Multi-armed bandit pick API

Selects the next arm to pull via UCB1 or epsilon-greedy given each arm's pull and reward stats, returning per-arm means and exploration scores. Answers 'Which option should I try next?', 'How do I balance explore vs exploit?'.

Price$0.03per request

MethodPOST

Route/v1/workflow/bandit-pick

StatusLive

MIME typeapplication/json

Rate limit120/minute

Cache0s public

banditmulti-armeducb1epsilon-greedyexplore-exploitoptimizationselectionworkflow

API URLhttps://x402.hexl.dev/v1/workflow/bandit-pick

Integration docs

Example request

{
  "arms": [
    {
      "name": "A",
      "pulls": 10,
      "reward": 6
    },
    {
      "name": "B",
      "pulls": 3,
      "reward": 2
    }
  ],
  "strategy": "ucb1"
}

Example response

{
  "chosenArm": "B",
  "strategy": "ucb1",
  "armStats": [
    {
      "name": "A",
      "pulls": 10,
      "mean": 0.6
    },
    {
      "name": "B",
      "pulls": 3,
      "mean": 0.666667
    }
  ],
  "scores": [
    {
      "name": "A",
      "score": 1.316233
    },
    {
      "name": "B",
      "score": 1.974323
    }
  ],
  "interpretation": "Strategy ucb1 selected arm 'B'."
}

Input schema

{
  "type": "object",
  "required": [
    "arms"
  ],
  "properties": {
    "arms": {
      "type": "array",
      "examples": [
        [
          {
            "name": "A",
            "pulls": 10,
            "reward": 6
          },
          {
            "name": "B",
            "pulls": 3,
            "reward": 2
          }
        ]
      ]
    },
    "strategy": {
      "type": "string",
      "enum": [
        "epsilon-greedy",
        "ucb1"
      ]
    },
    "epsilon": {
      "type": "number"
    },
    "seed": {
      "description": "Seed for epsilon-greedy explore draw"
    }
  }
}

Output schema

{
  "type": "object",
  "additionalProperties": true
}