feat(web,coder): arena pane — compare 2-6 AI competitors on same prompt

Arena is a new pane kind for competitive AI evaluation. A Battle runs
the same prompt against 2-6 Contestants across two concurrent lanes:
local lane (llama-swap models, serial) and cloud lane (parallel).

Added to all three registries: @boocode/contracts WsFrameSchema,
server InferenceFrame, and web WsFrame.

Backend (apps/coder):
- arena-runner: battle scheduler, lane classifier, benchmark, results
  writer, resume, user winner override
- arena-analyzer: two-stage digest→judge analysis on DEFAULT_MODEL
- arena-decisions: status transitions and resume logic (unit-tested)
- arena-analyzer-helpers: pure helper functions (unit-tested)
- arena-model-call: model call utility for analysis
- arena routes: create/get/list/stop/analyze/cross-examine/winner/diff
- schema: battles, contestants, cross_examinations tables (idempotent)
- remove old /api/arena* routes and tasks.arena_id column

Frontend (apps/web):
- ArenaLauncherDialog: battle type, prompt, contestant selection
- ArenaPane: live roster, streaming output, analysis, cross-exam
- DiffView: unified diff with line-by-line color for coding contests
- Winner override per-row dropdown (Trophy icon)
- battle_updated WS handler for live winner/analysis updates
- arena pane kind in Workspace, ChatTabBar, useSidebar

Cross-app:
- ArenaState and ArenaContestantShape/WsFrame types (contracts)
- battle_* frames in WsFrameSchema, InferenceFrame, and web WsFrame
- manifest.json written per battle results folder
- /Arena added to .gitignore

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

packages/contracts/package.json

@@ -28,6 +28,10 @@
     "./worktree-risk": {
       "types": "./dist/worktree-risk.d.ts",
       "default": "./dist/worktree-risk.js"
+    },
+    "./arena": {
+      "types": "./dist/arena.d.ts",
+      "default": "./dist/arena.js"
     }
   },
   "scripts": {

packages/contracts/src/arena.ts

@@ -0,0 +1,55 @@
+/** Arena types — single source of truth for cross-app Arena wire contracts. */
+
+export type BattleType = 'coding' | 'qa';
+export type BattleStatus = 'pending' | 'running' | 'completed' | 'failed' | 'cancelled';
+export type ContestantStatus = 'queued' | 'running' | 'done' | 'error';
+export type ContestantLane = 'local' | 'cloud';
+
+// Pane state — carried on the WorkspacePane row, mirrors OrchestratorState.
+export interface ArenaState {
+  battle_id: string;
+  battle_type: BattleType;
+  prompt: string;
+}
+
+export interface BattleShape {
+  id: string;
+  project_id: string;
+  battle_type: BattleType;
+  prompt: string;
+  status: BattleStatus;
+  winner_contestant_id: string | null;
+  results_path: string | null;
+  error: string | null;
+  created_at: string;
+  updated_at: string;
+}
+
+export interface ContestantShape {
+  id: string;
+  battle_id: string;
+  /** Backend name (coding) or persona name (qa). Unique per (battle, model) pair. */
+  identity: string;
+  model: string;
+  lane: ContestantLane;
+  task_id: string | null;
+  worktree_id: string | null;
+  status: ContestantStatus;
+  duration_ms: number | null;
+  tokens_per_sec: number | null;
+  cost_tokens: number | null;
+  result_path: string | null;
+  error: string | null;
+  created_at: string;
+  updated_at: string;
+}
+
+export interface CrossExaminationShape {
+  id: string;
+  battle_id: string;
+  /** Backend + model performing the cross-examination. */
+  identity: string;
+  model: string;
+  verdict: string | null;
+  created_at: string;
+}

packages/contracts/src/ws-frames.ts

@@ -358,6 +358,53 @@ export const FlowRunStepUpdatedFrame = z.object({
   report: z.string().optional(),
 });
 
+// ---- arena frames ----------------------------------------------------------
+
+const ContestantManifestEntry = z.object({
+  id: Uuid,
+  identity: z.string().min(1),
+  model: z.string().min(1),
+  lane: z.enum(['local', 'cloud']),
+});
+
+// Published once when a battle starts. Carries the contestant roster so the
+// ArenaPane can build its grid immediately.
+export const BattleStartedFrame = z.object({
+  type: z.literal('battle_started'),
+  battle_id: Uuid,
+  battle_type: z.enum(['coding', 'qa']),
+  prompt: z.string(),
+  contestants: z.array(ContestantManifestEntry),
+});
+
+// Published on every contestant status change or streaming update.
+// `delta` carries the latest chunk of streaming output while status='running'.
+// `battle_status` is present only on the final transition that closes the battle.
+export const ContestantUpdatedFrame = z.object({
+  type: z.literal('contestant_updated'),
+  battle_id: Uuid,
+  contestant_id: Uuid,
+  status: z.enum(['queued', 'running', 'done', 'error']).optional(),
+  duration_ms: z.number().int().nonnegative().optional(),
+  tokens_per_sec: z.number().nonnegative().optional(),
+  battle_status: z.enum(['pending', 'running', 'completed', 'failed', 'cancelled']).optional(),
+  delta: z.string().optional(),
+  error: z.string().optional(),
+});
+
+// Published when battle-level state changes that don't ride on a contestant
+// update: analysis finished, winner set, cross-exam verdict ready. The pane
+// uses this to update its analysis panel and winner badge without a refetch.
+// Fields are all optional — publishers include only what changed.
+export const BattleUpdatedFrame = z.object({
+  type: z.literal('battle_updated'),
+  battle_id: Uuid,
+  status: z.enum(['pending', 'running', 'completed', 'failed', 'cancelled']).optional(),
+  winner_contestant_id: Uuid.nullable().optional(),
+  analysis_ready: z.boolean().optional(),
+  cross_exam_id: Uuid.optional(),
+});
+
 // ---- discriminated union ---------------------------------------------------
 
 export const WsFrameSchema = z.discriminatedUnion('type', [
@@ -381,6 +428,10 @@ export const WsFrameSchema = z.discriminatedUnion('type', [
   // orchestrator
   FlowRunStartedFrame,
   FlowRunStepUpdatedFrame,
+  // arena
+  BattleStartedFrame,
+  ContestantUpdatedFrame,
+  BattleUpdatedFrame,
   // per-user
   ChatStatusFrame,
   SessionUpdatedFrame,
@@ -425,6 +476,9 @@ export const KNOWN_FRAME_TYPES: readonly WsFrame['type'][] = [
   'agent_status_updated',
   'flow_run_started',
   'flow_run_step_updated',
+  'battle_started',
+  'contestant_updated',
+  'battle_updated',
   'chat_status',
   'session_updated',
   'session_renamed',