feat: backend model load balancing support (#4927)
This commit is contained in:
@@ -54,4 +54,4 @@ from .explore import (
|
||||
from .tag import tags
|
||||
|
||||
# Import workspace controllers
|
||||
from .workspace import account, members, model_providers, models, tool_providers, workspace
|
||||
from .workspace import account, load_balancing_config, members, model_providers, models, tool_providers, workspace
|
||||
|
@@ -1,14 +1,19 @@
|
||||
from flask_login import current_user
|
||||
from flask_restful import Resource
|
||||
|
||||
from libs.login import login_required
|
||||
from services.feature_service import FeatureService
|
||||
|
||||
from . import api
|
||||
from .wraps import cloud_utm_record
|
||||
from .setup import setup_required
|
||||
from .wraps import account_initialization_required, cloud_utm_record
|
||||
|
||||
|
||||
class FeatureApi(Resource):
|
||||
|
||||
@setup_required
|
||||
@login_required
|
||||
@account_initialization_required
|
||||
@cloud_utm_record
|
||||
def get(self):
|
||||
return FeatureService.get_features(current_user.current_tenant_id).dict()
|
||||
|
@@ -17,13 +17,19 @@ class VersionApi(Resource):
|
||||
args = parser.parse_args()
|
||||
check_update_url = current_app.config['CHECK_UPDATE_URL']
|
||||
|
||||
if not check_update_url:
|
||||
return {
|
||||
'version': '0.0.0',
|
||||
'release_date': '',
|
||||
'release_notes': '',
|
||||
'can_auto_update': False
|
||||
result = {
|
||||
'version': current_app.config['CURRENT_VERSION'],
|
||||
'release_date': '',
|
||||
'release_notes': '',
|
||||
'can_auto_update': False,
|
||||
'features': {
|
||||
'can_replace_logo': current_app.config['CAN_REPLACE_LOGO'],
|
||||
'model_load_balancing_enabled': current_app.config['MODEL_LB_ENABLED']
|
||||
}
|
||||
}
|
||||
|
||||
if not check_update_url:
|
||||
return result
|
||||
|
||||
try:
|
||||
response = requests.get(check_update_url, {
|
||||
@@ -31,20 +37,15 @@ class VersionApi(Resource):
|
||||
})
|
||||
except Exception as error:
|
||||
logging.warning("Check update version error: {}.".format(str(error)))
|
||||
return {
|
||||
'version': args.get('current_version'),
|
||||
'release_date': '',
|
||||
'release_notes': '',
|
||||
'can_auto_update': False
|
||||
}
|
||||
result['version'] = args.get('current_version')
|
||||
return result
|
||||
|
||||
content = json.loads(response.content)
|
||||
return {
|
||||
'version': content['version'],
|
||||
'release_date': content['releaseDate'],
|
||||
'release_notes': content['releaseNotes'],
|
||||
'can_auto_update': content['canAutoUpdate']
|
||||
}
|
||||
result['version'] = content['version']
|
||||
result['release_date'] = content['releaseDate']
|
||||
result['release_notes'] = content['releaseNotes']
|
||||
result['can_auto_update'] = content['canAutoUpdate']
|
||||
return result
|
||||
|
||||
|
||||
api.add_resource(VersionApi, '/version')
|
||||
|
106
api/controllers/console/workspace/load_balancing_config.py
Normal file
106
api/controllers/console/workspace/load_balancing_config.py
Normal file
@@ -0,0 +1,106 @@
|
||||
from flask_restful import Resource, reqparse
|
||||
from werkzeug.exceptions import Forbidden
|
||||
|
||||
from controllers.console import api
|
||||
from controllers.console.setup import setup_required
|
||||
from controllers.console.wraps import account_initialization_required
|
||||
from core.model_runtime.entities.model_entities import ModelType
|
||||
from core.model_runtime.errors.validate import CredentialsValidateFailedError
|
||||
from libs.login import current_user, login_required
|
||||
from models.account import TenantAccountRole
|
||||
from services.model_load_balancing_service import ModelLoadBalancingService
|
||||
|
||||
|
||||
class LoadBalancingCredentialsValidateApi(Resource):
|
||||
@setup_required
|
||||
@login_required
|
||||
@account_initialization_required
|
||||
def post(self, provider: str):
|
||||
if not TenantAccountRole.is_privileged_role(current_user.current_tenant.current_role):
|
||||
raise Forbidden()
|
||||
|
||||
tenant_id = current_user.current_tenant_id
|
||||
|
||||
parser = reqparse.RequestParser()
|
||||
parser.add_argument('model', type=str, required=True, nullable=False, location='json')
|
||||
parser.add_argument('model_type', type=str, required=True, nullable=False,
|
||||
choices=[mt.value for mt in ModelType], location='json')
|
||||
parser.add_argument('credentials', type=dict, required=True, nullable=False, location='json')
|
||||
