def upload(name, **kwargs):
try:
app = application_detail(name)
if not app:
raise NotExistError("application not exist", "application %s not exist" % name)
bucket_name = app.buckets.split(",")[0]
accept_fields = [x for x, y in app.fields.items() if y.get('type') != "object"]
pipeline_fields = {x: y['pipeline'] for x, y in app.fields.items() if y.get('type') == "object"}
new_fields = app.fields.copy()
for k, v in kwargs.items():
if k in accept_fields:
new_fields[k]['value'] = v
res = []
for k, _ in kwargs.get('fields').items():
if k not in accept_fields and k not in pipeline_fields:
raise RequestError(f"fields {k} not in application", "")
for n, p in pipeline_fields.items():
pipe = pipeline_detail(p)
if not pipe:
raise NotExistError("pipeline not exist", "pipeline %s not exist" % p)
value = kwargs['fields'].get(n)
file_data = value.get('data')
url = value.get('url')
if not file_data and not url:
raise RequestError("can't find data or url from request", "")
file_name = "{}-{}".format(name, uuid.uuid4().hex)
file_path = save_tmp_file(file_name, file_data, url)
# begin to timing
start = time.time()
S3Ins.upload2bucket(bucket_name, file_path, file_name)
upload_time = time.time()
logger.debug("[timing] upload image to bucket costs: {:.3f}s".format(upload_time - start))
vectors = run_pipeline(pipe, data=file_data, url=url)
pipeline_time = time.time()
logger.debug("[timing] run pipeline costs: {:.3f}s".format(pipeline_time - upload_time))
milvus_collection_name = f"{pipe.name}_{pipe.encoder}"
vids = MilvusIns.insert_vectors(milvus_collection_name, vectors)
insert_time = time.time()
logger.debug("[timing] insert to milvus costs: {:.3f}s".format(insert_time - pipeline_time))
for vid in vids:
m = DB(id=vid, app_name=name,
image_url=gen_url(bucket_name, file_name),
fields=new_fields)
add_mapping_data(m)
res.append(new_mapping_ins(id=vid, app_name=name,
image_url=gen_url(bucket_name, file_name),
fields=new_fields))
final_time = time.time()
logger.debug("[timing] prepare result costs: {:.3f}s".format(final_time - insert_time))
return res
except Exception as e:
print(e)
return e
def upload(name, **kwargs):
try:
app = application_detail(name)
if not app:
raise NotExistError("application not exist",
"application %s not exist" % name)
bucket_name = app.buckets.split(",")[0]
accept_fields = [
x for x, y in app.fields.items() if y.get('type') != "object"
]
pipeline_fields = {
x: y['pipeline']
for x, y in app.fields.items() if y.get('type') == "object"
}
new_fields = app.fields.copy()
for k, v in kwargs.items():
if k in accept_fields:
new_fields[k]['value'] = v
res = []
for k, _ in kwargs.get('fields').items():
if k not in accept_fields and k not in pipeline_fields:
raise RequestError(f"fields {k} not in application", "")
for n, p in pipeline_fields.items():
pipe = pipeline_detail(p)
if not pipe:
raise NotExistError("pipeline not exist",
"pipeline %s not exist" % p)
value = kwargs['fields'].get(n)
file_data = value.get('data')
url = value.get('url')
if not file_data and not url:
raise RequestError("can't find data or url from request", "")
file_name = "{}-{}".format(name, uuid.uuid4().hex)
file_path = save_tmp_file(file_name, file_data, url)
S3Ins.upload2bucket(bucket_name, file_path, file_name)
vectors = run_pipeline(pipe, data=file_data, url=url)
if not vectors:
raise NoneVectorError(
"can't encode data by encoder, check input or encoder", "")
milvus_collection_name = f"{pipe.name}_{pipe.encoder}"
vids = MilvusIns.insert_vectors(milvus_collection_name, vectors)
for vid in vids:
m = DB(id=vid,
app_name=name,
image_url=gen_url(bucket_name, file_name),
fields=new_fields)
add_mapping_data(m)
res.append(
new_mapping_ins(id=vid,
app_name=name,
image_url=gen_url(bucket_name, file_name),
