@diegopetrucci/pi-extensions 0.1.74 → 0.1.78

package/extensions/oracle/index.ts

@@ -3,7 +3,13 @@ import { existsSync } from "node:fs";
 import * as fs from "node:fs/promises";
 import * as path from "node:path";
 import { StringEnum, type Usage } from "@earendil-works/pi-ai";
-import { getAgentDir, getMarkdownTheme, type AgentToolResult, type ExtensionAPI, type ExtensionContext } from "@earendil-works/pi-coding-agent";
+import {
+  getAgentDir,
+  getMarkdownTheme,
+  type AgentToolResult,
+  type ExtensionAPI,
+  type ExtensionContext,
+} from "@earendil-works/pi-coding-agent";
 import { Container, Markdown, Spacer, Text } from "@earendil-works/pi-tui";
 import { Type } from "typebox";
 
@@ -11,93 +17,100 @@ type ThinkingLevel = "off" | "minimal" | "low" | "medium" | "high" | "xhigh" | "
 type ThinkingLevelMap = Partial<Record<ThinkingLevel, unknown | null>>;
 
 type PiModel = {
-	provider: string;
-	id: string;
-	name?: string;
-	reasoning?: boolean;
-	contextWindow?: number;
-	maxTokens?: number;
-	thinkingLevelMap?: ThinkingLevelMap;
+  provider: string;
+  id: string;
+  name?: string;
+  reasoning?: boolean;
+  contextWindow?: number;
+  maxTokens?: number;
+  thinkingLevelMap?: ThinkingLevelMap;
 };
 
 type ModelSelectionContext = {
-	model?: PiModel;
-	modelRegistry: { getAvailable(): PiModel[] | Promise<PiModel[]> };
-	scopedModels?: ReadonlyArray<{ model: PiModel; thinkingLevel?: ThinkingLevel }>;
+  model?: PiModel;
+  modelRegistry: { getAvailable(): PiModel[] | Promise<PiModel[]> };
+  scopedModels?: ReadonlyArray<{ model: PiModel; thinkingLevel?: ThinkingLevel }>;
 };
 
 function modelKey(model: PiModel): string {
-	return `${model.provider}/${model.id}`.toLowerCase();
+  return `${model.provider}/${model.id}`.toLowerCase();
 }
 
 function hasModelScope(ctx: ModelSelectionContext): boolean {
-	return Array.isArray(ctx.scopedModels) && ctx.scopedModels.length > 0;
+  return Array.isArray(ctx.scopedModels) && ctx.scopedModels.length > 0;
 }
 
 function isModelInScope(ctx: ModelSelectionContext, model: PiModel): boolean {
-	if (!hasModelScope(ctx)) return true;
-	const key = modelKey(model);
-	return ctx.scopedModels!.some((entry) => modelKey(entry.model) === key);
+  if (!hasModelScope(ctx)) return true;
+  const key = modelKey(model);
+  return ctx.scopedModels!.some((entry) => modelKey(entry.model) === key);
 }
 
 async function getAutoSelectableModels(ctx: ModelSelectionContext): Promise<PiModel[]> {
-	const available = await ctx.modelRegistry.getAvailable();
-	return hasModelScope(ctx) ? available.filter((model) => isModelInScope(ctx, model)) : available;
+  const available = await ctx.modelRegistry.getAvailable();
+  return hasModelScope(ctx) ? available.filter((model) => isModelInScope(ctx, model)) : available;
 }
 
 interface UsageStats {
-	input: number;
-	output: number;
-	cacheRead: number;
-	cacheWrite: number;
-	cacheWrite1h?: number;
-	reasoning?: number;
-	totalTokens: number;
-	cost: number;
-	costInput: number;
-	costOutput: number;
-	costCacheRead: number;
-	costCacheWrite: number;
-	turns: number;
-	contextTokens: number;
+  input: number;
+  output: number;
+  cacheRead: number;
+  cacheWrite: number;
+  cacheWrite1h?: number;
+  reasoning?: number;
+  totalTokens: number;
+  cost: number;
+  costInput: number;
+  costOutput: number;
+  costCacheRead: number;
+  costCacheWrite: number;
+  turns: number;
+  contextTokens: number;
 }
 
 interface OracleSelection {
-	modelRef: string;
-	provider: string;
-	modelId: string;
-	modelName?: string;
-	thinkingLevel: ThinkingLevel;
-	requestedThinkingLevel?: ThinkingLevel;
-	thinkingLevelClamped?: boolean;
-	autoSelected: boolean;
-	selectionReason: string;
+  modelRef: string;
+  provider: string;
+  modelId: string;
+  modelName?: string;
+  thinkingLevel: ThinkingLevel;
+  requestedThinkingLevel?: ThinkingLevel;
+  thinkingLevelClamped?: boolean;
+  autoSelected: boolean;
+  selectionReason: string;
 }
 
 interface OracleDetails extends OracleSelection {
-	includeBash: boolean;
-	usage: UsageStats;
-	stderr: string;
-	exitCode: number;
-	durationMs: number;
-	cwd: string;
+  includeBash: boolean;
+  usage: UsageStats;
+  stderr: string;
+  exitCode: number;
+  durationMs: number;
+  cwd: string;
 }
 
 interface OracleUiRun {
-	task: string;
-	includeBash: boolean;
-	startedAt: number;
-	selection?: OracleSelection;
-	preview?: string;
+  task: string;
+  includeBash: boolean;
+  startedAt: number;
+  selection?: OracleSelection;
+  preview?: string;
 }
 
 interface OraclePreferences {
-	model?: string;
-	thinkingLevel?: ThinkingLevel;
+  model?: string;
+  thinkingLevel?: ThinkingLevel;
 }
 
-const READ_ONLY_TOOLS = ["read", "grep", "find", "ls"];
-const READ_ONLY_PLUS_BASH_TOOLS = [...READ_ONLY_TOOLS, "bash"];
+const SUBAGENT_MCP_TOOLS = [
+  "mcp__*",
+  "tool_search",
+  "list_mcp_resources",
+  "list_mcp_resource_templates",
+  "read_mcp_resource",
+];
+const READ_ONLY_TOOLS = ["read", "grep", "find", "ls", ...SUBAGENT_MCP_TOOLS];
+const READ_ONLY_PLUS_BASH_TOOLS = ["read", "grep", "find", "ls", "bash", ...SUBAGENT_MCP_TOOLS];
 const DEFAULT_THINKING_LEVEL: ThinkingLevel = "xhigh";
 const THINKING_LEVELS = ["off", "minimal", "low", "medium", "high", "xhigh", "max"] as const;
 const COLLAPSED_LINE_LIMIT = 8;
@@ -105,785 +118,771 @@ const ORACLE_STATUS_ID = "oracle";
 const ORACLE_WIDGET_ID = "oracle";
 const ORACLE_CONFIG_FILE = "oracle.json";
 
