Skip to main content

RLAlgorithm

A curated reinforcement-learning algorithm the FreqAI ReinforcementLearner can train with. The list is server-owned (ADR-0066) — the frontend derives its options from rlAlgorithmCatalogue and never hardcodes them. Power users can also supply a free-text model type outside this list via RLConfigInput.allowAdvancedModel.

type RLAlgorithm {
  id: String!
  label: String!
  family: String!
  supportsShort: Boolean!
  default: Boolean!
}

Fields​

RLAlgorithm.id ● String! non-null scalar​

Exact stable-baselines3 / sb3-contrib class name used as rl_config.model_type (e.g. "PPO").

RLAlgorithm.label ● String! non-null scalar​

Human-readable display name.

RLAlgorithm.family ● String! non-null scalar​

Library the algorithm ships in: "sb3" (stable-baselines3) or "sb3-contrib".

RLAlgorithm.supportsShort ● Boolean! non-null scalar​

Whether the algorithm supports short positions out of the box.

RLAlgorithm.default ● Boolean! non-null scalar​

Whether this is the recommended default algorithm (exactly one: PPO).

Returned By​

rlAlgorithmCatalogue query