args = parser.parse_args()
|
||||
|
||||
# validate model load balancing credentials
|
||||
model_load_balancing_service = ModelLoadBalancingService()
|
||||
|
||||
result = True
|
||||
error = None
|
||||
|
||||
try:
|
||||
model_load_balancing_service.validate_load_balancing_credentials(
|
||||
tenant_id=tenant_id,
|
||||
provider=provider,
|
||||
model=args['model'],
|
||||
model_type=args['model_type'],
|
||||
credentials=args['credentials']
|
||||
)
|
||||
except CredentialsValidateFailedError as ex:
|
||||
result = False
|
||||
error = str(ex)
|
||||
|
||||
response = {'result': 'success' if result else 'error'}
|
||||
|
||||
if not result:
|
||||
response['error'] = error
|
||||
|
||||
return response
|
||||
|
||||
|
||||
class LoadBalancingConfigCredentialsValidateApi(Resource):
|
||||
@setup_required
|
||||
@login_required
|
||||
@account_initialization_required
|
||||
def post(self, provider: str, config_id: str):
|
||||
if not TenantAccountRole.is_privileged_role(current_user.current_tenant.current_role):
|
||||
raise Forbidden()
|
||||
|
||||
tenant_id = current_user.current_tenant_id
|
||||
|
||||
parser = reqparse.RequestParser()
|
||||
parser.add_argument('model', type=str, required=True, nullable=False, location='json')
|
||||
parser.add_argument('model_type', type=str, required=True, nullable=False,
|
||||
choices=[mt.value for mt in ModelType], location='json')
|
||||
parser.add_argument('credentials', type=dict, required=True, nullable=False, location='json')
|
||||
args = parser.parse_args()
|
||||
|
||||
# validate model load balancing config credentials
|
||||
model_load_balancing_service = ModelLoadBalancingService()
|
||||
|
||||
result = True
|
||||
error = None
|
||||
|
||||
try:
|
||||
model_load_balancing_service.validate_load_balancing_credentials(
|
||||
tenant_id=tenant_id,
|
||||
provider=provider,
|
||||
model=args['model'],
|
||||
model_type=args['model_type'],
|
||||
credentials=args['credentials'],
|
||||
config_id=config_id,
|
||||
)
|
||||
except CredentialsValidateFailedError as ex:
|
||||
result = False
|
||||
error = str(ex)
|
||||
|
||||
response = {'result': 'success' if result else 'error'}
|
||||
|
||||
if not result:
|
||||
response['error'] = error
|
||||
|
||||
return response
|
||||
|
||||
|
||||
# Load Balancing Config
|
||||
api.add_resource(LoadBalancingCredentialsValidateApi,
|
||||
'/workspaces/current/model-providers/<string:provider>/models/load-balancing-configs/credentials-validate')
|
||||
|
||||
api.add_resource(LoadBalancingConfigCredentialsValidateApi,
|
||||
'/workspaces/current/model-providers/<string:provider>/models/load-balancing-configs/<string:config_id>/credentials-validate')
|
@@ -12,6 +12,7 @@ from core.model_runtime.errors.validate import CredentialsValidateFailedError
|
||||
from core.model_runtime.utils.encoders import jsonable_encoder
|
||||
from libs.login import login_required
|
||||
from models.account import TenantAccountRole
|
||||
from services.model_load_balancing_service import ModelLoadBalancingService
|
||||
from services.model_provider_service import ModelProviderService
|
||||
|
||||
|
||||
@@ -104,21 +105,56 @@ class ModelProviderModelApi(Resource):
|
||||
parser.add_argument('model', type=str, required=True, nullable=False, location='json')
|
||||
parser.add_argument('model_type', type=str, required=True, nullable=False,
|
||||
choices=[mt.value for mt in ModelType], location='json')
|
||||
parser.add_argument('credentials', type=dict, required=True, nullable=False, location='json')
|
||||
parser.add_argument('credentials', type=dict, required=False, nullable=True, location='json')
|
||||
parser.add_argument('load_balancing', type=dict, required=False, nullable=True, location='json')
|
||||
parser.add_argument('config_from', type=str, required=False, nullable=True, location='json')
|
||||
args = parser.parse_args()
|
||||
|
||||
model_provider_service = ModelProviderService()
|
||||
model_load_balancing_service = ModelLoadBalancingService()
|
||||
|
||||
try:
|
||||
model_provider_service.save_model_credentials(
|
||||
if ('load_balancing' in args and args['load_balancing'] and
|
||||
'enabled' in args['load_balancing'] and args['load_balancing']['enabled']):
|
||||
if 'configs' not in args['load_balancing']:
|
||||
raise ValueError('invalid load balancing configs')
|
||||
|
||||
# save load balancing configs
|
||||
model_load_balancing_service.update_load_balancing_configs(
|
||||
tenant_id=tenant_id,
|
||||
provider=provider,