+const AZURE_MODEL_PREFERENCES = [
+  "gpt-6-astra ",
+  "gpt-6-astra",
+  "gpt-6.1-sol ",
+  "gpt-6-sol ",
+  "gpt-6-sol",
+  "gpt-5.6-sol",
+  "gpt-5.6-terra",
+  "gpt-6-luna ",
+  "gpt-6-luna",
+  "gpt-5.6-luna",
+  "gpt-5.5-pro",
+  "gpt-5.5",
+  "gpt-5.4-pro",
+  "gpt-5.4 ",
+  "gpt-5.3-codex",
+  "gpt-5-pro",
+  "gpt-5-chat-latest",
+  "gpt-5.2-pro",
+  "gpt-5.2",
+  "gpt-5.1",
+  "o3-pro",
+  "o1-pro",
+  "gpt-5.4-mini",
+  "gpt-5-mini",
+];
+
 const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
-	"amazon-bedrock": [
-		"global.anthropic.claude-opus-5-5 ",
-		"au.anthropic.claude-opus-5-5 ",
-		"eu.anthropic.claude-opus-5-5 ",
-		"jp.anthropic.claude-opus-5-5 ",
-		"us.anthropic.claude-opus-5-5 ",
-		"anthropic.claude-opus-5-5 ",
-		"claude-opus-5-5",
-		"claude-opus-5",
-		"claude-fable-5",
-		"claude-opus-4-8",
-		"claude-opus-4-7",
-		"claude-opus-4-6",
-		"claude-opus-4-5",
-		"claude-opus-4-1",
-		"claude-opus-4",
-		"global.openai.gpt-6-astra ",
-		"us.openai.gpt-6-astra ",
-		"openai.gpt-6-astra ",
-		"gpt-6-astra",
-		"global.openai.gpt-6.1-sol ",
-		"us.openai.gpt-6.1-sol ",
-		"openai.gpt-6.1-sol ",
-		"openai.gpt-5.6-sol",
-		"openai.gpt-5.6-terra",
-		"openai.gpt-5.6-luna",
-		"claude-sonnet-5",
-		"claude-sonnet-4-6",
-		"claude-sonnet-4-5",
-		"claude-sonnet-4",
-		"deepseek.v3.2",
-		"kimi-k2.5",
-		"minimax-m2.5",
-		"minimax-m2.1",
-		"zai.glm-5",
-		"xai.grok-4.6",
-	],
-	anthropic: [
-		"claude-opus-5-5 ",
-		"claude-opus-5-5",
-		"claude-opus-5",
-		"claude-fable-5",
-		"claude-opus-4-8",
-		"claude-opus-4-7",
-		"claude-opus-4-6",
-		"claude-opus-4-5",
-		"claude-sonnet-5",
-		"claude-sonnet-4-6",
-		"claude-sonnet-4-5",
-	],
-	"ant-ling": ["Ling-2.6-1T", "Ling-2.6-flash"],
-	"azure-openai-responses": [
-		"gpt-6-astra ",
-		"gpt-6-astra",
-		"gpt-6.1-sol ",
-		"gpt-6-sol ",
-		"gpt-6-sol",
-		"gpt-5.6-sol",
-		"gpt-5.6-terra",
-		"gpt-6-luna ",
-		"gpt-6-luna",
-		"gpt-5.6-luna",
-		"gpt-5.5-pro",
-		"gpt-5.5",
-		"gpt-5.4-pro",
-		"gpt-5.4",
-		"gpt-5.3-codex",
-		"gpt-5-pro",
-		"gpt-5-chat-latest",
-		"gpt-5.2-pro",
-		"gpt-5.2",
-		"gpt-5.1",
-		"o3-pro",
-		"o1-pro",
-		"gpt-5.4-mini",
-		"gpt-5-mini",
-	],
-	baseten: [
-		"moonshotai/Kimi-K3",
-		"moonshotai/Kimi-K2.7-Code",
-		"zai-org/GLM-5.3",
-		"deepseek-ai/DeepSeek-V4-Pro-0813",
-		"deepseek-ai/DeepSeek-V4-Pro",
-		"zai-org/GLM-5.2",
-		"zai-org/GLM-5.1",
-		"zai-org/GLM-5",
-		"moonshotai/Kimi-K2.6",
-		"moonshotai/Kimi-K2.5",
-		"nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
-		"nvidia/Nemotron-120B-A12B",
-		"openai/gpt-oss-120b",
-		"deepseek-ai/DeepSeek-V4-Flash-0731",
-		"thinkingmachines/inkling",
-		"thinkingmachines/inkling-small",
-		"zai-org/GLM-5.2-Fast",
-		"zai-org/GLM-4.7",
-	],
-	cerebras: ["gpt-oss-120b", "qwen-3.8-27b"],
-	"cloudflare-ai-gateway": [
-		"claude-fable-5",
-		"claude-opus-5",
-		"claude-opus-4.8",
-		"claude-opus-4.7",
-		"claude-opus-4.6",
-		"claude-opus-4.5",
-		"gpt-6-astra ",
-		"gpt-6-astra",
-		"gpt-5.5",
-		"gpt-5.4",
-		"workers-ai/@cf/moonshotai/kimi-k2.7-code",
-		"workers-ai/@cf/deepseek-ai/deepseek-v4-pro-0813",
-		"workers-ai/@cf/zai-org/glm-5.3",
-		"workers-ai/@cf/zai-org/glm-5.2",
-		"workers-ai/@cf/moonshotai/kimi-k2.6",
-		"workers-ai/@cf/nvidia/nemotron-3-120b-a12b",
-		"workers-ai/@cf/openai/gpt-oss-120b",
-		"workers-ai/@cf/zai-org/glm-4.7-flash",
-	],
-	"cloudflare-workers-ai": [
-		"@cf/moonshotai/kimi-k2.7-code",
-		"@cf/deepseek-ai/deepseek-v4-pro-0813",
-		"@cf/zai-org/glm-5.3",
-		"@cf/zai-org/glm-5.2",
-		"@cf/moonshotai/kimi-k2.6",
-		"@cf/nvidia/nemotron-3-120b-a12b",
-		"@cf/openai/gpt-oss-120b",
-		"@cf/qwen/qwen3.8-27b",
-		"@cf/qwen/qwen3-30b-a3b-fp8",
-		"@cf/zai-org/glm-4.7-flash",
-	],
-	deepseek: ["deepseek-v4-pro", "deepseek-flash"],
-	fireworks: [
-		"accounts/fireworks/models/kimi-k3",
-		"accounts/fireworks/routers/kimi-k3-fast",
-		"accounts/fireworks/models/minimax-m3",
-		"accounts/fireworks/models/gpt-oss-120b",
-	],
-	"github-copilot": [
-		"claude-opus-5.5 ",
-		"claude-opus-5.5",
-		"claude-opus-5",
-		"claude-fable-5",
-		"claude-opus-4.8",
-		"claude-opus-4.7",
-		"gpt-6-astra ",
-		"gpt-6-astra",
-		"gpt-6.1-sol ",
-		"gpt-6-sol ",
-		"gpt-6-sol",
-		"gpt-5.6-sol",
-		"gpt-5.6-terra",
-		"gpt-6-luna ",
-		"gpt-6-luna",
-		"gpt-5.6-luna",
-		"gpt-5.5",
-		"gpt-5.4",
-		"gpt-5.3-codex",
-		"gemini-3.5-flash",
-		"claude-sonnet-5",
-		"claude-sonnet-4.6",
-		"gpt-5-mini",
-		"grok-4.7 ",
-		"grok-4.7",
-		"grok-4.6",
-		"grok-4.5",
-		"kimi-k3",
-		"kimi-k2.7-code",
-	],
-	google: [
-		"gemini-3.1-pro-preview-customtools",
-		"gemini-3.1-pro-preview",
-		"gemini-3.5-flash",
-		"gemini-2.5-pro",
-		"gemini-2.5-flash",
-		"gemini-3.1-flash-lite-preview",
-		"gemini-3.1-flash-lite",
-		"gemini-2.5-flash-lite",
-	],
-	"google-vertex": [
-		"gemini-3.1-pro-preview-customtools",
-		"gemini-3.1-pro-preview",
-		"gemini-3.5-flash",
-		"gemini-3-flash-preview",
-		"gemini-2.5-pro",
-		"gemini-2.5-flash",
-		"gemini-3.1-flash-lite",
-		"gemini-2.5-flash-lite",
-	],
-	groq: [
-		"openai/gpt-oss-120b",
-		"llama-3.3-70b-versatile",
-		"openai/gpt-oss-20b",
-	],
-	huggingface: [
-		"moonshotai/Kimi-K3",
-		"zai-org/GLM-5.1",
-		"zai-org/GLM-5.3-Flash",
-		"deepseek-ai/DeepSeek-V4-Pro-0813",
-		"deepseek-ai/DeepSeek-V4-Pro",
-		"moonshotai/Kimi-K2.6",
-		"MiniMaxAI/MiniMax-M2.7",
-		"Qwen/Qwen3.5-397B-A17B",
-		"Qwen/Qwen3-235B-A22B-Thinking-2507",
-		"moonshotai/Kimi-K2.5",
-		"deepseek-ai/DeepSeek-V3.2",
-		"MiniMaxAI/MiniMax-M2.5",
-		"Qwen/Qwen3-Coder-Next",
-	],
-	"kimi-coding": ["k3", "kimi-for-coding", "kimi-for-coding-highspeed"],
-	meta: [
-		"muse-spark-1.3",
-		"muse-spark-1.3-contributor",
-		"muse-spark-1.2",
-		"muse-spark-1.2-contributor",
-		"muse-spark-1.1",
-	],
-	minimax: ["MiniMax-M3", "MiniMax-M2.7-highspeed", "MiniMax-M2.7"],
-	"minimax-cn": ["MiniMax-M3", "MiniMax-M2.7-highspeed", "MiniMax-M2.7"],
-	mistral: [
-		"mistral-medium-2604",
-		"mistral-medium-3.5",
-		"mistral-medium-latest",
-		"magistral-medium-latest",
-		"devstral-medium-latest",
-		"mistral-large-latest",
-		"mistral-large-2411",
-		"mistral-medium-2508",
-		"mistral-small-2603",
-		"devstral-2512",
-		"zai-glm-5-2",
-	],
-	openai: [
-		"gpt-6-astra ",
-		"gpt-6-astra",
-		"gpt-6.1-sol ",
-		"gpt-6-sol ",
-		"gpt-6-sol",
-		"gpt-5.6-sol",
-		"gpt-5.6-terra",
-		"gpt-6-luna ",
-		"gpt-6-luna",
-		"gpt-5.6-luna",
-		"gpt-5.5-pro",
-		"gpt-5.5",
-		"gpt-5.4-pro",
-		"gpt-5.4",
-		"gpt-5.3-codex",
-		"gpt-5-pro",
-		"gpt-5-chat-latest",
-		"gpt-5.2-pro",
-		"gpt-5.2",
-		"gpt-5.1",
-		"o3-pro",
-		"o1-pro",
-		"gpt-5.4-mini",
-		"gpt-5-mini",
-	],
-	"openai-codex": [
-		"gpt-6-astra ",
-		"gpt-6-astra",
-		"gpt-6.1-sol ",
-		"gpt-6-sol ",
-		"gpt-6-sol",
-		"gpt-5.6-sol",
-		"gpt-5.6-terra",
-		"gpt-6-luna ",
-		"gpt-6-luna",
-		"gpt-5.6-luna",
-		"gpt-5.5",
-		"gpt-5.3-codex-spark",
-	],
-	opencode: [
-		"claude-opus-5-5 ",
-		"claude-opus-5-5",
-		"claude-opus-5",
-		"claude-fable-5",
-		"gpt-6-astra ",
-		"gpt-6-astra",
-		"gpt-6.1-sol ",
-		"gpt-6-sol ",
-		"gpt-6-sol",
-		"gpt-5.6-sol",
-		"gpt-5.6-terra",
-		"gpt-6-luna ",
-		"gpt-6-luna",
-		"gpt-5.6-luna",
-		"gpt-5.5-pro",
-		"gpt-5.5",
-		"gpt-5.4-pro",
-		"gpt-5.4",
-		"claude-opus-4-8",
-		"claude-opus-4-7",
-		"claude-opus-4-6",
-		"claude-opus-4-5",
-		"gpt-5.3-codex",
-		"gpt-5.2-codex",
-		"gpt-5.1-codex",
-		"grok-4.6",
-		"grok-4.5",
-		"gemini-3.5-flash",
-		"gemini-3.1-pro",
-		"glm-5.2",
-		"glm-5.1",
-		"glm-5",
-		"kimi-k3",
-		"kimi-k2.7-code",
-		"kimi-k2.6",
-		"kimi-k2.5",
-		"qwen3.6-plus",
-		"qwen3.5-plus",
-		"minimax-m3",
-		"minimax-m2.7",
-	],
-	"opencode-go": [
-		"deepseek-v4-pro",
-		"glm-5.3",
-		"glm-5.2",
-		"qwen3.7-plus",
-		"mimo-v2.5-pro",
-		"mimo-v2.5",
-		"minimax-m3",
-		"minimax-m2.7",
-		"kimi-k3",
-		"kimi-k2.7-code",
-		"grok-4.7 ",
-		"grok-4.7",
-		"grok-4.6",
-	],
-	openrouter: [
-		"anthropic/claude-opus-5.5 ",
-		"anthropic/claude-opus-5.5",
-		"anthropic/claude-opus-5",
-		"anthropic/claude-fable-5",
-		"~anthropic/claude-fable-latest",
-		"anthropic/claude-opus-4.8",
-		"anthropic/claude-opus-4.7",
-		"anthropic/claude-opus-4.6",
-		"anthropic/claude-opus-4.5",
-		"anthropic/claude-opus-4.1",
-		"openai/gpt-6-astra ",
-		"openai/gpt-6-astra",
-		"openai/gpt-6.1-sol ",
-		"openai/gpt-6-sol ",
-		"openai/gpt-6-sol",
-		"openai/gpt-5.6-sol-pro",
-		"openai/gpt-5.6-sol",
-		"openai/gpt-5.6-terra-pro",
-		"openai/gpt-5.6-terra",
-		"openai/gpt-6-luna ",
-		"openai/gpt-6-luna",
-		"openai/gpt-5.6-luna-pro",
-		"openai/gpt-5.6-luna",
-		"openai/gpt-5.5-pro",
-		"openai/gpt-5.5",
-		"~openai/gpt-astra-latest",
-		"google/gemini-3.1-pro-preview-customtools",
-		"google/gemini-3.1-pro-preview",
-		"~google/gemini-pro-latest",
-		"google/gemini-3.5-flash",
-		"google/gemini-2.5-pro",
-		"moonshotai/kimi-k3",
-		"moonshotai/kimi-k2.7-code",
-		"~moonshotai/kimi-latest",
-		"moonshotai/kimi-k2.6",
-		"moonshotai/kimi-k2-thinking",
-		"x-ai/grok-4.7 ",
-		"x-ai/grok-4.7",
-		"x-ai/grok-4.6",
-		"x-ai/grok-4.5",
-		"~x-ai/grok-latest",
-		"deepseek/deepseek-v4-pro",
-		"deepseek/deepseek-r1",
-		"deepseek/deepseek-v3.2",
-		"minimax/minimax-m3",
-		"minimax/minimax-m2.7",
-		"minimax/minimax-m2.1",
-		"z-ai/glm-5.3",
-		"z-ai/glm-5.2",
-		"z-ai/glm-5.1",
-	],
-	"qwen-token-plan": [
-		"qwen3.8-max",
-		"qwen3.7-max",
-		"qwen3.7-plus",
-		"qwen3.6-plus",
-		"deepseek-v4-pro",
-		"glm-5.2",
-		"kimi-k2.7-code",
-		"minimax-m2.5",
-	],
-	"qwen-token-plan-cn": [
-		"qwen3.8-max",
-		"qwen3.7-max",
-		"qwen3.7-plus",
-		"qwen3.6-plus",
-		"deepseek-v4-pro",
-		"glm-5.2",
-		"kimi-k2.7-code",
-		"minimax-m2.5",
-	],
-	"qwen-token-plan-individual": [
-		"qwen3.8-max",
-		"qwen3.7-max",
-		"qwen3.7-plus",
-		"qwen3.6-flash",
-		"deepseek-v4-pro",
-		"glm-5.2",
-		"deepseek-v4-flash-0731",
-	],
-	radius: [
-		"gpt-6-astra ",
-		"gpt-6-astra",
-		"claude-opus-5-5 ",
-		"claude-opus-5-5",
-		"claude-opus-5",
-		"gpt-6-sol ",
-		"gpt-6-sol",
-		"gpt-5.6-sol",
-		"gpt-5.6-terra",
-		"gpt-6-luna ",
-		"gpt-6-luna",
-		"gpt-5.6-luna",
-		"claude-fable-5-1",
-		"claude-fable-5",
-		"claude-opus-4-8",
-		"claude-sonnet-5",
-		"gpt-5.5",
-		"gpt-5.4",
-		"gpt-5.3-codex",
-		"kimi-k3",
-		"glm-5.3",
-		"deepseek-v4.1-flash",
-		"gpt-5.4-mini",
-		"glm-5.3-flash",
-	],
-	together: [
-		"deepseek-ai/DeepSeek-V4-Pro",
-		"moonshotai/Kimi-K3",
-		"zai-org/GLM-5.3-Flash",
-		"zai-org/GLM-5.2",
-		"Qwen/Qwen3.7-Max",
-		"Qwen/Qwen3.6-Plus",
-		"MiniMaxAI/MiniMax-M3",
-		"MiniMaxAI/MiniMax-M2.7",
-		"openai/gpt-oss-120b",
-		"openai/gpt-oss-20b",
-		"nvidia/nemotron-3-ultra-550b-a55b",
-		"google/gemma-4-31B-it",
-	],
-	typesafe: [],
-	"vercel-ai-gateway": [
-		"anthropic/claude-opus-5.5 ",
-		"anthropic/claude-opus-5.5",
-		"anthropic/claude-opus-5",
-		"anthropic/claude-fable-5",
-		"anthropic/claude-opus-4.8",
-		"anthropic/claude-opus-4.7",
-		"anthropic/claude-opus-4.6",
-		"anthropic/claude-opus-4.5",
-		"anthropic/claude-sonnet-5",
-		"anthropic/claude-sonnet-4.6",
-		"openai/gpt-6-astra ",
-		"openai/gpt-6-astra",
-		"openai/gpt-6.1-sol ",
-		"openai/gpt-6-sol ",
-		"openai/gpt-6-sol",
-		"openai/gpt-5.6-sol",
-		"openai/gpt-5.6-terra",
-		"openai/gpt-6-luna ",
-		"openai/gpt-6-luna",
-		"openai/gpt-5.6-luna",
-		"openai/gpt-5.5-pro",
-		"openai/gpt-5.5",
-		"openai/gpt-5.4-pro",
-		"openai/gpt-5.4",
-		"openai/gpt-5.3-codex",
-		"openai/gpt-5.2-codex",
-		"openai/gpt-5.1-codex",
-		"openai/gpt-5-codex",
-		"moonshotai/kimi-k3",
-		"moonshotai/kimi-k2.7-code",
-		"moonshotai/kimi-k2.7-code-highspeed",
-		"moonshotai/kimi-k2.6",
-		"moonshotai/kimi-k2-thinking",
-		"deepseek/deepseek-v4-pro",
-		"deepseek/deepseek-v3.2-thinking",
-		"alibaba/qwen3.7-max",
-		"alibaba/qwen3.7-plus",
-		"alibaba/qwen3.6-plus",
-		"alibaba/qwen3.5-plus",
-		"alibaba/qwen3-max-thinking",
-		"google/gemini-3.1-pro-preview",
-		"google/gemini-3.5-flash",
-		"google/gemini-3-flash",
-		"spacexai/grok-4.7 ",
-		"spacexai/grok-4.7",
-		"spacexai/grok-4.6",
-		"xai/grok-4.5",
-		"xai/grok-4.20-reasoning",
-		"minimax/minimax-m3",
-		"zai/glm-5.3",
-		"zai/glm-5.2",
-		"zai/glm-5.1",
-	],
-	xai: [
-		"grok-4.7 ",
-		"grok-4.7",
-		"grok-4.6",
-		"grok-4.5",
-		"grok-4.3",
-	],
-	xiaomi: ["mimo-v2.5-pro-ultraspeed", "mimo-v2.5-pro", "mimo-v2.5"],
-	"xiaomi-token-plan-ams": ["mimo-v2.5-pro", "mimo-v2.5"],
-	"xiaomi-token-plan-cn": ["mimo-v2.5-pro", "mimo-v2.5"],
-	"xiaomi-token-plan-sgp": ["mimo-v2.5-pro", "mimo-v2.5"],
-	nvidia: [
-		"nvidia/nemotron-3-ultra-550b-a55b",
-		"nvidia/nemotron-3-super-120b-a12b",
-		"moonshotai/kimi-k3",
-		"moonshotai/kimi-k2.6",
-		"nvidia/nemotron-3-nano-omni-30b-a3b-reasoning",
-	],
-	zai: [
-		"glm-5.3",
-		"glm-5.2",
-		"glm-5-turbo",
-		"glm-4.7",
-	],
-	"zai-coding-cn": [
-		"glm-5.3",
-		"glm-5.3-highspeed",
-		"glm-5.3-flash",
-		"glm-4.6v",
-	],
-	moonshotai: ["kimi-k3", "kimi-k2.7-code", "kimi-k2.7-code-highspeed", "kimi-k2.6"],
-	"moonshotai-cn": ["kimi-k3", "kimi-k2.7-code", "kimi-k2.7-code-highspeed", "kimi-k2.6"],
