def create_pipeline(name, processors=None, encoder=None, description=None):
try:
p = MongoIns.search_by_name(PIPELINE_COLLECTION_NAME, name)
if p:
raise ExistError(f"pipeline <{name}> already exists", "")
pro = []
encoder_res = {}
processor_res = {}
for processor in processors:
pr = operator_detail(processor["name"])
processor_res["operator"] = pr.to_dict()
processor_res["instance"] = pr.inspect_instance(
processor["instance"])
pro.append(processor_res)
encoder_info = operator_detail(encoder["name"])
encoder_res["operator"] = encoder_info.to_dict()
encoder_res["instance"] = encoder_info.inspect_instance(
encoder["instance"])
pipe = Pipeline(name, description, pro, encoder_res)
pipe.metadata = pipe._metadata()
if pipeline_illegal(pipe):
raise PipelineIllegalError("Pipeline illegal check error", "")
MongoIns.insert_documents(PIPELINE_COLLECTION_NAME, pipe.to_dict())
return pipe
except Exception as e:
logger.error(e, exc_info=True)
raise e
def fetch_operators(url, overwrite=True):
"""fetch operators from origin market
url -- origin url
overwrite -- Whether to overwrite local information if the same name exists
"""
origin = []
try:
r = requests.get(url)
if r.headers.get(MARKET_IDENTITY_HEADER) != "0.1.0":
raise RequestError("Uncertified market", "")
if r.status_code != 200:
raise RequestError(r.text, r.status_code)
except Exception as e:
raise RequestError(e.args[0], e)
for op in r.json():
origin.append(Operator(op['name'], op['addr'], op['author'],
op['version'], op['type'], op['description']))
local_operators = all_operators()
local_operator_names = [x.name for x in local_operators]
for x in origin:
if x.name not in local_operator_names:
local_operators.append(x)
else:
if overwrite:
for lop in local_operators:
if lop.name == x.name:
local_operators.remove(lop)
local_operators.append(x)
MongoIns.delete_mongo_collection(OPERATOR_COLLECTION_NAME)
for x in local_operators:
MongoIns.insert_documents(OPERATOR_COLLECTION_NAME, x.to_dict())
return local_operators
def new_application(app_name, fields, s3_bucket):
ok, message = fields_check(fields)
if not ok:
raise ArgsCheckError(message, "")
try:
# check application exist
if MongoIns.search_by_name(APPLICATION_COLLECTION_NAME, app_name):
raise ExistError(f"application <{app_name}> had exist", "")
except ExistError:
raise
try:
for _, value in fields.items():
if value.get("type") == "pipeline":
pipe = MongoIns.search_by_name(PIPELINE_COLLECTION_NAME,
value.get("value"))[0]
ei = identity(
pipe.get("encoder").get("instance").get("endpoint"))
name = f"{app_name}_{pipe.get('encoder').get('instance').get('name').replace('phantoscope_', '')}"
MilvusIns.new_milvus_collection(name, int(ei["dimension"]),
1024, "l2")
# create a application entity collection
MongoIns.new_mongo_collection(f"{app_name}_entity")
S3Ins.new_s3_buckets(s3_bucket)
# create milvus collections
app = Application(name=app_name, fields=fields, bucket=s3_bucket)
app.metadata = app._metadata()
MongoIns.insert_documents(APPLICATION_COLLECTION_NAME, app.to_dict())
return app
except Exception as e:
logger.error("error happen during create app: %s",
str(e),
exc_info=True)
raise e
def register_operators(name, addr, author, version, type, description):
try:
op = Operator(name, addr, author, version, type, description)
op.metadata = op._metadata()
if MongoIns.search_by_name(OPERATOR_COLLECTION_NAME, name):
raise ExistError(f"operator {name} had exist", "")
MongoIns.insert_documents(OPERATOR_COLLECTION_NAME, op.to_dict())
return op.to_dict()
except Exception as e:
logger.error(f"Unexpected error happen during register operator, {str(e)}", exc_info=True)
raise 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