fields=new_fields))
return res
except Exception as e:
print(e)
return e
def upload(name, **kwargs):
try:
app = application_detail(name)
if not app:
raise NotExistError("application not exist", "application %s not exist" % name)
bucket_name = app.buckets.split(",")[0]
accept_fields = [x for x, y in app.fields.items() if y.get('type') != "pipeline"]
pipeline_fields = {x: y['value'] for x, y in app.fields.items() if y.get('type') == "pipeline"}
new_fields = app.fields.copy()
for k, v in kwargs.items():
if k in accept_fields:
new_fields[k]['value'] = v
res = []
for k, _ in kwargs.get('fields').items():
if k not in accept_fields and k not in pipeline_fields:
raise RequestError(f"fields {k} not in application", "")
docs = {}
valid_field_flag = False
for n, p in pipeline_fields.items():
pipe = pipeline_detail(p)
if not pipe:
raise NotExistError("pipeline not exist", "pipeline %s not exist" % p)
value = kwargs['fields'].get(n)
if not value:
continue
valid_field_flag = True
file_data = value.get('data')
url = value.get('url')
if not file_data and not url:
raise RequestError("can't find data or url from request", "")
file_name = "{}-{}".format(name, uuid.uuid4().hex)
file_path = save_tmp_file(file_name, file_data, url)
S3Ins.upload2bucket(bucket_name, file_path, file_name)
vectors = run_pipeline(pipe, data=file_data, url=url)
if not vectors:
raise NoneVectorError("can't encode data by encoder, check input or encoder", "")
milvus_collection_name = f"{app.name}_{pipe.encoder['name']}_{pipe.encoder['instance']}"
vids = MilvusIns.insert_vectors(milvus_collection_name, vectors)
docs[n] = {"ids": vids, "url": gen_url(bucket_name, file_name)}
doc_id = MongoIns.insert_documents(f"{app.name}_entity", docs)
res.append(new_mapping_ins(docs))
if not valid_field_flag:
raise RequestError("none valid field exist", "")
return res
except Exception as e:
err_msg = f"Unexpected error happen when upload: {str(e)}"
logger.error(err_msg, exc_info=True)
raise UnexpectedError(err_msg, e)
def upload(name, **kwargs):
try:
app = application_detail(name)
if not app:
raise NotExistError("application not exist", "application %s not exist" % name)
bucket_name = app.buckets.split(",")[0]
accept_fields = [x for x, y in app.fields.items() if y.get('type') != "pipeline"]
pipeline_fields = {x: y['value'] for x, y in app.fields.items() if y.get('type') == "pipeline"}
new_fields = app.fields.copy()
for k, v in kwargs.items():
if k in accept_fields:
new_fields[k]['value'] = v
res = []
for k, _ in kwargs.get('fields').items():
if k not in accept_fields and k not in pipeline_fields:
raise RequestError(f"fields {k} not in application", "")
docs = {}
for n, p in pipeline_fields.items():
pipe = pipeline_detail(p)
if not pipe:
raise NotExistError("pipeline not exist", "pipeline %s not exist" % p)
value = kwargs['fields'].get(n)
file_data = value.get('data')
url = value.get('url')
if not file_data and not url:
raise RequestError("can't find data or url from request", "")
file_name = "{}-{}".format(name, uuid.uuid4().hex)
file_path = save_tmp_file(file_name, file_data, url)
S3Ins.upload2bucket(bucket_name, file_path, file_name)
vectors = run_pipeline(pipe, data=file_data, url=url)
milvus_collection_name = f"{app.name}_{pipe.encoder['name']}_{pipe.encoder['instance']}"
vids = MilvusIns.insert_vectors(milvus_collection_name, vectors)
docs[n] = {"ids": vids, "url": gen_url(bucket_name, file_name)}
doc_id = MongoIns.insert_documents(f"{app.name}_entity", docs)
res.append(new_mapping_ins(docs))
fields=new_fields))
return res
def test_insert_vectors(self):
"""test insert vectors"""
vectors = [[random.random() for _ in range(self.dimension)]
for _ in range(20)]
rv = MilvusIns.insert_vectors(self.name, vectors)
assert len(rv) == 20