-
-
Notifications
You must be signed in to change notification settings - Fork 3.8k
Creating a new agent
Please note: Huginn's API is evolving and at some point Agents will likely be extracted into Ruby Gems. We'd very much like your input into how this should work and what should be changed in this API.
Huginn's Agents can create and receive events, and can be scheduled to run code at certain times or intervals. Creating a new Huginn Agent is not difficult, you simply create a new subclass of Agent which defines a set of required methods.
Use the description
class method to set a Markdown description for your agent. For example:
description <<-MD
The WeatherAgent creates an event for the following day's weather at `zipcode`.
You must setup an API key for Wunderground in order to use this Agent.
MD
Agents are configured with a JSON structure from the user, accessible raw via options
and available with Formatting Events using Liquid interpolation performed in interpolated
. You should define a method called default_options
that returns an example default configuration for your type of Agent. Additionally, you should define a method called validate_options
that performs Rails validation on the contents of options
, if any fields are required. You generally don't want to look at interpolated
within validate_options
. Here's an example of default_options
:
def default_options
{ 'zipcode' => '94103' }
end
def validate_options
errors.add(:base, 'zipcode is required') unless options['zipcode'].present?
end
Agents can be scheduled to run at certain times, or on certain intervals. When a schedule is triggered, the check
method in your Agent will be called. If your agent should be schedulable, use the default_schedule
class method to declare a default, otherwise you should call cannot_be_scheduled!
. Possible schedules are: every_1m
, every_2m
, every_5m
, every_10m
, every_30m
, every_1h
, every_2h
, every_5h
, every_12h
, every_1d
, every_2d
, every_7d
, midnight
, 1am
, 2am
, 3am
, 4am
, 5am
, 6am
, 7am
, 8am
, 9am
, 10am
, 11am
, noon
, 1pm
, 2pm
, 3pm
, 4pm
, 5pm
, 6pm
, 7pm
, 8pm
, 9pm
, 10pm
, and 11pm
default_schedule "8pm"
def check
wunderground.forecast_for(interpolated['zipcode'])['forecast']['simpleforecast']['forecastday'].each do |day|
if is_tomorrow?(day)
create_event :payload => day.merge('zipcode' => interpolated['zipcode'])
end
end
end
If your Agent creates events, as this example from the WeatherAgent does, then you should use the event_description
class method to detail what data those events contain.
event_description <<-MD
Events look like this:
{
'zipcode' => 12345,
...
'maxhumidity' => 93,
'minhumidity' => 63
}
MD
If your Agent can receive events, define a method called receive
that accepts an array of incoming events. Otherwise, please annotate it with a call to cannot_receive_events!
.
In code, your Agent can create events with create_event :payload => { ... }
. If your Agent will never create events, please annotate it with a call to cannot_create_events!
.
Agents have memory that can be used to maintain state between scheduled intervals or received events. It will be loaded and saved automatically for you and is available in memory
.
Your Agent should create AgentLogs when interesting things happen, especially errors. Call log
or error
with a log message and, optionally, :outbound_event
or :inbound_event
to keep track of events tied to the log message.
It's nice to be able to tell the user if their instance of your Agent is working correctly. You should define a method called working?
that returns true when everything seems good. Here's an example for an Agent that primarily creates events and has an expected_update_period_in_days
option:
def working?
event_created_within?(interpolated['expected_update_period_in_days']) && !recent_error_logs?
end
And here is an example for an Agent that primarily receives events and has an expected_receive_period_in_days
option:
def working?
last_receive_at && last_receive_at > interpolated['expected_receive_period_in_days'].to_i.days.ago && !recent_error_logs?
end
You can, of course, write Agent-specific code in working?
.
Agents can have a custom UI by defining a show view at: app/views/agents/agent_views/<agent name>/_show.html.erb
. If they need to have server-side functionality, you may POST data to the handle_details_post_agent_path
and handle it with handle_details_post
in your Agent. See the ManualEventAgent and its details view for an example.
Your Agent can receive web requests by implementing receive_web_request
. The URL for your Agent will be something like http://yourserver.com/users/:user_id/web_requests/:agent_id/:secret
where :user_id
is a User's id, :agent_id
is an Agent's id, and :secret
is a token that should be user-specifiable in your Agent's configuration and checked by receive_web_request
. It is highly recommended that every Agent verify this token whenever receive_web_request
is called. For example, one of your Agent's options could be secret
and you could compare this value to params[:secret]
whenever receive_web_request
is called on your Agent, rejecting invalid requests.
Your Agent's receive_web_request
method should return an Array containing a response, a status code, and an optional MIME type. For example:
[{ status: "success" }, 200]
or
["not found", 404, 'text/plain']
Here is an example implementation of receive_web_request
:
def receive_web_request(params, method, format)
secret = params.delete('secret')
return ["Please use POST requests only", 401] unless method == "post"
return ["Not Authorized", 401] unless secret == interpolated['secret']
# do something with params here
['Done!', 200, 'text/plain']
end
Please see the WebRequestsController for more documentation, as well as the implementations of receive_web_request
in WebhookAgent and DataOutputAgent.