Configuration Module

Hummingbot's config module helps onboard users who are new to running a trading bot. It is helpful in - Compiling a list of absolutely essential config variables. - Skip or provide default variables that are more advanced / have the potential to confuse new users. - Collecting and validating user inputs as config values. - Check if all configs are present before running any strategy.

Architecture

Currently, we split all the configuration variables into three different types.

Config type In-memory location Saved to local yml Description
in_memory hummingbot/client/config/in_memory_config_map.py No Configs that are never saved and prompted every time (currently, only the strategy and strategy_config_path are in this config map.
global hummingbot/client/config/global_config_map.py Yes Strategy-agnostic configs such as exchange API keys, wallet selection, etc.
strategy hummingbot/strategy/{STRATEGY_NAME}/{STRATEGY_NAME}_strategy_config_map.py Yes Strategy-specific configs.

Default Configuration Flow

  1. When the bot starts, it automatically reads all the global configurations from a file named conf_global.yml. If such a file does not exist, it will copy the empty template from hummingbot/templates/conf_global_TEMPLATE.yml. The bot populates global_config_map object in hummingbot/client/config/global_config_map.py with any values previously saved in the yml file.
  2. When the user enters config command, the bot prompts the user all the items in in_memory_config_map.
  3. Once a user inputs her desired strategy, she can choose to import or create a configuration file.
    • if import is chosen, the user will be prompted to select a strategy config file. The bot will load all variables values and save them in-memory.
    • if create is chosen, the bot will copy a strategy config template from hummingbot/templates/conf_{STRATEGY_NAME}_strategy_{COUNT}.yml. The user will then fill out each of the configs required by that specific strategy.
  4. With each user input, the bot will validate the input with a custom checker (more details on that in the ConfigVar class). If an input is invalid, the user is prompted the same question again.
  5. After all the strategy configs are filled out, the bot uses the newly acquired info to figure out which variables are require in global config settings (Which exchange API keys to prompt, etc).
  6. Since altering certain variables will trigger requirement for other variables, this prompt-and-config process loops until all variables are complete.
  7. When the loop finishes. The bot writes all of the config variables saved in memory to local yml files so that they can be reused for another session.
  8. The user can then start running her selected strategy with the set of configs currently stored in memory.

ConfigVar Class

The ConfigVar Class is located in hummingbot/client/config/config_var.py. It standardizes each config setting with a set of attributes.

Attribute Attribute Type Use
key str Unique key that identifies a config variable.
prompt str or callable Question displayed in the client when the bot collects user input for this config setting. You can modify the string in run time by passing a function rather than a static string.
is_secure bool Whether the user input needs to be masked with "***".
default any Default value for this variable if user input is None.
type_str str One of {"str", "list", "dict", "float", "int", "bool"}. Defaults to "str". This is used by parse_cvar_value to parse user input into correct data type.
required_if callable A condition check for whether this config setting needs to be prompted during the configuration flow.
validator callable A condition check for whether an input is a valid value for this config setting.
on_validated callable A function hook that gets activated if an input passes the validation check (e.g. set wallet requirement to True when a valid DEX name is entered.)

Check hummingbot/client/config/config_var.py for more details.

Config definition conventions

  1. Always place configs that will alters requirement state first. Example: telegram_token should only be required if telegram_enabled is set to True. Therefore telegram_enabled should be listed before telegram_token.
  2. For exchange-specific configurations, use using_exchange("exchange_name") as the required_if condition.
  3. When writing prompt questions, be sure to add examples for a better user experience.
  4. When prompting for a boolean value, add (y/n) as options so that the user knows what to enter.
  5. When prompting a question with a few choices as answers e.g. ["import", "create", etc], make sure to include all options in the format of (OPTION_1/OPTION_2/OPTION_3). This pattern is recognized by our autocomplete system, and the user can hit Tab to have the option autofilled.