if results[0][0].distance == 0.0 or results[0][0].id == ids[0]:
print('Query result is correct')
else:
print('Query result isn\'t correct')
# print results
print(results)
# Delete demo_collection
status = milvus.drop_collection(collection_name)
# Disconnect from Milvus
status = milvus.disconnect()
'dimension': _DIM,
'index_file_size': 10, # optional
'metric_type': MetricType.L2 # optional
}
client.create_collection(param)
# randomly generate 100000 vectors and insert collection
vectors = [[random.random() for _ in range(_DIM)] for _ in range(10000)]
client.insert(collection_name, vectors)
# flush data to disk
client.flush([collection_name])
# query collection's statistical information.
status, info = client.collection_info(collection_name)
if not status.OK():
print("Query collection statistical information fail. exiting ....")
sys.exit(1)
# show collection information
print("Total amount of vectors in collection {} is {}".format(
collection_name, info.count))
for par in info.partitions_stat:
print("\tpartition tag: {}, vector count: {}".format(
par.tag, par.count))
# show segment information
for seg in par.segments_stat:
print(
"\t\tsegment name: {}, vector count: {}, index: {}, storage size {:.3f} MB"
.format(seg.segment_name, seg.count, seg.index_name,