|
||||
model=args['model'],
|
||||
model_type=args['model_type'],
|
||||
credentials=args['credentials']
|
||||
configs=args['load_balancing']['configs']
|
||||
)
|
||||
except CredentialsValidateFailedError as ex:
|
||||
raise ValueError(str(ex))
|
||||
|
||||
# enable load balancing
|
||||
model_load_balancing_service.enable_model_load_balancing(
|
||||
tenant_id=tenant_id,
|
||||
provider=provider,
|
||||
model=args['model'],
|
||||
model_type=args['model_type']
|
||||
)
|
||||
else:
|
||||
# disable load balancing
|
||||
model_load_balancing_service.disable_model_load_balancing(
|
||||
tenant_id=tenant_id,
|
||||
provider=provider,
|
||||
model=args['model'],
|
||||
model_type=args['model_type']
|
||||
)
|
||||
|
||||
if args.get('config_from', '') != 'predefined-model':
|
||||
model_provider_service = ModelProviderService()
|
||||
|
||||
try:
|
||||
model_provider_service.save_model_credentials(
|
||||
tenant_id=tenant_id,
|
||||
provider=provider,
|
||||
model=args['model'],
|
||||
model_type=args['model_type'],
|
||||
credentials=args['credentials']
|
||||
)
|
||||
except CredentialsValidateFailedError as ex:
|
||||
raise ValueError(str(ex))
|
||||
|
||||
return {'result': 'success'}, 200
|
||||
|
||||
@@ -170,11 +206,73 @@ class ModelProviderModelCredentialApi(Resource):
|
||||
model=args['model']
|
||||
)
|
||||
|
||||
model_load_balancing_service = ModelLoadBalancingService()
|
||||
is_load_balancing_enabled, load_balancing_configs = model_load_balancing_service.get_load_balancing_configs(
|
||||
tenant_id=tenant_id,
|
||||
provider=provider,
|
||||
model=args['model'],
|
||||
model_type=args['model_type']
|
||||
)
|
||||
|
||||
return {
|
||||
"credentials": credentials
|
||||
"credentials": credentials,
|
||||
"load_balancing": {
|
||||
"enabled": is_load_balancing_enabled,
|
||||
"configs": load_balancing_configs
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
class ModelProviderModelEnableApi(Resource):
|
||||
|
||||
@setup_required
|
||||
@login_required
|
||||
@account_initialization_required
|
||||
def patch(self, provider: str):
|
||||
tenant_id = current_user.current_tenant_id
|
||||
|
||||
parser = reqparse.RequestParser()
|
||||
parser.add_argument('model', type=str, required=True, nullable=False, location='json')
|
||||
parser.add_argument('model_type', type=str, required=True, nullable=False,
|
||||
choices=[mt.value for mt in ModelType], location='json')
|
||||
args = parser.parse_args()
|
||||
|
||||
model_provider_service = ModelProviderService()
|
||||
model_provider_service.enable_model(
|
||||
tenant_id=tenant_id,
|
||||
provider=provider,
|
||||
model=args['model'],
|
||||
model_type=args['model_type']
|
||||
)
|
||||
|
||||
return {'result': 'success'}
|
||||
|
||||
|
||||
class ModelProviderModelDisableApi(Resource):
|
||||
|
||||
@setup_required
|
||||
@login_required
|
||||
@account_initialization_required
|
||||
def patch(self, provider: str):
|
||||
tenant_id = current_user.current_tenant_id
|
||||
|
||||
parser = reqparse.RequestParser()
|
||||
parser.add_argument('model', type=str, required=True, nullable=False, location='json')
|
||||
parser.add_argument('model_type', type=str, required=True, nullable=False,
|
||||
choices=[mt.value for mt in ModelType], location='json')
|
||||
args = parser.parse_args()
|
||||
|
||||
model_provider_service = ModelProviderService()
|
||||
model_provider_service.disable_model(
|
||||
tenant_id=tenant_id,
|
||||
provider=provider,
|
||||
model=args['model'],
|
||||
model_type=args['model_type']
|
||||
)
|
||||
|
||||
return {'result': 'success'}
|
||||
|
||||
|
||||
class ModelProviderModelValidateApi(Resource):
|
||||
|
||||
@setup_required
|
||||
@@ -259,6 +357,10 @@ class ModelProviderAvailableModelApi(Resource):
|
||||
|
||||
|
||||
api.add_resource(ModelProviderModelApi, '/workspaces/current/model-providers/<string:provider>/models')
|
||||
api.add_resource(ModelProviderModelEnableApi, '/workspaces/current/model-providers/<string:provider>/models/enable',
|
||||
endpoint='model-provider-model-enable')
|
||||
api.add_resource(ModelProviderModelDisableApi, '/workspaces/current/model-providers/<string:provider>/models/disable',
|
||||
endpoint='model-provider-model-disable')
|
||||
api.add_resource(ModelProviderModelCredentialApi,
|
||||
'/workspaces/current/model-providers/<string:provider>/models/credentials')
|
||||
api.add_resource(ModelProviderModelValidateApi,
|
||||
|
Reference in New Issue
Block a user