Yandex Alice integration

Any model specified by a DeepPavlov config can be launched as a skill for Yandex.Alice. You can do it using command line interface or using python.

Command line interface

To interact with Alice you will require your own HTTPS certificate. To generate a new one – run:

openssl req -new -newkey rsa:4096 -days 365 -nodes -x509 -subj "/CN=MY_DOMAIN_OR_IP" -keyout my.key -out my.crt

Then run:

python -m deeppavlov riseapi --api-mode alice --https --key my.key --cert my.crt  <config_path> [-d] [-p <port>]
  • -d: download model specific data before starting the service.

  • -p <port>: sets the port to <port>. Overrides default settings from deeppavlov/utils/settings/server_config.json.

Now set up and test your dialog ( Detailed documentation of the platform could be found on Advanced API configuration is described in REST API section.


To run a model specified by a DeepPavlov config <config_path> as an Alice skill, firstly, you have to turn it to a Skill and then make it an Agent.

from deeppavlov import build_model
from deeppavlov.skills.default_skill.default_skill import DefaultStatelessSkill
from deeppavlov.agents.default_agent.default_agent import DefaultAgent
from deeppavlov.agents.processors.highest_confidence_selector import HighestConfidenceSelector
from deeppavlov.utils.alice import start_agent_server

model = build_model("<config_path>", download=True)

# Step 1: make it a Skill
skill = DefaultStatelessSkill(model)
# Step 2: make it an Agent
agent = DefaultAgent(skills=[skill])
# Step 3: run server
start_agent_server(agent, host='', port=7051, endpoint='/agent', ssl_key='my.key', ssl_cert='my.crt')

If your model is already a subclass of Skill or a subclass of Agent (see skills and agents) you can skip corresponding steps.