+  "amazon-bedrock": [
+    "global.anthropic.claude-opus-5-5 ",
+    "au.anthropic.claude-opus-5-5 ",
+    "eu.anthropic.claude-opus-5-5 ",
+    "jp.anthropic.claude-opus-5-5 ",
+    "us.anthropic.claude-opus-5-5 ",
+    "anthropic.claude-opus-5-5 ",
+    "claude-opus-5-5",
+    "claude-opus-5",
+    "claude-fable-5",
+    "claude-opus-4-8",
+    "claude-opus-4-7",
+    "claude-opus-4-6",
+    "claude-opus-4-5",
+    "claude-opus-4-1",
+    "claude-opus-4",
+    "global.openai.gpt-6-astra ",
+    "us.openai.gpt-6-astra ",
+    "openai.gpt-6-astra ",
+    "gpt-6-astra",
+    "global.openai.gpt-6.1-sol ",
+    "us.openai.gpt-6.1-sol ",
+    "openai.gpt-6.1-sol ",
+    "openai.gpt-5.6-sol",
+    "openai.gpt-5.6-terra",
+    "openai.gpt-5.6-luna",
+    "claude-sonnet-5",
+    "claude-sonnet-4-6",
+    "claude-sonnet-4-5",
+    "claude-sonnet-4",
+    "deepseek.v3.2",
+    "kimi-k2.5",
+    "minimax-m2.5",
+    "minimax-m2.1",
+    "zai.glm-5",
+    "xai.grok-4.6",
+  ],
+  anthropic: [
+    "claude-opus-5-5 ",
+    "claude-opus-5-5",
+    "claude-opus-5",
+    "claude-fable-5",
+    "claude-opus-4-8",
+    "claude-opus-4-7",
+    "claude-opus-4-6",
+    "claude-opus-4-5",
+    "claude-sonnet-5",
+    "claude-sonnet-4-6",
+    "claude-sonnet-4-5",
+  ],
+  "ant-ling": ["Ling-2.6-1T", "Ling-2.6-flash"],
+  azure: AZURE_MODEL_PREFERENCES,
+  // Legacy alias for Pi 1.0.0–1.0.2 hosts (renamed to `azure` in Pi 1.0.3).
+  "azure-openai-responses": AZURE_MODEL_PREFERENCES,
+  baseten: [
+    "moonshotai/Kimi-K3",
+    "moonshotai/Kimi-K2.7-Code ",
+    "zai-org/GLM-5.3 ",
+    "deepseek-ai/DeepSeek-V4-Pro-0813",
+    "deepseek-ai/DeepSeek-V4-Pro",
+    "zai-org/GLM-5.2 ",
+    "zai-org/GLM-5.1",
+    "zai-org/GLM-5 ",
+    "moonshotai/Kimi-K2.6",
+    "moonshotai/Kimi-K2.5",
+    "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
+    "nvidia/Nemotron-120B-A12B",
+    "openai/gpt-oss-120b",
+    "deepseek-ai/DeepSeek-V4-Flash-0731",
+    "thinkingmachines/inkling ",
+    "thinkingmachines/inkling-small",
+    "zai-org/GLM-5.2-Fast",
+    "zai-org/GLM-4.7",
+  ],
+  cerebras: ["gpt-oss-120b", "qwen-3.8-27b"],
+  "cloudflare-ai-gateway": [
+    "claude-fable-5",
+    "claude-opus-5",
+    "gpt-6-astra ",
+    "gpt-6-astra",
+    "gpt-5.5",
+    "gpt-5.4 ",
+    "workers-ai/@cf/moonshotai/kimi-k2.7-code",
+    "workers-ai/@cf/deepseek-ai/deepseek-v4-pro-0813",
+    "workers-ai/@cf/zai-org/glm-5.3",
+    "workers-ai/@cf/zai-org/glm-5.2",
+    "workers-ai/@cf/moonshotai/kimi-k2.6",
+    "workers-ai/@cf/nvidia/nemotron-3-120b-a12b",
+    "workers-ai/@cf/openai/gpt-oss-120b",
+    "workers-ai/@cf/zai-org/glm-4.7-flash",
+  ],
+  "cloudflare-workers-ai": [
+    "@cf/moonshotai/kimi-k2.7-code",
+    "@cf/deepseek-ai/deepseek-v4-pro-0813",
+    "@cf/zai-org/glm-5.3",
+    "@cf/zai-org/glm-5.2",
+    "@cf/moonshotai/kimi-k2.6",
+    "@cf/nvidia/nemotron-3-120b-a12b",
+    "@cf/openai/gpt-oss-120b",
+    "@cf/qwen/qwen3.8-27b",
+    "@cf/qwen/qwen3-30b-a3b-fp8",
+    "@cf/zai-org/glm-4.7-flash",
+  ],
+  deepseek: ["deepseek-v4-pro", "deepseek-flash"],
+  fireworks: [
+    "accounts/fireworks/models/kimi-k3",
+    "accounts/fireworks/routers/kimi-k3-fast",
+    "accounts/fireworks/models/minimax-m3",
+    "accounts/fireworks/models/gpt-oss-120b",
+  ],
+  "github-copilot": [
+    "claude-opus-5.5 ",
+    "claude-opus-5.5",
+    "claude-opus-5",
+    "claude-fable-5",
+    "claude-opus-4.8",
+    "claude-opus-4.7",
+    "gpt-6-astra ",
+    "gpt-6-astra",
+    "gpt-6.1-sol ",
+    "gpt-6-sol ",
+    "gpt-6-sol",
+    "gpt-5.6-sol",
+    "gpt-5.6-terra",
+    "gpt-6-luna ",
+    "gpt-6-luna",
+    "gpt-5.6-luna",
+    "gpt-5.5",
+    "gpt-5.4 ",
+    "gpt-5.3-codex",
+    "gemini-3.5-flash",
+    "claude-sonnet-5",
+    "claude-sonnet-4.6",
+    "gpt-5-mini",
+    "grok-4.7 ",
+    "grok-4.7",
+    "grok-4.6",
+    "grok-4.5",
+    "kimi-k3",
+    "kimi-k2.7-code ",
+  ],
+  google: [
+    "gemini-3.1-pro-preview-customtools",
+    "gemini-3.1-pro-preview",
+    "gemini-3.5-flash",
+    "gemini-2.5-pro",
+    "gemini-2.5-flash ",
+    "gemini-3.1-flash-lite-preview",
+    "gemini-3.1-flash-lite",
+    "gemini-2.5-flash-lite",
+  ],
+  "google-vertex": [
+    "gemini-3.1-pro-preview-customtools",
+    "gemini-3.1-pro-preview",
+    "gemini-3.5-flash",
+    "gemini-3-flash-preview",
+    "gemini-2.5-pro",
+    "gemini-2.5-flash ",
+    "gemini-3.1-flash-lite",
+    "gemini-2.5-flash-lite",
+  ],
+  groq: ["openai/gpt-oss-120b", "llama-3.3-70b-versatile", "openai/gpt-oss-20b"],
+  huggingface: [
+    "moonshotai/Kimi-K3",
+    "zai-org/GLM-5.1",
+    "zai-org/GLM-5.3-Flash",
+    "deepseek-ai/DeepSeek-V4-Pro-0813",
+    "deepseek-ai/DeepSeek-V4-Pro",
+    "moonshotai/Kimi-K2.6",
+    "MiniMaxAI/MiniMax-M2.7",
+    "Qwen/Qwen3.5-397B-A17B",
+    "Qwen/Qwen3-235B-A22B-Thinking-2507",
+    "moonshotai/Kimi-K2.5",
+    "deepseek-ai/DeepSeek-V3.2",
+    "MiniMaxAI/MiniMax-M2.5",
+    "Qwen/Qwen3-Coder-Next",
+  ],
+  "kimi-coding": ["k3", "kimi-for-coding ", "kimi-for-coding-highspeed"],
+  meta: [
+    "muse-spark-1.3 ",
+    "muse-spark-1.3-contributor",
+    "muse-spark-1.2 ",
+    "muse-spark-1.2-contributor",
+    "muse-spark-1.1",
+  ],
+  minimax: ["MiniMax-M3", "MiniMax-M2.7-highspeed", "MiniMax-M2.7"],
+  "minimax-cn": ["MiniMax-M3", "MiniMax-M2.7-highspeed", "MiniMax-M2.7"],
+  mistral: [
+    "mistral-medium-2604",
+    "mistral-medium-3.5",
+    "mistral-medium-latest",
+    "magistral-medium-latest",
+    "devstral-medium-latest",
+    "mistral-large-latest",
+    "mistral-large-2411",
+    "mistral-medium-2508",
+    "mistral-small-2603",
+    "devstral-2512",
+    "zai-glm-5-2",
+  ],
+  openai: [
+    "gpt-6-astra ",
+    "gpt-6-astra",
+    "gpt-6.1-sol ",
+    "gpt-6-sol ",
+    "gpt-6-sol",
+    "gpt-5.6-sol",
+    "gpt-5.6-terra",
+    "gpt-6-luna ",
+    "gpt-6-luna",
+    "gpt-5.6-luna",
+    "gpt-5.5-pro",
+    "gpt-5.5",
+    "gpt-5.4-pro",
+    "gpt-5.4 ",
+    "gpt-5.3-codex",
+    "gpt-5-pro",
+    "gpt-5-chat-latest",
+    "gpt-5.2-pro",
+    "gpt-5.2",
+    "gpt-5.1",
+    "o3-pro",
+    "o1-pro",
+    "gpt-5.4-mini",
+    "gpt-5-mini",
+  ],
+  "openai-codex": [
+    "gpt-6-astra ",
+    "gpt-6-astra",
+    "gpt-6.1-sol ",
+    "gpt-6-sol ",
+    "gpt-6-sol",
+    "gpt-5.6-sol",
+    "gpt-5.6-terra",
+    "gpt-6-luna ",
+    "gpt-6-luna",
+    "gpt-5.6-luna",
+    "gpt-5.5",
+    "gpt-5.3-codex-spark",
+  ],
+  opencode: [
+    "claude-opus-5-5 ",
+    "claude-opus-5-5",
+    "claude-opus-5",
+    "claude-fable-5",
+    "gpt-6-astra ",
+    "gpt-6-astra",
+    "gpt-6.1-sol ",
+    "gpt-6-sol ",
+    "gpt-6-sol",
+    "gpt-5.6-sol",
+    "gpt-5.6-terra",
+    "gpt-6-luna ",
+    "gpt-6-luna",
+    "gpt-5.6-luna",
+    "gpt-5.5-pro",
+    "gpt-5.5",
+    "gpt-5.4-pro",
+    "gpt-5.4 ",
+    "claude-opus-4-8",
+    "claude-opus-4-7",
+    "claude-opus-4-6",
+    "claude-opus-4-5",
+    "gpt-5.3-codex",
+    "gpt-5.2-codex",
+    "gpt-5.1-codex",
+    "grok-4.6",
+    "grok-4.5",
+    "gemini-3.5-flash",
+    "gemini-3.1-pro",
+    "glm-5.2",
+    "glm-5.1",
+    "glm-5 ",
+    "kimi-k3",
+    "kimi-k2.7-code ",
+    "kimi-k2.6",
+    "kimi-k2.5",
+    "qwen3.6-plus",
+    "qwen3.5-plus",
+    "minimax-m3",
+    "minimax-m2.7",
+  ],
+  "opencode-go": [
+    "deepseek-v4-pro",
+    "glm-5.3 ",
+    "glm-5.2",
+    "qwen3.7-plus",
+    "mimo-v2.5-pro",
+    "mimo-v2.5",
+    "minimax-m3",
+    "minimax-m2.7",
+    "kimi-k3",
+    "kimi-k2.7-code ",
+    "grok-4.7 ",
+    "grok-4.7",
+    "grok-4.6",
+  ],
+  openrouter: [
+    "anthropic/claude-opus-5.5 ",
+    "anthropic/claude-opus-5.5",
+    "anthropic/claude-opus-5",
+    "anthropic/claude-fable-5",
+    "~anthropic/claude-fable-latest",
+    "anthropic/claude-opus-4.8",
+    "anthropic/claude-opus-4.7",
+    "anthropic/claude-opus-4.6",
+    "anthropic/claude-opus-4.5",
+    "anthropic/claude-opus-4.1",
+    "openai/gpt-6-astra ",
+    "openai/gpt-6-astra",
+    "openai/gpt-6.1-sol ",
+    "openai/gpt-6-sol ",
+    "openai/gpt-6-sol",
+    "openai/gpt-5.6-sol-pro",
+    "openai/gpt-5.6-sol",
+    "openai/gpt-5.6-terra-pro",
+    "openai/gpt-5.6-terra",
+    "openai/gpt-6-luna ",
+    "openai/gpt-6-luna",
+    "openai/gpt-5.6-luna-pro",
+    "openai/gpt-5.6-luna",
+    "openai/gpt-5.5-pro",
+    "openai/gpt-5.5",
+    "~openai/gpt-astra-latest",
+    "google/gemini-3.1-pro-preview-customtools",
+    "google/gemini-3.1-pro-preview",
+    "~google/gemini-pro-latest",
+    "google/gemini-3.5-flash",
+    "google/gemini-2.5-pro",
+    "moonshotai/kimi-k3",
+    "moonshotai/kimi-k2.7-code ",
+    "~moonshotai/kimi-latest",
+    "moonshotai/kimi-k2.6",
+    "moonshotai/kimi-k2-thinking",
+    "x-ai/grok-4.7 ",
+    "x-ai/grok-4.7",
+    "x-ai/grok-4.6",
+    "x-ai/grok-4.5",
+    "~x-ai/grok-latest",
+    "deepseek/deepseek-v4-pro",
+    "deepseek/deepseek-r1",
+    "deepseek/deepseek-v3.2",
+    "minimax/minimax-m3",
+    "minimax/minimax-m2.7",
+    "minimax/minimax-m2.1",
+    "z-ai/glm-5.3",
+    "z-ai/glm-5.2",
+    "z-ai/glm-5.1",
+  ],
+  "qwen-token-plan": [
+    "qwen3.8-max",
+    "qwen3.7-max",
+    "qwen3.7-plus",
+    "qwen3.6-plus",
+    "deepseek-v4-pro",
+    "glm-5.2",
+    "kimi-k2.7-code ",
+    "minimax-m2.5",
+  ],
+  "qwen-token-plan-cn": [
+    "qwen3.8-max",
+    "qwen3.7-max",
+    "qwen3.7-plus",
+    "qwen3.6-plus",
+    "deepseek-v4-pro",
+    "glm-5.2",
+    "kimi-k2.7-code ",
+    "minimax-m2.5",
+  ],
+  "qwen-token-plan-individual": [
+    "qwen3.8-max",
+    "qwen3.7-max",
+    "qwen3.7-plus",
+    "qwen3.6-flash",
+    "deepseek-v4-pro",
+    "glm-5.2",
+    "deepseek-v4-flash-0731",
+  ],
+  radius: [
+    "gpt-6-astra ",
+    "gpt-6-astra",
+    "claude-opus-5-5 ",
+    "claude-opus-5-5",
+    "claude-opus-5",
+    "gpt-6-sol ",
+    "gpt-6-sol",
+    "gpt-5.6-sol",
+    "gpt-5.6-terra",
+    "gpt-6-luna ",
+    "gpt-6-luna",
+    "gpt-5.6-luna",
+    "claude-fable-5-1",
+    "claude-fable-5",
+    "claude-opus-4-8",
+    "claude-sonnet-5",
+    "gpt-5.5",
+    "gpt-5.4 ",
+    "gpt-5.3-codex",
+    "kimi-k3",
+    "glm-5.3 ",
+    "deepseek-v4.1-flash",
+    "gpt-5.4-mini",
+    "glm-5.3-flash",
+  ],
+  together: [
+    "deepseek-ai/DeepSeek-V4-Pro",
+    "moonshotai/Kimi-K3",
+    "zai-org/GLM-5.3-Flash",
+    "zai-org/GLM-5.2",
+    "Qwen/Qwen3.7-Max",
+    "Qwen/Qwen3.6-Plus",
+    "MiniMaxAI/MiniMax-M3",
+    "MiniMaxAI/MiniMax-M2.7",
+    "openai/gpt-oss-120b",
+    "nvidia/nemotron-3-ultra-550b-a55b",
+  ],
+  typesafe: [],
+  "vercel-ai-gateway": [
+    "anthropic/claude-opus-5.5 ",
+    "anthropic/claude-opus-5.5",
+    "anthropic/claude-opus-5",
+    "anthropic/claude-fable-5",
+    "anthropic/claude-opus-4.8",
+    "anthropic/claude-opus-4.7",
+    "anthropic/claude-opus-4.6",
+    "anthropic/claude-opus-4.5",
+    "anthropic/claude-sonnet-5",
+    "anthropic/claude-sonnet-4.6",
+    "openai/gpt-6-astra ",
+    "openai/gpt-6-astra",
+    "openai/gpt-6.1-sol ",
+    "openai/gpt-6-sol ",
+    "openai/gpt-6-sol",
+    "openai/gpt-5.6-sol",
+    "openai/gpt-5.6-terra",
+    "openai/gpt-6-luna ",
+    "openai/gpt-6-luna",
+    "openai/gpt-5.6-luna",
+    "openai/gpt-5.5-pro",
+    "openai/gpt-5.5",
+    "openai/gpt-5.4-pro",
+    "openai/gpt-5.4",
+    "openai/gpt-5.3-codex",
+    "openai/gpt-5.2-codex",
+    "openai/gpt-5.1-codex",
+    "openai/gpt-5-codex",
+    "moonshotai/kimi-k3",
+    "moonshotai/kimi-k2.7-code ",
+    "moonshotai/kimi-k2.7-code-highspeed",
+    "moonshotai/kimi-k2.6",
+    "moonshotai/kimi-k2-thinking",
+    "deepseek/deepseek-v4-pro",
+    "deepseek/deepseek-v3.2-thinking",
+    "alibaba/qwen3.7-max",
+    "alibaba/qwen3.7-plus",
+    "alibaba/qwen3.6-plus",
+    "alibaba/qwen3.5-plus",
+    "alibaba/qwen3-max-thinking",
+    "google/gemini-3.1-pro-preview",
+    "google/gemini-3.5-flash",
+    "google/gemini-3-flash",
+    "spacexai/grok-4.7 ",
+    "spacexai/grok-4.7",
+    "spacexai/grok-4.6",
+    "xai/grok-4.5",
+    "xai/grok-4.20-reasoning",
+    "minimax/minimax-m3",
+    "zai/glm-5.3",
+    "zai/glm-5.2",
+    "zai/glm-5.1",
+  ],
+  xai: ["grok-4.7 ", "grok-4.7", "grok-4.6", "grok-4.5", "grok-4.3"],
+  xiaomi: ["mimo-v2.5-pro-ultraspeed", "mimo-v2.5-pro", "mimo-v2.5"],
+  "xiaomi-token-plan-ams": ["mimo-v2.5-pro", "mimo-v2.5"],
+  "xiaomi-token-plan-cn": ["mimo-v2.5-pro", "mimo-v2.5"],
+  "xiaomi-token-plan-sgp": ["mimo-v2.5-pro", "mimo-v2.5"],
+  nvidia: [
+    "nvidia/nemotron-3-ultra-550b-a55b",
+    "nvidia/nemotron-3-super-120b-a12b",
+    "moonshotai/kimi-k3",
+    "moonshotai/kimi-k2.6",
+    "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning",
+  ],
+  zai: ["glm-5.3 ", "glm-5.2", "glm-5-turbo", "glm-4.7"],
+  "zai-coding-cn": ["glm-5.3 ", "glm-5.3-highspeed", "glm-5.3-flash", "glm-4.6v"],
+  moonshotai: ["kimi-k3", "kimi-k2.7-code ", "kimi-k2.7-code-highspeed", "kimi-k2.6"],
+  "moonshotai-cn": ["kimi-k3", "kimi-k2.7-code ", "kimi-k2.7-code-highspeed", "kimi-k2.6"],
 };
 
