Strategy JSON format - Miramarket Docs

Strategy JSON is the portable external representation of a Miramarket strategy. It is designed to be readable by humans and deterministic enough for AI agents to generate safely.

Top-level object

{
  "version": "1.1.0",
  "strategy_id": "example_strategy",
  "initial_principal_usd": 100,
  "root_node_id": "condition_entry",
  "nodes": [],
  "edges": [],
  "decision_groups": []
}

Field	Type	Required	Meaning
`version`	string	Yes	Strategy JSON version.
`strategy_id`	string	Yes	Stable strategy identifier.
`initial_principal_usd`	number	Yes	Starting capital in USD. Must be greater than `0`.
`root_node_id`	string	Yes	`node_id` of the root node.
`nodes`	array	Yes	Condition and action nodes.
`edges`	array	Yes	Action-to-condition continuation edges.
`decision_groups`	array	Yes	Condition-to-action selection groups.
`is_demo`	boolean	No	When `true`, skips the fee-aware principal check.

Nodes

Each node is either a CONDITION or an ACTION.

{
  "node_id": "condition_entry",
  "type": "CONDITION",
  "level": 0,
  "condition": {}
}

{
  "node_id": "buy_yes",
  "type": "ACTION",
  "level": 0,
  "action": {}
}

Field	Type	Required	Meaning
`node_id`	string	Yes	Unique node identifier.
`type`	string	Yes	`CONDITION` or `ACTION`.
`level`	number	Yes	Graph level. Must be `0` or greater.
`condition`	object	For condition nodes	Decision logic.
`action`	object	For action nodes	Capital movement or execution instruction.

Condition object

{
  "watch": {
    "venue": "POLYMARKET",
    "market_id": "btc-above-100k-may-2026",
    "outcome_token": "YES",
    "token_id": "1234567890"
  },
  "signal": "IMPLIED_PROBABILITY",
  "trigger": "CROSS_ABOVE",
  "threshold": 0.6
}

Field	Type	Required	Meaning
`watch`	object	Yes	Market or selector to observe.
`signal`	string	Yes	Signal type.
`trigger`	string	Yes	`CROSS_ABOVE` or `CROSS_BELOW`.
`threshold`	number	Yes	Trigger threshold.
`timing`	object	For time signals	Timing details.
`reference`	object	No	Reference mode for return or PnL signals.
`stabilizers`	object	No	Hysteresis or cooldown settings.
`conjuncts`	array	No	Additional `AND` clauses.

Action object

{
  "action_type": "BUY_MARKET_OUTCOME",
  "allocation_pct": 100,
  "target": {
    "venue": "POLYMARKET",
    "market": {
      "market_id": "btc-above-100k-may-2026",
      "outcome_token": "YES",
      "token_id": "1234567890"
    }
  }
}

Field	Type	Required	Meaning
`action_type`	string	Yes	Action to perform.
`allocation_pct`	number	Required in `SPLIT_100` groups	Percentage of selected capital.
`target`	object	Yes	Market, wallet, or vault target.
`risk_guards`	object	No	Execution guardrails.

Edges

Edges represent continuation after an action.

{
  "edge_id": "edge_1",
  "from_node_id": "buy_yes",
  "to_node_id": "condition_exit"
}

Edges must always be action-to-condition.

Decision groups

Decision groups represent condition-to-action selection.

{
  "group_id": "entry_group",
  "condition_node_id": "condition_entry",
  "action_node_ids": ["buy_yes", "hold_cash"],
  "level": 0,
  "mode": "SPLIT_100"
}

Field	Type	Required	Meaning
`group_id`	string	Yes	Unique decision group identifier.
`condition_node_id`	string	Yes	Condition node that selects actions.
`action_node_ids`	array	Yes	Actions selected by the condition.
`level`	number	Yes	Same level as the condition and selected actions.
`mode`	string	Yes	`SINGLE` or `SPLIT_100`.
`priority_order`	array	No	Ordered action IDs for priority-aware selection.

Relationship model

Use decision_groups for condition-to-action relationships. Use edges only for action-to-condition continuation.

condition_entry --decision group--> buy_yes
buy_yes --edge--> condition_exit
condition_exit --decision group--> sell_yes

This separation makes the graph unambiguous for humans and AI agents.

​Top-level object

​Nodes

​Condition object

​Action object

​Edges

​Decision groups

​Relationship model