async def chat(request_data: TextData,
current_user: User = Depends(auth.get_current_user)):
"""
Fetches a bot response for a given text/query.
It is basically used to test the chat functionality of the agent
"""
model = AgentProcessor.get_agent(current_user.get_bot())
response = await model.handle_text(request_data.data,
sender_id=current_user.get_user())
return {"data": {"response": response}}
def start_training(bot: str, user: str, token: str = None, reload=True):
"""
prevents training of the bot,
if the training session is in progress otherwise start training
:param reload: whether to reload model in the cache
:param bot: bot id
:param token: JWT token for remote model reload
:param user: user id
:return: model path
"""
exception = None
model_file = None
training_status = None
if Utility.environment.get('model') and Utility.environment['model'][
'train'].get('event_url'):
Utility.train_model_event(bot, user, token)
else:
try:
model_file = train_model_for_bot(bot)
training_status = MODEL_TRAINING_STATUS.DONE.value
agent_url = Utility.environment['model']['train'].get('agent_url')
if agent_url:
Utility.http_request(
'get', urljoin(agent_url, "/api/bot/model/reload"), token,
user)
else:
if reload:
AgentProcessor.reload(bot)
except Exception as e:
logging.exception(e)
training_status = MODEL_TRAINING_STATUS.FAIL.value
exception = str(e)
finally:
ModelProcessor.set_training_status(
bot=bot,
user=user,
status=training_status,
model_path=model_file,
exception=exception,
)
return model_file
async def predict_intent(request_data: TextData,
current_user: User = Depends(auth.get_current_user)):
"""
Fetches the predicted intent of the entered text form the loaded agent
"""
model = AgentProcessor.get_agent(current_user.get_bot())
response = await model.parse_message_using_nlu_interpreter(
request_data.data)
intent = response.get("intent").get("name") if response else None
confidence = response.get("intent").get("confidence") if response else None
return {"data": {"intent": intent, "confidence": confidence}}
def start_training(bot: str, user: str, reload=True):
"""
prevents training of the bot,
if the training session is in progress otherwise start training
:param reload: whether to reload model in the cache
:param bot: bot id
:param user: user id
:return: model path
"""
exception = None
model_file = None
training_status = None
ModelProcessor.set_training_status(
bot=bot, user=user, status=MODEL_TRAINING_STATUS.INPROGRESS.value,
)
try:
model_file = train_model_for_bot(bot)
training_status = MODEL_TRAINING_STATUS.DONE.value
except Exception as e:
logging.exception(e)
training_status = MODEL_TRAINING_STATUS.FAIL.value
exception = str(e)
raise AppException(exception)
finally:
ModelProcessor.set_training_status(
bot=bot,
user=user,
status=training_status,
model_path=model_file,
exception=exception,
)
if reload:
AgentProcessor.reload(bot)
return model_file
async def chat(
request_data: TextData, current_user: User = Depends(auth.get_current_user)
):
"""
Fetches a bot response for a given text/query.
It is basically used to test the chat functionality of the agent
"""
if Utility.environment.get('model') and Utility.environment['model']['train'].get('agent_url'):
agent_url = Utility.environment['model']['train'].get('agent_url')
token = auth.create_access_token(data={"sub": current_user.email})
response = Utility.http_request('post', urljoin(agent_url, "/api/bot/chat"), token.decode('utf8'), current_user.get_user(), json={'data': request_data.data})
else:
model = AgentProcessor.get_agent(current_user.get_bot())
response = await model.handle_text(
request_data.data, sender_id=current_user.get_user()
)
response = {"data": {"response": response}}
return response
async def download_model(
background_tasks: BackgroundTasks,
current_user: User = Depends(auth.get_current_user),
):
"""
Downloads latest trained model file
"""
try:
model_path = AgentProcessor.get_latest_model(current_user.get_bot())
response = FileResponse(
model_path,
filename=os.path.basename(model_path),
background=background_tasks,
)
response.headers[
"Content-Disposition"] = "attachment; filename=" + os.path.basename(
model_path)
return response
except Exception as e:
raise AppException(str(e))