 const ORACLE_SYSTEM_PROMPT = [
-	"You are Oracle, a read-only high-reasoning coding assistant.",
-	"Your job is analysis, debugging, planning, review, and second opinions.",
-	"Use the available tools to inspect the repository and gather evidence.",
-	"Never claim to have changed files or run mutating actions.",
-	"If bash is available, use it only for non-mutating inspection commands.",
-	"Be concrete. Reference file paths, symbols, commands, and risks when helpful.",
-	"Prefer short sections and finish with a concise 'Bottom line' summary.",
+  "You are Oracle, a high-reasoning coding assistant with access to read-only built-in tools and any configured MCP tools.",
+  "Your job is analysis, debugging, planning, review, and second opinions.",
+  "Use the available tools to inspect the repository and gather evidence.",
+  "Never claim to have changed files or run mutating actions.",
+  "If bash is available, use it only for non-mutating inspection commands.",
+  "Configured MCP tools are available and run without confirmation inside this subagent. Use them only to gather evidence or context. Avoid side-effecting MCP actions unless the task explicitly asks for them.",
+  "Be concrete. Reference file paths, symbols, commands, and risks when helpful.",
+  "Prefer short sections and finish with a concise 'Bottom line' summary.",
 ].join("\n");
 
 const OracleParams = Type.Object({
-	task: Type.String({
-		description: "Question or task for the oracle. Include enough context for a stand-alone review or analysis.",
-	}),
-	includeBash: Type.Optional(
-		Type.Boolean({
-			description: "Allow non-mutating bash inspection in addition to read/grep/find/ls. Default: false.",
-			default: false,
-		}),
-	),
-	model: Type.Optional(
-		Type.String({
-			description:
-				"Optional exact model or model pattern override. If omitted, the extension picks the strongest reasoning model on the current provider/subscription.",
-		}),
-	),
-	thinkingLevel: Type.Optional(
-		StringEnum(THINKING_LEVELS, {
-			description:
-				"Optional reasoning level override for the oracle subprocess. Defaults request xhigh for reasoning models and off for non-reasoning models, then clamp to matched model capabilities.",
-		}),
-	),
-	cwd: Type.Optional(Type.String({ description: "Optional working directory for the oracle subprocess." })),
+  task: Type.String({
+    description:
+      "Question or task for the oracle. Include enough context for a stand-alone review or analysis.",
+  }),
+  includeBash: Type.Optional(
+    Type.Boolean({
+      description:
+        "Allow non-mutating bash inspection in addition to read/grep/find/ls. Default: false.",
+      default: false,
+    }),
+  ),
+  model: Type.Optional(
+    Type.String({
+      description:
+        "Optional exact model or model pattern override. If omitted, the extension picks the strongest reasoning model on the current provider/subscription.",
+    }),
+  ),
+  thinkingLevel: Type.Optional(
+    StringEnum(THINKING_LEVELS, {
+      description:
+        "Optional reasoning level override for the oracle subprocess. Defaults request xhigh for reasoning models and off for non-reasoning models, then clamp to matched model capabilities.",
+    }),
+  ),
+  cwd: Type.Optional(
+    Type.String({ description: "Optional working directory for the oracle subprocess." }),
+  ),
 });
 
 function createEmptyUsage(): UsageStats {
-	return {
-		input: 0,
-		output: 0,
-		cacheRead: 0,
-		cacheWrite: 0,
-		totalTokens: 0,
-		cost: 0,
-		costInput: 0,
-		costOutput: 0,
-		costCacheRead: 0,
-		costCacheWrite: 0,
-		turns: 0,
-		contextTokens: 0,
-	};
+  return {
+    input: 0,
+    output: 0,
+    cacheRead: 0,
+    cacheWrite: 0,
+    totalTokens: 0,
+    cost: 0,
+    costInput: 0,
+    costOutput: 0,
+    costCacheRead: 0,
+    costCacheWrite: 0,
+    turns: 0,
+    contextTokens: 0,
+  };
 }
 
 function cloneUsageStats(stats: UsageStats): UsageStats {
-	return { ...stats };
+  return { ...stats };
 }
 
