Contender Run

Overview

The Contender Run node executes a single AI lens as one side of a battle, feeding it the rendered battle prompt and capturing the provider's output as text. It requires an AI provider to be wired by the workflow engine; if the provider is absent (e.g. in dry-run or test mode) it falls back to echoing the resolved prompt. The node attaches battle context metadata — battleId, contenderId, slot, and execution timing — to its output for downstream nodes such as judge_battle or score_aggregator. Provider errors propagate unconditionally so the engine's retry or error_catch layer can handle them.

Configuration

Field	Type	Required	Description
lensId	string (UUID)	Yes	The lens that generates this contender's response. Must be a published lens UUID.
slot	enum: 'A'	'B'	Yes
battleId	string (UUID)	No	The parent battle ID. Attached to output metadata so downstream nodes (judge_battle, leaderboard_update) can correlate results without re-wiring.
contenderId	string (UUID)	No	The contender record ID. Attached to output metadata for audit trails and replay events.
maxTokens	number	No	Output token cap for this contender. Falls back to the lens default when omitted. Shorter limits reduce cost but may truncate complex answers.
temperature	number (0–2)	No	Sampling temperature override. Overrides the lens default when set. For fair comparisons, use the same value on both contender nodes.
personalityNote	string	No	Appended as a system-prompt prefix to the provider call. Use to give a contender a persona without modifying the underlying lens.

Inputs

Port	Type	Description
prompt	text	The rendered battle prompt the contender must respond to. Typically wired from a prompt_template or battle_execute node. [[label]] placeholders must be resolved by the engine before this node runs.

Outputs

Port	Type	Description
output	text	The contender's generated response text. In fallback mode (no provider), contains the raw resolved prompt.
metadata	json	Execution context attached by the runner: battleId, contenderId, slot, durationMs, token counts, provider finish_reason, and executedBy marker. Consumed by judge_battle and score_aggregator.

Example

{
  "nodeType": "contender_run",
  "config": {
    "lensId": "d4e5f6a7-b8c9-0123-def0-123456789abc",
    "slot": "A",
    "battleId": "{{n1.battle.id}}",
    "maxTokens": 2048,
    "temperature": 0.7
  }
}