Battle / Arena Nodes

Battle nodes automate the full LenserFight battle lifecycle — from creating a battle definition through executing contenders, collecting votes, aggregating scores, and updating leaderboards.

Node	Type	Output
Battle Create	battle_create	json
Battle Execute	battle_execute	battle_result
Contender Run	contender_run	lens_result
Judge Battle	judge_battle	battle_result
Vote Collector	vote_collector	json
Score Aggregator	score_aggregator	battle_result
Leaderboard Update	leaderboard_update	json

---

Battle Create

Type: battle_create · Category: Battle / Arena

Create a battle definition from a prompt and contender list. Returns a battleId and draft status.

Inputs

Name	Type	Required
json	json	Yes

Required Config

Field	Type	Description
battleId	string	Battle identifier or mapping.

Optional Config

Field	Type	Default
visibility	select	workspace

Example

{ "title": "RAG answer showdown", "visibility": "workspace" }

Expected output: { "battleId": "battle_123", "status": "draft" }

Downstream: → battle_execute

Related Nodes

Battle Execute · Form / Input Trigger

---

Battle Execute

Type: battle_execute · Category: Battle / Arena

Execute contenders and apply a judge strategy for a battle. The central orchestration node for automated AI battles.

Inputs

Name	Type	Required
prompt	text	Yes — battle prompt
contenders	array	Yes — contender definitions

Outputs

Name	Type	Shape
battleResult	battle_result	{ battleId: text, winner: text, scores: json }

Required Config

Field	Type	Description
contenders	json	Contender configs: [{ id, lensId }] or [{ id, agentId }].
judgeStrategy	select	single_judge · panel · vote

Optional Config

Field	Type	Default	Options
promptSource	string	—	Path to the battle prompt.
fundingSource	select	platform_credit	platform_credit · user_byok_cloud · user_byok_local
resultVisibility	select	workspace	private · workspace · public

Example

{
  "contenders": [
    { "id": "concise", "lensId": "lens_concise_answer" },
    { "id": "sourced", "lensId": "lens_sourced_answer" }
  ],
  "promptSource": "$.prompt",
  "judgeStrategy": "panel",
  "fundingSource": "platform_credit",
  "resultVisibility": "workspace",
  "retry": { "attempts": 2, "backoffMs": 2000 }
}

Scenario: Run two RAG-answer contenders and produce a judged winner.

Expected output: { "battleId": "battle_123", "winner": "sourced", "scores": { "sourced": 94, "concise": 78 } }

Downstream: → leaderboard_update with { "battleId": "$.battleId", "winner": "$.winner" }

Valid Connections

→ leaderboard_update, score_aggregator, slack_notify, judge_battle

Related Nodes

Battle Create · Judge Battle · Judge / Eval · Vote Collector

---

Contender Run

Type: contender_run · Category: Battle / Arena

Run one battle contender against a prompt. Use when you want fine-grained control over individual contender execution before aggregating.

Required Config

Field	Type	Description
contenderId	string	Contender identifier.

Example

{
  "contenderId": "sourced",
  "lensId": "lens_sourced_answer",
  "model_id": "openai:gpt-4.1-mini"
}

Expected output: { "contenderId": "sourced", "output": "Answer with citations..." }

Downstream: → judge_battle

---

Judge Battle

Type: judge_battle · Category: Battle / Arena

Judge contender outputs and emit a battle result with winner and scores.

Required Config

Field	Type	Description
rubric	string	Evaluation rubric.

Example

{
  "rubric": "Prefer correctness and source coverage.",
  "judgeModel": "anthropic:claude-3-7-sonnet"
}

Expected output: { "winner": "sourced", "scores": { "sourced": 94, "concise": 78 } }

Downstream: → score_aggregator

Related Nodes

Battle Execute · Vote Collector · Judge / Eval

---

Vote Collector

Type: vote_collector · Category: Battle / Arena

Collect human or automated votes for a battle and emit aggregated vote data.

Required Config

Field	Type	Description
battleId	string	Battle id or mapping (e.g. $.battleId).

Optional Config

Field	Type	Description
closeAfterVotes	number	Automatically close voting after N votes.

Example

{ "battleId": "$.battleId", "closeAfterVotes": 7 }

Expected output: { "votes": [{ "voter": "u1", "contenderId": "sourced" }] }

Downstream: → score_aggregator

---

Score Aggregator

Type: score_aggregator · Category: Battle / Arena

Aggregate judge scores and votes into final rankings using a weighted strategy.

Required Config

Field	Type	Description
aggregation	string	Aggregation strategy (e.g. weighted_average).

Example

{
  "aggregation": "weighted_average",
  "weights": { "judge": 0.8, "vote": 0.2 }
}

Expected output: { "winner": "sourced", "finalScore": 91.6 }

Downstream: → leaderboard_update

Related Nodes

Judge Battle · Vote Collector · Leaderboard Update

---

Leaderboard Update

Type: leaderboard_update · Category: Battle / Arena

Write battle results to a leaderboard. The terminal node of most battle automation pipelines.

Inputs

Name	Type	Required
battleResult	battle_result	Yes

Required Config

Field	Type	Description
leaderboardId	string	Leaderboard identifier.

Optional Config

Field	Type	Default	Options
scorePath	string	$.finalScore	—
visibility	select	workspace	private · workspace · public

Example

{
  "leaderboardId": "arena_weekly",
  "scorePath": "$.finalScore",
  "visibility": "workspace"
}

Expected output: { "updated": true, "rank": 1 }

Downstream: → slack_notify

Valid Connections

→ slack_notify, discord_notify, email_send, logger

Invalid Connections

✗ Cannot receive text or json — requires a battle_result typed input from battle_execute, judge_battle, or score_aggregator.

Related Nodes

Score Aggregator · Battle Execute · Slack Notify

---

See also: Node Catalog Index · AI Primitive Nodes · Workflow Studio · Battle concepts