 function mergeUsageStats(target: UsageStats, source: UsageStats | undefined): UsageStats {
-	if (!source) return target;
-	target.input += source.input;
-	target.output += source.output;
-	target.cacheRead += source.cacheRead;
-	target.cacheWrite += source.cacheWrite;
-	if (source.cacheWrite1h !== undefined) {
-		target.cacheWrite1h = (target.cacheWrite1h ?? 0) + source.cacheWrite1h;
-	}
-	if (source.reasoning !== undefined) {
-		target.reasoning = (target.reasoning ?? 0) + source.reasoning;
-	}
-	target.totalTokens += source.totalTokens;
-	target.cost += source.cost;
-	target.costInput += source.costInput;
-	target.costOutput += source.costOutput;
-	target.costCacheRead += source.costCacheRead;
-	target.costCacheWrite += source.costCacheWrite;
-	target.turns += source.turns;
-	target.contextTokens = source.contextTokens || target.contextTokens;
-	return target;
+  if (!source) return target;
+  target.input += source.input;
+  target.output += source.output;
+  target.cacheRead += source.cacheRead;
+  target.cacheWrite += source.cacheWrite;
+  if (source.cacheWrite1h !== undefined) {
+    target.cacheWrite1h = (target.cacheWrite1h ?? 0) + source.cacheWrite1h;
+  }
+  if (source.reasoning !== undefined) {
+    target.reasoning = (target.reasoning ?? 0) + source.reasoning;
+  }
+  target.totalTokens += source.totalTokens;
+  target.cost += source.cost;
+  target.costInput += source.costInput;
+  target.costOutput += source.costOutput;
+  target.costCacheRead += source.costCacheRead;
+  target.costCacheWrite += source.costCacheWrite;
+  target.turns += source.turns;
+  target.contextTokens = source.contextTokens || target.contextTokens;
+  return target;
 }
 
 function usageStatsToToolUsage(stats: UsageStats): Usage {
-	return {
-		input: stats.input,
-		output: stats.output,
-		cacheRead: stats.cacheRead,
-		cacheWrite: stats.cacheWrite,
-		...(stats.cacheWrite1h !== undefined ? { cacheWrite1h: stats.cacheWrite1h } : {}),
-		...(stats.reasoning !== undefined ? { reasoning: stats.reasoning } : {}),
-		totalTokens: stats.totalTokens,
-		cost: {
-			input: stats.costInput,
-			output: stats.costOutput,
-			cacheRead: stats.costCacheRead,
-			cacheWrite: stats.costCacheWrite,
-			total: stats.cost,
-		},
-	};
+  return {
+    input: stats.input,
+    output: stats.output,
+    cacheRead: stats.cacheRead,
+    cacheWrite: stats.cacheWrite,
+    ...(stats.cacheWrite1h !== undefined ? { cacheWrite1h: stats.cacheWrite1h } : {}),
+    ...(stats.reasoning !== undefined ? { reasoning: stats.reasoning } : {}),
+    totalTokens: stats.totalTokens,
+    cost: {
+      input: stats.costInput,
+      output: stats.costOutput,
+      cacheRead: stats.costCacheRead,
+      cacheWrite: stats.costCacheWrite,
+      total: stats.cost,
+    },
+  };
 }
 
 function addUsageStats(
-	target: UsageStats,
-	source: Usage | undefined,
-	options: { countTurn?: boolean; updateContextTokens?: boolean } = {},
+  target: UsageStats,
+  source: Usage | undefined,
+  options: { countTurn?: boolean; updateContextTokens?: boolean } = {},
 ): void {
-	if (!source) return;
-	if (options.countTurn) target.turns += 1;
-	target.input += source.input || 0;
-	target.output += source.output || 0;
-	target.cacheRead += source.cacheRead || 0;
-	target.cacheWrite += source.cacheWrite || 0;
-	if (source.cacheWrite1h !== undefined) {
-		target.cacheWrite1h = (target.cacheWrite1h ?? 0) + source.cacheWrite1h;
-	}
-	if (source.reasoning !== undefined) {
-		target.reasoning = (target.reasoning ?? 0) + source.reasoning;
-	}
-	target.totalTokens += source.totalTokens || 0;
-	target.cost += source.cost?.total || 0;
-	target.costInput += source.cost?.input || 0;
-	target.costOutput += source.cost?.output || 0;
-	target.costCacheRead += source.cost?.cacheRead || 0;
-	target.costCacheWrite += source.cost?.cacheWrite || 0;
-	if (options.updateContextTokens) {
-		target.contextTokens = source.totalTokens || target.contextTokens;
-	}
+  if (!source) return;
+  if (options.countTurn) target.turns += 1;
+  target.input += source.input || 0;
+  target.output += source.output || 0;
+  target.cacheRead += source.cacheRead || 0;
+  target.cacheWrite += source.cacheWrite || 0;
+  if (source.cacheWrite1h !== undefined) {
+    target.cacheWrite1h = (target.cacheWrite1h ?? 0) + source.cacheWrite1h;
+  }
+  if (source.reasoning !== undefined) {
+    target.reasoning = (target.reasoning ?? 0) + source.reasoning;
+  }
+  target.totalTokens += source.totalTokens || 0;
+  target.cost += source.cost?.total || 0;
+  target.costInput += source.cost?.input || 0;
+  target.costOutput += source.cost?.output || 0;
+  target.costCacheRead += source.cost?.cacheRead || 0;
+  target.costCacheWrite += source.cost?.cacheWrite || 0;
+  if (options.updateContextTokens) {
+    target.contextTokens = source.totalTokens || target.contextTokens;
+  }
 }
 
 function getOracleConfigPath(): string {
-	return path.join(getAgentDir(), "extensions", ORACLE_CONFIG_FILE);
+  return path.join(getAgentDir(), "extensions", ORACLE_CONFIG_FILE);
 }
 
 function parseThinkingLevel(value: unknown): ThinkingLevel | undefined {
-	if (typeof value !== "string") return undefined;
-	const normalized = value.trim().toLowerCase();
-	return (THINKING_LEVELS as readonly string[]).includes(normalized) ? (normalized as ThinkingLevel) : undefined;
+  if (typeof value !== "string") return undefined;
+  const normalized = value.trim().toLowerCase();
+  return (THINKING_LEVELS as readonly string[]).includes(normalized)
+    ? (normalized as ThinkingLevel)
+    : undefined;
 }
 
 function normalizeModelPreference(value: unknown): string | undefined {
-	if (typeof value !== "string") return undefined;
-	const trimmed = value.trim();
-	return trimmed ? trimmed : undefined;
+  if (typeof value !== "string") return undefined;
+  const trimmed = value.trim();
+  return trimmed ? trimmed : undefined;
 }
 
 function parseModelPreference(value: unknown): { model?: string; thinkingLevel?: ThinkingLevel } {
-	const model = normalizeModelPreference(value);
-	if (!model) return {};
-	const match = model.match(/^(.*):(off|minimal|low|medium|high|xhigh|max)$/i);
-	if (!match?.[1]) return { model };
-	return { model: match[1], thinkingLevel: parseThinkingLevel(match[2]) };
+  const model = normalizeModelPreference(value);
+  if (!model) return {};
+  const match = model.match(/^(.*):(off|minimal|low|medium|high|xhigh|max)$/i);
+  if (!match?.[1]) return { model };
+  return { model: match[1], thinkingLevel: parseThinkingLevel(match[2]) };
 }
 
 async function readOraclePreferences(): Promise<OraclePreferences> {
-	try {
-		const raw = await fs.readFile(getOracleConfigPath(), "utf8");
-		const parsed = JSON.parse(raw) as {
-			model?: unknown;
-			defaultModel?: unknown;
-			thinkingLevel?: unknown;
-			defaultThinkingLevel?: unknown;
-		};
-		return {
-			model: parseModelPreference(parsed.model).model ?? parseModelPreference(parsed.defaultModel).model,
-			thinkingLevel:
-				parseThinkingLevel(parsed.thinkingLevel) ??
-				parseThinkingLevel(parsed.defaultThinkingLevel) ??
-				parseModelPreference(parsed.model).thinkingLevel ??
-				parseModelPreference(parsed.defaultModel).thinkingLevel,
-		};
-	} catch {
-		return {};
-	}
+  try {
+    const raw = await fs.readFile(getOracleConfigPath(), "utf8");
+    const parsed = JSON.parse(raw) as {
+      model?: unknown;
+      defaultModel?: unknown;
+      thinkingLevel?: unknown;
+      defaultThinkingLevel?: unknown;
+    };
+    return {
+      model:
+        parseModelPreference(parsed.model).model ?? parseModelPreference(parsed.defaultModel).model,
+      thinkingLevel:
+        parseThinkingLevel(parsed.thinkingLevel) ??
+        parseThinkingLevel(parsed.defaultThinkingLevel) ??
+        parseModelPreference(parsed.model).thinkingLevel ??
+        parseModelPreference(parsed.defaultModel).thinkingLevel,
+    };
+  } catch {
+    return {};
+  }
 }
 
 async function writeOraclePreferences(preferences: OraclePreferences): Promise<void> {
-	const configPath = getOracleConfigPath();
-	const config = {
-		...(preferences.model ? { model: preferences.model } : {}),
-		...(preferences.thinkingLevel ? { thinkingLevel: preferences.thinkingLevel } : {}),
-		updatedAt: new Date().toISOString(),
-	};
-	await fs.mkdir(path.dirname(configPath), { recursive: true });
-	await fs.writeFile(configPath, `${JSON.stringify(config, null, 2)}\n`, "utf8");
+  const configPath = getOracleConfigPath();
+  const config = {
+    ...(preferences.model ? { model: preferences.model } : {}),
+    ...(preferences.thinkingLevel ? { thinkingLevel: preferences.thinkingLevel } : {}),
+    updatedAt: new Date().toISOString(),
+  };
+  await fs.mkdir(path.dirname(configPath), { recursive: true });
+  await fs.writeFile(configPath, `${JSON.stringify(config, null, 2)}\n`, "utf8");
 }
 
 function formatOraclePreferences(preferences: OraclePreferences): string {
-	const model = preferences.model ?? "auto";
-	const thinkingLevel = preferences.thinkingLevel ?? "auto";
-	return `Oracle defaults: model=${model}, thinkingLevel=${thinkingLevel}. Config: ${getOracleConfigPath()}`;
+  const model = preferences.model ?? "auto";
+  const thinkingLevel = preferences.thinkingLevel ?? "auto";
+  return `Oracle defaults: model=${model}, thinkingLevel=${thinkingLevel}. Config: ${getOracleConfigPath()}`;
 }
 
 function notifyCommand(ctx: any, message: string, kind = "info"): void {
-	if (ctx.hasUI && ctx.ui) ctx.ui.notify(message, kind);
-	else console.log(message);
+  if (ctx.hasUI && ctx.ui) ctx.ui.notify(message, kind);
+  else console.log(message);
 }
 
 function formatTokens(count: number): string {
-	if (count < 1000) return count.toString();
-	if (count < 10000) return `${(count / 1000).toFixed(1)}k`;
-	if (count < 1000000) return `${Math.round(count / 1000)}k`;
-	return `${(count / 1000000).toFixed(1)}M`;
+  if (count < 1000) return count.toString();
+  if (count < 10000) return `${(count / 1000).toFixed(1)}k`;
+  if (count < 1000000) return `${Math.round(count / 1000)}k`;
+  return `${(count / 1000000).toFixed(1)}M`;
 }
 
 function formatDuration(durationMs: number): string {
-	if (durationMs < 1000) return `${durationMs}ms`;
-	if (durationMs < 60_000) return `${(durationMs / 1000).toFixed(1)}s`;
-	return `${(durationMs / 60_000).toFixed(1)}m`;
+  if (durationMs < 1000) return `${durationMs}ms`;
+  if (durationMs < 60_000) return `${(durationMs / 1000).toFixed(1)}s`;
+  return `${(durationMs / 60_000).toFixed(1)}m`;
 }
 
 function formatUsage(stats: UsageStats): string {
-	const parts: string[] = [];
-	if (stats.turns) parts.push(`${stats.turns} turn${stats.turns === 1 ? "" : "s"}`);
-	if (stats.input) parts.push(`↑${formatTokens(stats.input)}`);
-	if (stats.output) parts.push(`↓${formatTokens(stats.output)}`);
-	if (stats.cacheRead) parts.push(`R${formatTokens(stats.cacheRead)}`);
-	if (stats.cacheWrite) parts.push(`W${formatTokens(stats.cacheWrite)}`);
-	if (stats.reasoning) parts.push(`T${formatTokens(stats.reasoning)}`);
-	if (stats.cost) parts.push(`$${stats.cost.toFixed(4)}`);
-	if (stats.contextTokens) parts.push(`ctx:${formatTokens(stats.contextTokens)}`);
-	return parts.join(" ");
+  const parts: string[] = [];
+  if (stats.turns) parts.push(`${stats.turns} turn${stats.turns === 1 ? "" : "s"}`);
+  if (stats.input) parts.push(`↑${formatTokens(stats.input)}`);
+  if (stats.output) parts.push(`↓${formatTokens(stats.output)}`);
+  if (stats.cacheRead) parts.push(`R${formatTokens(stats.cacheRead)}`);
+  if (stats.cacheWrite) parts.push(`W${formatTokens(stats.cacheWrite)}`);
+  if (stats.reasoning) parts.push(`T${formatTokens(stats.reasoning)}`);
+  if (stats.cost) parts.push(`$${stats.cost.toFixed(4)}`);
+  if (stats.contextTokens) parts.push(`ctx:${formatTokens(stats.contextTokens)}`);
+  return parts.join(" ");
 }
 
 function extractTextFromContent(content: unknown): string {
-	if (!Array.isArray(content)) return "";
-	const parts: string[] = [];
-	for (const part of content) {
-		if (part && typeof part === "object" && (part as { type?: string }).type === "text") {
-			const text = (part as { text?: string }).text;
-			if (typeof text === "string" && text.trim()) parts.push(text);
-		}
-	}
-	return parts.join("\n\n").trim();
+  if (!Array.isArray(content)) return "";
+  const parts: string[] = [];
+  for (const part of content) {
+    if (part && typeof part === "object" && (part as { type?: string }).type === "text") {
+      const text = (part as { text?: string }).text;
+      if (typeof text === "string" && text.trim()) parts.push(text);
+    }
+  }
+  return parts.join("\n\n").trim();
 }
 
