slavakurilyak
commented
5 years ago Closed ghost closed 5 years ago
slavakurilyak
commented
5 years ago 
ghost
commented
5 years ago thank you, I no longer get that error. However was getting this since I don't have the google cloud json.
google.auth.exceptions.DefaultCredentialsError
google.auth.exceptions.DefaultCredentialsError: File /app/Kryptos-Staging-3deb8860d570.json was not found.
Traceback (most recent call last)
File "/app/autoapp.py", line 4, in <module>
from app.app import create_app
File "/app/app/app.py", line 14, in <module>
from app.settings import DockerDevConfig, ProdConfig
File "/app/app/settings.py", line 20, in <module>
class Config(object):
File "/app/app/settings.py", line 41, in Config
CLIENT_ACCESS_TOKEN = get_from_datastore("CLIENT_ACCESS_TOKEN", "production")
File "/app/app/settings.py", line 11, in get_from_datastore
ds = datastore.Client()
File "/usr/local/lib/python3.6/site-packages/google/cloud/datastore/client.py", line 205, in __init__
project=project, credentials=credentials, _http=_http)
File "/usr/local/lib/python3.6/site-packages/google/cloud/client.py", line 215, in __init__
_ClientProjectMixin.__init__(self, project=project)
File "/usr/local/lib/python3.6/site-packages/google/cloud/client.py", line 169, in __init__
project = self._determine_default(project)
File "/usr/local/lib/python3.6/site-packages/google/cloud/datastore/client.py", line 222, in _determine_default
return _determine_default_project(project)
File "/usr/local/lib/python3.6/site-packages/google/cloud/datastore/client.py", line 75, in _determine_default_project
project = _base_default_project(project=project)
File "/usr/local/lib/python3.6/site-packages/google/cloud/_helpers.py", line 179, in _determine_default_project
_, project = google.auth.default()
File "/usr/local/lib/python3.6/site-packages/google/auth/_default.py", line 294, in default
credentials, project_id = checker()
File "/usr/local/lib/python3.6/site-packages/google/auth/_default.py", line 165, in _get_explicit_environ_credentials
Open an interactive python shell in this frameos.environ[environment_vars.CREDENTIALS])
File "/usr/local/lib/python3.6/site-packages/google/auth/_default.py", line 89, in _load_credentials_from_file
'File {} was not found.'.format(filename))
google.auth.exceptions.DefaultCredentialsError: File /app/Kryptos-Staging-3deb8860d570.json was not found.I commented out the get_from_datastore parts on settings.py, could see the main page but then when I was registering I was getting this
sqlalchemy.exc.ProgrammingError
sqlalchemy.exc.ProgrammingError: (psycopg2.ProgrammingError) relation "users" does not exist
LINE 2: FROM users
^
[SQL: 'SELECT users.is_active AS users_is_active, users.id AS users_id, users.email AS users_email, users.email_confirmed_at AS users_email_confirmed_at, users.password AS users_password, users.telegram_id AS users_telegram_id, users.telegram_username AS users_telegram_username, users.telegram_photo AS users_telegram_photo, users.telegram_auth_date AS users_telegram_auth_date \nFROM users \nWHERE users.email ILIKE %(email_1)s \n LIMIT %(param_1)s'] [parameters: {'email_1': 'canercak@gmail.com', 'param_1': 1}] (Background on this error at: http://sqlalche.me/e/f405) then I created postgresql://localhost/kryptos manually to get this error.
sqlalchemy.exc.OperationalError: (psycopg2.OperationalError) could not connect to server: Connection refused
Is the server running on host "localhost" (127.0.0.1) and accepting
TCP/IP connections on port 5432?then I changed postgres config from localhost to accept all connections, tried some more, etc but could not make it work.
slavakurilyak
commented
5 years ago @canercak did you try the command $ docker-compose up -d after running $ docker-compose build?
We are currently relying on Google Cloud Platform (GCP) for the deployment of this AI agent. We need to move away from GCP and implement a one-click deploy to Heroku. If anyone is up for the challenge, take a look at #39.
ghost
commented
5 years ago I tried. Getting the same result. I think not relying on deployment credentials just to see the system should be a bigger priority. And just want to add my idea because what you've done looks like nice and runs on proper foundations such as zipline/catalyst and I also want to apply alphazero or gym as you do. However I've run my own bot based on technical indicators and used almost all systems such as haas, gunbot, 3commas, cryptohopper, gekko, freqtrade just to lose money as all others do. Anyone who will get into this will understand that ml is the way to go but it's complicated so people need a simple bot who runs certain ml models. For example ml strategies in gekko work well but the bot is complicated, amateur and not-trustable (crashes, cannot buy bid/ask, cannot run multiple pairs with partial amounts, etc). People need something simple that they can control that doesn't rely on 16th century indicators. That's all. Otherwise no one will ever hand his cryptocurrency portfolio to a virtual assistant. I've handed several times and have seen what happens. Going to write my own bot anyway. Wish you good luck.
slavakurilyak
commented
5 years ago I think not relying on deployment credentials just to see the system should be a bigger priority
I agree. We need to move towards Heroku to simplify Kryptos. See: #39
And just want to add my idea because what you've done looks like nice and runs on proper foundations such as zipline/catalyst and I also want to apply alphazero or gym as you do.
Reinforcement learning has a lot of promise within Kryptos. See: #104
However I've run my own bot based on technical indicators and used almost all systems such as haas, gunbot, 3commas, cryptohopper, gekko, freqtrade just to lose money as all others do. Anyone who will get into this will understand that ml is the way to go but it's complicated so people need a simple bot who runs certain ml models. For example ml strategies in gekko work well but the bot is complicated, amateur and not-trustable (crashes, cannot buy bid/ask, cannot run multiple pairs with partial amounts, etc). People need something simple that they can control that doesn't rely on 16th century indicators. That's all. Otherwise no one will ever hand his cryptocurrency portfolio to a virtual assistant. I've handed several times and have seen what happens.
We believe that the only way to survive in the cryptocurrency-investment space is to use AI and machine learning. We began describing our visions for the application of these technologies here: machine learning, deep learning, data science, and reinforcement learning.
Thanks to the work completed by @bukosabino, we already use gradient tree boosting algorithms (XGBoost and LiteGBM) to perform machine learning strategies at Kryptos.
@slavakurilyak thank you for your work.
When I try to run the app by
docker-compose upI getImportError: cannot import name 'choices'error. Output is below. Do you have an idea why this is happening?Terminal Output
Flask output