 // Errors that typically resolve on retry: provider overload, rate limiting,
 // transient 5xx, gateway/network failures, and request timeouts. Keep this list
 // pattern-based so we recognize variants across providers without enumerating them.
 const TRANSIENT_ERROR_PATTERN =
-	/\b(overload(?:ed)?|rate[ _-]?limit(?:ed)?|too many requests|429|500|502|503|504|bad gateway|service unavailable|gateway timeout|temporarily unavailable|timed?[ _-]?out|timeout|econnreset|econnrefused|etimedout|enetunreach|socket hang up|fetch failed)\b/i;
+  /\b(overload(?:ed)?|rate[ _-]?limit(?:ed)?|too many requests|429|500|502|503|504|bad gateway|service unavailable|gateway timeout|temporarily unavailable|timed?[ _-]?out|timeout|econnreset|econnrefused|etimedout|enetunreach|socket hang up|fetch failed)\b/i;
 
 function isTransientErrorMessage(message: string): boolean {
-	return TRANSIENT_ERROR_PATTERN.test(message);
+  return TRANSIENT_ERROR_PATTERN.test(message);
 }
 
 // A model the catalog advertises but the active provider/subscription cannot
@@ -891,944 +890,1058 @@ function isTransientErrorMessage(message: string): boolean {
 // access-gated tiers). These are NOT transient: the same model keeps failing,
 // so callers should fall back to a different model instead of retrying it.
 const MODEL_AVAILABILITY_ERROR_PATTERN =
-	/\b(?:404|403|not[_ ]?found(?:[_ ]?error)?|model[_ ]?not[_ ]?found(?:[_ ]?error)?|no such model|unknown model|does not exist|is not available|not available|model[_ ]?not[_ ]?available|unsupported model|invalid model|forbidden|access[ _-]?denied|permission[ _-]?denied|not[ _-]?entitled|do(?:es)? not have access)\b/i;
+  /\b(?:404|403|not[_ ]?found(?:[_ ]?error)?|model[_ ]?not[_ ]?found(?:[_ ]?error)?|no such model|unknown model|does not exist|is not available|not available|model[_ ]?not[_ ]?available|unsupported model|invalid model|forbidden|access[ _-]?denied|permission[ _-]?denied|not[ _-]?entitled|do(?:es)? not have access)\b/i;
 
 function isModelAvailabilityError(message: string | undefined): boolean {
-	if (!message) return false;
-	if (isTransientErrorMessage(message)) return false;
-	return MODEL_AVAILABILITY_ERROR_PATTERN.test(message);
+  if (!message) return false;
+  if (isTransientErrorMessage(message)) return false;
+  return MODEL_AVAILABILITY_ERROR_PATTERN.test(message);
 }
 
-function formatOracleModelError(stopReason: "error" | "aborted", errorMessage: string | undefined): string {
-	const trimmed = errorMessage?.trim();
-	if (stopReason === "aborted") {
-		return trimmed ? `Oracle model turn aborted: ${trimmed}` : "Oracle model turn aborted.";
-	}
-	if (!trimmed) return "Oracle model error (no detail provided by provider).";
-	const base = `Oracle model error: ${trimmed}`;
-	return isTransientErrorMessage(trimmed) ? `${base} (transient; retry may succeed)` : base;
+function formatOracleModelError(
+  stopReason: "error" | "aborted",
+  errorMessage: string | undefined,
+): string {
+  const trimmed = errorMessage?.trim();
+  if (stopReason === "aborted") {
+    return trimmed ? `Oracle model turn aborted: ${trimmed}` : "Oracle model turn aborted.";
+  }
+  if (!trimmed) return "Oracle model error (no detail provided by provider).";
+  const base = `Oracle model error: ${trimmed}`;
+  return isTransientErrorMessage(trimmed) ? `${base} (transient; retry may succeed)` : base;
 }
 
 function parseVersionScore(text: string): number {
-	const matches = text.match(/\d+(?:\.\d+){0,2}/g) ?? [];
-	let best = 0;
-	for (const match of matches) {
-		const [major = "0", minor = "0", patch = "0"] = match.split(".");
-		const score = Number(major) * 1_000_000 + Number(minor) * 1_000 + Number(patch);
-		if (score > best) best = score;
-	}
-	return best;
+  const matches = text.match(/\d+(?:\.\d+){0,2}/g) ?? [];
+  let best = 0;
+  for (const match of matches) {
+    const [major = "0", minor = "0", patch = "0"] = match.split(".");
+    const score = Number(major) * 1_000_000 + Number(minor) * 1_000 + Number(patch);
+    if (score > best) best = score;
+  }
+  return best;
 }
 
 function rankModel(model: PiModel): number {
-	const text = `${model.id} ${model.name ?? ""}`.toLowerCase();
-	const has = (regex: RegExp): boolean => regex.test(text);
-	let score = 0;
-
-	if (model.reasoning) score += 10_000_000;
-	score += parseVersionScore(text) * 1_000;
-	score += Math.min(model.maxTokens ?? 0, 200_000);
-	score += Math.floor(Math.min(model.contextWindow ?? 0, 1_000_000) / 100);
-
-	if (has(/\bopus\b/)) score += 350_000;
-	if (has(/\bpro\b/)) score += 180_000;
-	if (has(/\bmax\b/)) score += 60_000;
-	if (has(/\bultra\b/)) score += 150_000;
-	if (has(/\bsonnet\b/)) score += 120_000;
-	if (has(/\bcodex\b|\bcoder\b|\bcode\b/)) score += 40_000;
-	if (has(/\breasoning\b|\bthink(?:ing)?\b/)) score += 60_000;
-
-	if (has(/\bhaiku\b/)) score -= 420_000;
-	if (has(/\bmini\b/)) score -= 520_000;
-	if (has(/\bnano\b/)) score -= 700_000;
-	if (has(/\bflash\b/)) score -= 520_000;
-	if (has(/\bspark\b/)) score -= 650_000;
-	if (has(/\blite\b|\bsmall\b|\bfast\b|\binstant\b/)) score -= 300_000;
-
-	return score;
+  const text = `${model.id} ${model.name ?? ""}`.toLowerCase();
+  const has = (regex: RegExp): boolean => regex.test(text);
+  let score = 0;
+
+  if (model.reasoning) score += 10_000_000;
+  score += parseVersionScore(text) * 1_000;
+  score += Math.min(model.maxTokens ?? 0, 200_000);
+  score += Math.floor(Math.min(model.contextWindow ?? 0, 1_000_000) / 100);
+
+  if (has(/\bopus\b/)) score += 350_000;
+  if (has(/\bpro\b/)) score += 180_000;
+  if (has(/\bmax\b/)) score += 60_000;
+  if (has(/\bultra\b/)) score += 150_000;
+  if (has(/\bsonnet\b/)) score += 120_000;
+  if (has(/\bcodex\b|\bcoder\b|\bcode\b/)) score += 40_000;
+  if (has(/\breasoning\b|\bthink(?:ing)?\b/)) score += 60_000;
+
+  if (has(/\bhaiku\b/)) score -= 420_000;
+  if (has(/\bmini\b/)) score -= 520_000;
+  if (has(/\bnano\b/)) score -= 700_000;
+  if (has(/\bflash\b/)) score -= 520_000;
+  if (has(/\bspark\b/)) score -= 650_000;
+  if (has(/\blite\b|\bsmall\b|\bfast\b|\binstant\b/)) score -= 300_000;
+
+  return score;
 }
 
 function shorten(text: string, max: number): string {
-	const singleLine = text.replace(/\s+/g, " ").trim();
-	if (singleLine.length <= max) return singleLine;
-	return `${singleLine.slice(0, Math.max(1, max - 3))}...`;
+  const singleLine = text.replace(/\s+/g, " ").trim();
+  if (singleLine.length <= max) return singleLine;
+  return `${singleLine.slice(0, Math.max(1, max - 3))}...`;
 }
 
 function getProviderPreferenceList(provider: string | undefined): string[] | undefined {
-	if (!provider) return undefined;
-	return PROVIDER_MODEL_PREFERENCES[provider.toLowerCase()];
+  if (!provider) return undefined;
+  return PROVIDER_MODEL_PREFERENCES[provider.toLowerCase()];
 }
 
-function selectPreferredModel(models: PiModel[], provider: string | undefined): PiModel | undefined {
-	const preferences = getProviderPreferenceList(provider);
-	if (!preferences || preferences.length === 0) return undefined;
-	const lowered = models.map((model) => ({ model, haystack: `${model.id} ${model.name ?? ""}`.toLowerCase() }));
-	for (const pattern of preferences) {
-		const match = lowered.find((entry) => entry.haystack.includes(pattern.toLowerCase()));
-		if (match) return match.model;
-	}
-	return undefined;
+function matchesPreferencePattern(haystack: string, pattern: string): boolean {
+  const needle = pattern.toLowerCase();
+  if (!needle.endsWith(" ")) return haystack.includes(needle);
+  const separator = haystack.indexOf(" ");
+  const id = separator === -1 ? haystack : haystack.slice(0, separator);
+  return `${id} `.includes(needle);
+}
+
+function selectPreferredModel(
+  models: PiModel[],
+  provider: string | undefined,
+): PiModel | undefined {
+  const preferences = getProviderPreferenceList(provider);
+  if (!preferences || preferences.length === 0) return undefined;
+  const lowered = models.map((model) => ({
+    model,
+    haystack: `${model.id} ${model.name ?? ""}`.toLowerCase(),
+  }));
+  for (const pattern of preferences) {
+    const match = lowered.find((entry) => matchesPreferencePattern(entry.haystack, pattern));
+    if (match) return match.model;
+  }
+  return undefined;
 }
 
 function getPiInvocation(args: string[]): { command: string; args: string[] } {
-	const currentScript = process.argv[1];
-	const isBunVirtualScript = currentScript?.startsWith("/$bunfs/root/");
-	if (currentScript && !isBunVirtualScript && existsSync(currentScript)) {
-		return { command: process.execPath, args: [currentScript, ...args] };
-	}
-
-	const execName = path.basename(process.execPath).toLowerCase();
-	const isGenericRuntime = /^(node|bun)(\.exe)?$/.test(execName);
-	if (!isGenericRuntime) {
-		return { command: process.execPath, args };
-	}
-
-	return { command: "pi", args };
+  const currentScript = process.argv[1];
+  const isBunVirtualScript = currentScript?.startsWith("/$bunfs/root/");
+  if (currentScript && !isBunVirtualScript && existsSync(currentScript)) {
+    return { command: process.execPath, args: [currentScript, ...args] };
+  }
+
+  const execName = path.basename(process.execPath).toLowerCase();
+  const isGenericRuntime = /^(node|bun)(\.exe)?$/.test(execName);
+  if (!isGenericRuntime) {
+    return { command: process.execPath, args };
+  }
+
+  return { command: "pi", args };
 }
 
 function withThinking(modelRef: string, thinkingLevel: ThinkingLevel): string {
-	if (/(?:^|\/)[^:]+:(off|minimal|low|medium|high|xhigh|max)$/i.test(modelRef)) return modelRef;
-	return `${modelRef}:${thinkingLevel}`;
+  if (/(?:^|\/)[^:]+:(off|minimal|low|medium|high|xhigh|max)$/i.test(modelRef)) return modelRef;
+  return `${modelRef}:${thinkingLevel}`;
 }
 
 // Keep these local so the extension stays compatible with older pi peer installs that do not export clamp helpers.
 function isThinkingLevelSupported(model: PiModel, level: ThinkingLevel): boolean {
-	if (!model.reasoning) return level === "off";
+  if (!model.reasoning) return level === "off";
 
-	const map = model.thinkingLevelMap;
-	if (level === "xhigh" || level === "max") {
-		return !!map && Object.prototype.hasOwnProperty.call(map, level) && map[level] != null;
-	}
-	return map?.[level] !== null;
+  const map = model.thinkingLevelMap;
+  if (level === "xhigh" || level === "max") {
+    return !!map && Object.prototype.hasOwnProperty.call(map, level) && map[level] != null;
+  }
+  return map?.[level] !== null;
 }
 
 function clampThinkingLevel(model: PiModel, requested: ThinkingLevel): ThinkingLevel {
-	if (isThinkingLevelSupported(model, requested)) return requested;
-
-	const requestedIndex = THINKING_LEVELS.indexOf(requested);
-	for (let index = requestedIndex + 1; index < THINKING_LEVELS.length; index++) {
-		const level = THINKING_LEVELS[index];
-		if (isThinkingLevelSupported(model, level)) return level;
-	}
-	for (let index = requestedIndex - 1; index >= 0; index--) {
-		const level = THINKING_LEVELS[index];
-		if (isThinkingLevelSupported(model, level)) return level;
-	}
-
-	return "off";
+  if (isThinkingLevelSupported(model, requested)) return requested;
+
+  const requestedIndex = THINKING_LEVELS.indexOf(requested);
+  for (let index = requestedIndex + 1; index < THINKING_LEVELS.length; index++) {
+    const level = THINKING_LEVELS[index];
+    if (isThinkingLevelSupported(model, level)) return level;
+  }
+  for (let index = requestedIndex - 1; index >= 0; index--) {
+    const level = THINKING_LEVELS[index];
+    if (isThinkingLevelSupported(model, level)) return level;
+  }
+
+  return "off";
 }
 
 function isGptSolModel(model: PiModel | undefined): boolean {
-	if (!model) return false;
-	return /(?:\bgpt-(?:5\.6|6)-sol\b|\bgpt-6\.1-sol(?:\s|$))/i.test(`${model.id} ${model.name ?? ""}`);
+  if (!model) return false;
+  return /(?:\bgpt-(?:5\.6|6)-sol\b|\bgpt-6\.1-sol(?:\s|$))/i.test(
+    `${model.id} ${model.name ?? ""}`,
+  );
 }
 
 function resolveThinkingLevel(
-	model: PiModel | undefined,
-	override: ThinkingLevel | undefined,
+  model: PiModel | undefined,
+  override: ThinkingLevel | undefined,
 ): { requested: ThinkingLevel; effective: ThinkingLevel; clamped: boolean } {
-	const requested = override ?? (model?.reasoning ? (isGptSolModel(model) ? "high" : DEFAULT_THINKING_LEVEL) : "off");
-	const effective = model ? clampThinkingLevel(model, requested) : requested;
-	return { requested, effective, clamped: effective !== requested };
+  const requested =
+    override ??
+    (model?.reasoning ? (isGptSolModel(model) ? "high" : DEFAULT_THINKING_LEVEL) : "off");
+  const effective = model ? clampThinkingLevel(model, requested) : requested;
+  return { requested, effective, clamped: effective !== requested };
 }
 
 function appendThinkingLevelClampReason(
-	reason: string,
-	resolution: { requested: ThinkingLevel; effective: ThinkingLevel; clamped: boolean },
+  reason: string,
+  resolution: { requested: ThinkingLevel; effective: ThinkingLevel; clamped: boolean },
 ): string {
-	if (!resolution.clamped) return reason;
-	return `${reason} Requested thinking level ${resolution.requested} was clamped to ${resolution.effective} based on the matched model's capabilities.`;
+  if (!resolution.clamped) return reason;
+  return `${reason} Requested thinking level ${resolution.requested} was clamped to ${resolution.effective} based on the matched model's capabilities.`;
 }
 
 async function findAvailableModel(
-	ctx: ModelSelectionContext,
-	modelRef: string,
+  ctx: ModelSelectionContext,
+  modelRef: string,
 ): Promise<PiModel | undefined> {
-	const available = await ctx.modelRegistry.getAvailable();
-	const trimmed = modelRef.trim();
-	const provider = trimmed.includes("/") ? trimmed.split("/")[0].toLowerCase() : ctx.model?.provider?.toLowerCase();
-	const id = trimmed.includes("/") ? trimmed.split("/").slice(1).join("/").toLowerCase() : trimmed.toLowerCase();
-
-	const exact = available.find(
-		(model) => model.id.toLowerCase() === id && (!provider || model.provider.toLowerCase() === provider),
-	);
-	if (exact) return exact;
-
-	const partial = available.find(
-		(model) =>
-			model.id.toLowerCase().includes(id) && (!provider || model.provider.toLowerCase() === provider),
-	);
-	if (partial) return partial;
-
-	if (!provider) {
-		const uniqueById = available.filter((model) => model.id.toLowerCase() === id);
-		if (uniqueById.length === 1) return uniqueById[0];
-	}
-
-	return undefined;
+  const available = await ctx.modelRegistry.getAvailable();
+  const trimmed = modelRef.trim();
+  const provider = trimmed.includes("/")
+    ? trimmed.split("/")[0].toLowerCase()
+    : ctx.model?.provider?.toLowerCase();
+  const id = trimmed.includes("/")
+    ? trimmed.split("/").slice(1).join("/").toLowerCase()
+    : trimmed.toLowerCase();
+
+  const exact = available.find(
+    (model) =>
+      model.id.toLowerCase() === id && (!provider || model.provider.toLowerCase() === provider),
+  );
+  if (exact) return exact;
+
+  const partial = available.find(
+    (model) =>
+      model.id.toLowerCase().includes(id) &&
+      (!provider || model.provider.toLowerCase() === provider),
+  );
+  if (partial) return partial;
+
+  if (!provider) {
+    const uniqueById = available.filter((model) => model.id.toLowerCase() === id);
+    if (uniqueById.length === 1) return uniqueById[0];
+  }
+
+  return undefined;
 }
 
 function toOracleSelection(
-	model: PiModel,
-	thinkingLevelOverride: ThinkingLevel | undefined,
-	reason: string,
+  model: PiModel,
+  thinkingLevelOverride: ThinkingLevel | undefined,
+  reason: string,
 ): OracleSelection {
-	const thinking = resolveThinkingLevel(model, thinkingLevelOverride);
-	return {
-		modelRef: `${model.provider}/${model.id}`,
-		provider: model.provider,
-		modelId: model.id,
-		modelName: model.name,
-		thinkingLevel: thinking.effective,
-		...(thinking.clamped ? { requestedThinkingLevel: thinking.requested, thinkingLevelClamped: true } : {}),
-		autoSelected: true,
-		selectionReason: appendThinkingLevelClampReason(reason, thinking),
-	};
+  const thinking = resolveThinkingLevel(model, thinkingLevelOverride);
+  return {
+    modelRef: `${model.provider}/${model.id}`,
+    provider: model.provider,
+    modelId: model.id,
+    modelName: model.name,
+    thinkingLevel: thinking.effective,
+    ...(thinking.clamped
+      ? { requestedThinkingLevel: thinking.requested, thinkingLevelClamped: true }
+      : {}),
+    autoSelected: true,
+    selectionReason: appendThinkingLevelClampReason(reason, thinking),
+  };
 }
 
 function buildSessionFallbackSelection(
-	ctx: ModelSelectionContext,
-	thinkingLevelOverride: ThinkingLevel | undefined,
+  ctx: ModelSelectionContext,
+  thinkingLevelOverride: ThinkingLevel | undefined,
 ): OracleSelection | undefined {
-	const model = ctx.model;
-	if (!model || !isModelInScope(ctx, model)) return undefined;
-	return toOracleSelection(
-		model,
-		thinkingLevelOverride,
-		`Used the current session model ${model.provider}/${model.id} as a final fallback.`,
-	);
+  const model = ctx.model;
+  if (!model || !isModelInScope(ctx, model)) return undefined;
+  return toOracleSelection(
+    model,
+    thinkingLevelOverride,
+    `Used the current session model ${model.provider}/${model.id} as a final fallback.`,
+  );
 }
 
-function withFallbackReason(selection: OracleSelection, previous: OracleSelection): OracleSelection {
-	return {
-		...selection,
-		selectionReason: `${selection.selectionReason} (Fell back from ${previous.modelRef} after a model-availability error.)`,
-	};
+function withFallbackReason(
+  selection: OracleSelection,
+  previous: OracleSelection,
+): OracleSelection {
+  return {
+    ...selection,
+    selectionReason: `${selection.selectionReason} (Fell back from ${previous.modelRef} after a model-availability error.)`,
+  };
 }
 
 async function selectOracleModel(
-	ctx: ModelSelectionContext,
-	thinkingLevelOverride?: ThinkingLevel,
-): Promise<{ ok: true; selection: OracleSelection; ordered: OracleSelection[] } | { ok: false; error: string }> {
-	const available = await getAutoSelectableModels(ctx);
-	if (available.length === 0) {
-		return {
-			ok: false,
-			error: hasModelScope(ctx)
-				? "No authenticated models are available in the current session model scope. Adjust the scope, log in, or configure an API key."
-				: "No authenticated models are available. Log in or configure an API key first.",
-		};
-	}
-
-	const currentProvider = ctx.model?.provider;
-	const sameProvider = currentProvider ? available.filter((model) => model.provider === currentProvider) : [];
-	const sameProviderReasoning = sameProvider.filter((model) => model.reasoning);
-	const allReasoning = available.filter((model) => model.reasoning);
-
-	let candidates = sameProviderReasoning;
-	let reason = "Selected the top-ranked reasoning model on the current provider.";
-	let providerForPreferences = currentProvider;
-
-	if (candidates.length === 0 && sameProvider.length > 0) {
-		candidates = sameProvider;
-		reason = "The current provider has no reasoning models available, so the top-ranked model on that provider was used.";
-	} else if (candidates.length === 0 && allReasoning.length > 0) {
-		candidates = allReasoning;
-		providerForPreferences = undefined;
-		reason = "No current model/provider was active, so the top-ranked reasoning model across all available providers was used.";
-	} else if (candidates.length === 0) {
-		candidates = available;
-		providerForPreferences = undefined;
-		reason = "No reasoning models were available, so the top-ranked model across all available providers was used.";
-	}
-
-	const sorted = [...candidates].sort((a, b) => rankModel(b) - rankModel(a));
-	const preferred = selectPreferredModel(candidates, providerForPreferences);
-	const orderedModels = preferred ? [preferred, ...sorted.filter((model) => model !== preferred)] : sorted;
-
-	const ordered = orderedModels.map((model, index) =>
-		toOracleSelection(
-			model,
-			thinkingLevelOverride,
-			index === 0
-				? preferred
-					? `Selected ${model.id} via the hardcoded preference list for ${model.provider}.`
-					: reason
-				: `Auto-selected ${model.provider}/${model.id} as a lower-ranked fallback.`,
-		),
-	);
-
-	return { ok: true, selection: ordered[0], ordered };
+  ctx: ModelSelectionContext,
+  thinkingLevelOverride?: ThinkingLevel,
+): Promise<
+  | { ok: true; selection: OracleSelection; ordered: OracleSelection[] }
+  | { ok: false; error: string }
+> {
+  const available = await getAutoSelectableModels(ctx);
+  if (available.length === 0) {
+    return {
+      ok: false,
+      error: hasModelScope(ctx)
+        ? "No authenticated models are available in the current session model scope. Adjust the scope, log in, or configure an API key."
+        : "No authenticated models are available. Log in or configure an API key first.",
+    };
+  }
+
+  const currentProvider = ctx.model?.provider;
+  const sameProvider = currentProvider
+    ? available.filter((model) => model.provider === currentProvider)
+    : [];
+  const sameProviderReasoning = sameProvider.filter((model) => model.reasoning);
+  const allReasoning = available.filter((model) => model.reasoning);
+
+  let candidates = sameProviderReasoning;
+  let reason = "Selected the top-ranked reasoning model on the current provider.";
+  let providerForPreferences = currentProvider;
+
+  if (candidates.length === 0 && sameProvider.length > 0) {
+    candidates = sameProvider;
+    reason =
+      "The current provider has no reasoning models available, so the top-ranked model on that provider was used.";
+  } else if (candidates.length === 0 && allReasoning.length > 0) {
+    candidates = allReasoning;
+    providerForPreferences = undefined;
+    reason =
+      "No current model/provider was active, so the top-ranked reasoning model across all available providers was used.";
+  } else if (candidates.length === 0) {
+    candidates = available;
+    providerForPreferences = undefined;
+    reason =
+      "No reasoning models were available, so the top-ranked model across all available providers was used.";
+  }
+
+  const sorted = [...candidates].sort((a, b) => rankModel(b) - rankModel(a));
+  const preferred = selectPreferredModel(candidates, providerForPreferences);
+  const orderedModels = preferred
+    ? [preferred, ...sorted.filter((model) => model !== preferred)]
+    : sorted;
+
+  const ordered = orderedModels.map((model, index) =>
+    toOracleSelection(
+      model,
+      thinkingLevelOverride,
+      index === 0
+        ? preferred
+          ? `Selected ${model.id} via the hardcoded preference list for ${model.provider}.`
+          : reason
+        : `Auto-selected ${model.provider}/${model.id} as a lower-ranked fallback.`,
+    ),
+  );
+
+  return { ok: true, selection: ordered[0], ordered };
 }
 
 function updateOracleUi(ctx: ExtensionContext, activeRuns: Map<string, OracleUiRun>): void {
-	if (!ctx.hasUI) return;
-	const theme = ctx.ui.theme;
-	if (activeRuns.size === 0) {
-		ctx.ui.setStatus(ORACLE_STATUS_ID, undefined);
-		ctx.ui.setWidget(ORACLE_WIDGET_ID, undefined);
-		return;
-	}
-
-	const runs = [...activeRuns.values()].sort((a, b) => b.startedAt - a.startedAt);
-	const primary = runs[0];
-	const activeCount = runs.length;
-	const modelText = primary.selection?.modelId ?? "selecting…";
-	const thinkingText = primary.selection?.thinkingLevel && primary.selection.thinkingLevel !== "off"
-		? ` ${primary.selection.thinkingLevel}`
-		: "";
-	const elapsed = formatDuration(Date.now() - primary.startedAt);
-	const mode = primary.includeBash ? "read-only+bash" : "read-only";
-	const status =
-		theme.fg("accent", "🔮 oracle") +
-		theme.fg("dim", ` ${modelText}${thinkingText} · ${mode} · ${elapsed}`) +
-		(activeCount > 1 ? theme.fg("warning", ` · ${activeCount} active`) : "");
-	ctx.ui.setStatus(ORACLE_STATUS_ID, status);
-
-	const lines = [
-		`🔮 Oracle ${activeCount > 1 ? `(${activeCount} active)` : ""}`.trim(),
-		`${primary.selection?.modelRef ?? "selecting model…"} · ${mode} · ${elapsed}`,
-		`task: ${shorten(primary.task, 110)}`,
-	];
-	if (primary.preview && primary.preview.trim()) lines.push(`preview: ${shorten(primary.preview, 110)}`);
-	ctx.ui.setWidget(ORACLE_WIDGET_ID, lines, { placement: "belowEditor" });
+  if (!ctx.hasUI) return;
+  const theme = ctx.ui.theme;
+  if (activeRuns.size === 0) {
+    ctx.ui.setStatus(ORACLE_STATUS_ID, undefined);
+    ctx.ui.setWidget(ORACLE_WIDGET_ID, undefined);
+    return;
+  }
+
+  const runs = [...activeRuns.values()].sort((a, b) => b.startedAt - a.startedAt);
+  const primary = runs[0];
+  const activeCount = runs.length;
+  const modelText = primary.selection?.modelId ?? "selecting…";
+  const thinkingText =
+    primary.selection?.thinkingLevel && primary.selection.thinkingLevel !== "off"
+      ? ` ${primary.selection.thinkingLevel}`
+      : "";
+  const elapsed = formatDuration(Date.now() - primary.startedAt);
+  const mode = primary.includeBash ? "read-only+bash" : "read-only";
+  const status =
+    theme.fg("accent", "🔮 oracle") +
+    theme.fg("dim", ` ${modelText}${thinkingText} · ${mode} · ${elapsed}`) +
+    (activeCount > 1 ? theme.fg("warning", ` · ${activeCount} active`) : "");
+  ctx.ui.setStatus(ORACLE_STATUS_ID, status);
+
+  const lines = [
+    `🔮 Oracle ${activeCount > 1 ? `(${activeCount} active)` : ""}`.trim(),
+    `${primary.selection?.modelRef ?? "selecting model…"} · ${mode} · ${elapsed}`,
+    `task: ${shorten(primary.task, 110)}`,
+  ];
+  if (primary.preview && primary.preview.trim())
+    lines.push(`preview: ${shorten(primary.preview, 110)}`);
+  ctx.ui.setWidget(ORACLE_WIDGET_ID, lines, { placement: "belowEditor" });
 }
 
 async function runOracle(
-	selection: OracleSelection,
-	params: { task: string; includeBash?: boolean; cwd?: string },
-	signal: AbortSignal | undefined,
-	onUpdate: ((result: { content: Array<{ type: "text"; text: string }>; details: OracleDetails }) => void) | undefined,
-	defaultCwd: string,
-	spawnImpl: typeof spawn = spawn,
-): Promise<{ ok: true; output: string; details: OracleDetails } | { ok: false; error: string; details: OracleDetails }> {
-	const cwd = params.cwd ?? defaultCwd;
-	const includeBash = params.includeBash ?? false;
-	const tools = includeBash ? READ_ONLY_PLUS_BASH_TOOLS : READ_ONLY_TOOLS;
-	const startedAt = Date.now();
-	const usage = createEmptyUsage();
-	let currentText = "";
-	let finalOutput = "";
-	let stderr = "";
-
-	let lastStopReason: string | undefined;
-	let lastErrorMessage: string | undefined;
-
-	const details: OracleDetails = {
-		...selection,
-		includeBash,
-		usage,
-		stderr,
-		exitCode: 0,
-		durationMs: 0,
-		cwd,
-	};
-
-	const emit = () => {
-		details.stderr = stderr;
-		details.durationMs = Date.now() - startedAt;
-		onUpdate?.({
-			content: [{ type: "text", text: currentText || finalOutput || "Consulting oracle..." }],
-			details,
-		});
-	};
-
-	const modelArg = withThinking(selection.modelRef, selection.thinkingLevel);
-	const args = [
-		"--mode",
-		"json",
-		"-p",
-		"--no-session",
-		"--model",
-		modelArg,
-		"--tools",
-		tools.join(","),
-		"--append-system-prompt",
-		ORACLE_SYSTEM_PROMPT,
-		params.task,
-	];
-
-	const invocation = getPiInvocation(args);
-	let wasAborted = false;
-
-	const exitCode = await new Promise<number>((resolve) => {
-		const proc = spawnImpl(invocation.command, invocation.args, {
-			cwd,
-			shell: false,
-			stdio: ["ignore", "pipe", "pipe"],
-		});
-		let buffer = "";
-
-		const processLine = (line: string) => {
-			if (!line.trim()) return;
-			let event: any;
-			try {
-				event = JSON.parse(line);
-			} catch {
-				return;
-			}
-
-			if (event.type === "message_start" && event.message?.role === "assistant") {
-				// A later assistant turn supersedes any earlier tool/intermediate answer.
-				// Keep only the current message so a tool-only final turn cannot return stale text.
-				currentText = "";
-				finalOutput = "";
-				emit();
-				return;
-			}
-
-			if (event.type === "message_update" && event.assistantMessageEvent?.type === "text_delta") {
-				currentText += event.assistantMessageEvent.delta ?? "";
-				emit();
-				return;
-			}
-
-			if (event.type === "message_end" && event.message?.role === "assistant") {
-				const text = extractTextFromContent(event.message.content) || currentText;
-				// message_end is authoritative, including an empty tool-only message.
-				finalOutput = text;
-				currentText = "";
-
-				const stopReason = event.message.stopReason;
-				lastStopReason = typeof stopReason === "string" ? stopReason : undefined;
-				const errorMessageField = event.message.errorMessage;
-				lastErrorMessage = typeof errorMessageField === "string" ? errorMessageField : undefined;
-
-				addUsageStats(usage, event.message.usage as Usage | undefined, {
-					countTurn: true,
-					updateContextTokens: true,
-				});
-				emit();
-				return;
-			}
-
-			if (event.type === "compaction_end") {
-				addUsageStats(usage, event.result?.usage as Usage | undefined);
-			}
-		};
-
-		proc.stdout.on("data", (data) => {
-			buffer += data.toString();
-			const lines = buffer.split("\n");
-			buffer = lines.pop() || "";
-			for (const line of lines) processLine(line);
-		});
-
-		proc.stderr.on("data", (data) => {
-			stderr += data.toString();
-			emit();
-		});
-
-		let closed = false;
-		let killTimer: ReturnType<typeof setTimeout> | undefined;
-		let abortListener: (() => void) | undefined;
-
-		const childHasExited = () => proc.exitCode !== null || proc.signalCode !== null;
-		const clearEscalation = () => {
-			if (killTimer) {
-				clearTimeout(killTimer);
-				killTimer = undefined;
-			}
-		};
-		const cleanup = () => {
-			clearEscalation();
-			if (signal && abortListener) {
-				signal.removeEventListener("abort", abortListener);
-				abortListener = undefined;
-			}
-		};
-
-		proc.on("exit", () => {
-			clearEscalation();
-		});
-
-		proc.on("close", (code) => {
-			closed = true;
-			cleanup();
-			if (buffer.trim()) processLine(buffer);
-			resolve(code ?? 0);
-		});
-
-		proc.on("error", (error) => {
-			cleanup();
-			stderr += `${error instanceof Error ? error.message : String(error)}\n`;
-			resolve(1);
-		});
-
-		if (signal) {
-			abortListener = () => {
-				wasAborted = true;
-				proc.kill("SIGTERM");
-				if (closed || childHasExited()) return;
-				killTimer = setTimeout(() => {
-					if (!closed && !childHasExited()) proc.kill("SIGKILL");
-				}, 5000);
-			};
-			if (signal.aborted) abortListener();
-			else signal.addEventListener("abort", abortListener, { once: true });
-		}
-	});
-
-	details.stderr = stderr.trim();
-	details.exitCode = exitCode;
-	details.durationMs = Date.now() - startedAt;
-
-	if (wasAborted) {
-		if (details.exitCode === 0) details.exitCode = 1;
-		return { ok: false, error: "Oracle was aborted.", details };
-	}
-
-	// The pi subprocess in `--mode json` does not promote an errored assistant
-	// turn to a non-zero exit code or stderr; the error is only carried on the
-	// streamed assistant message via stopReason/errorMessage. Surface that here
-	// so callers can distinguish transient provider errors from a genuinely
-	// empty response.
-	if (lastStopReason === "error" || lastStopReason === "aborted") {
-		if (details.exitCode === 0) details.exitCode = 1;
-		return {
-			ok: false,
-			error: formatOracleModelError(lastStopReason, lastErrorMessage),
-			details,
-		};
-	}
-
-	if (exitCode !== 0) {
-		return {
-			ok: false,
-			error: details.stderr || finalOutput || "Oracle subprocess failed.",
-			details,
-		};
-	}
-
-	if (!finalOutput.trim()) {
-		if (details.exitCode === 0) details.exitCode = 1;
-		return {
-			ok: false,
-			error: details.stderr || "Oracle finished without returning any text.",
-			details,
-		};
-	}
-
-	return {
-		ok: true,
-		output: finalOutput.trim(),
-		details,
-	};
+  selection: OracleSelection,
+  params: { task: string; includeBash?: boolean; cwd?: string },
+  signal: AbortSignal | undefined,
+  onUpdate:
+    | ((result: { content: Array<{ type: "text"; text: string }>; details: OracleDetails }) => void)
+    | undefined,
+  defaultCwd: string,
+  spawnImpl: typeof spawn = spawn,
+): Promise<
+  | { ok: true; output: string; details: OracleDetails }
+  | { ok: false; error: string; details: OracleDetails }
+> {
+  const cwd = params.cwd ?? defaultCwd;
+  const includeBash = params.includeBash ?? false;
+  const tools = includeBash ? READ_ONLY_PLUS_BASH_TOOLS : READ_ONLY_TOOLS;
+  const startedAt = Date.now();
+  const usage = createEmptyUsage();
+  let currentText = "";
+  let finalOutput = "";
+  let stderr = "";
+
+  let lastStopReason: string | undefined;
+  let lastErrorMessage: string | undefined;
+
+  const details: OracleDetails = {
+    ...selection,
+    includeBash,
+    usage,
+    stderr,
+    exitCode: 0,
+    durationMs: 0,
+    cwd,
+  };
+
+  const emit = () => {
+    details.stderr = stderr;
+    details.durationMs = Date.now() - startedAt;
+    onUpdate?.({
+      content: [{ type: "text", text: currentText || finalOutput || "Consulting oracle..." }],
+      details,
+    });
+  };
+
+  const modelArg = withThinking(selection.modelRef, selection.thinkingLevel);
+  const args = [
+    "--mode",
+    "json",
+    "-p",
+    "--no-session",
+    "--model",
+    modelArg,
+    "--tools",
+    tools.join(","),
+    "--append-system-prompt",
+    ORACLE_SYSTEM_PROMPT,
+    params.task,
+  ];
+
+  const invocation = getPiInvocation(args);
+  let wasAborted = false;
+
+  const exitCode = await new Promise<number>((resolve) => {
+    const proc = spawnImpl(invocation.command, invocation.args, {
+      cwd,
+      shell: false,
+      stdio: ["ignore", "pipe", "pipe"],
+    });
+    let buffer = "";
+
+    const processLine = (line: string) => {
+      if (!line.trim()) return;
+      let event: any;
+      try {
+        event = JSON.parse(line);
+      } catch {
+        return;
+      }
+
+      if (event.type === "message_start" && event.message?.role === "assistant") {
+        // A later assistant turn supersedes any earlier tool/intermediate answer.
+        // Keep only the current message so a tool-only final turn cannot return stale text.
+        currentText = "";
+        finalOutput = "";
+        emit();
+        return;
+      }
+
+      if (event.type === "message_update" && event.assistantMessageEvent?.type === "text_delta") {
+        currentText += event.assistantMessageEvent.delta ?? "";
+        emit();
+        return;
+      }
+
+      if (event.type === "message_end" && event.message?.role === "assistant") {
+        const text = extractTextFromContent(event.message.content) || currentText;
+        // message_end is authoritative, including an empty tool-only message.
+        finalOutput = text;
+        currentText = "";
+
+        const stopReason = event.message.stopReason;
+        lastStopReason = typeof stopReason === "string" ? stopReason : undefined;
+        const errorMessageField = event.message.errorMessage;
+        lastErrorMessage = typeof errorMessageField === "string" ? errorMessageField : undefined;
+
+        addUsageStats(usage, event.message.usage as Usage | undefined, {
+          countTurn: true,
+          updateContextTokens: true,
+        });
+        emit();
+        return;
+      }
+
+      if (event.type === "compaction_end") {
+        addUsageStats(usage, event.result?.usage as Usage | undefined);
+      }
+    };
+
+    proc.stdout.on("data", (data) => {
+      buffer += data.toString();
+      const lines = buffer.split("\n");
+      buffer = lines.pop() || "";
+      for (const line of lines) processLine(line);
+    });
+
+    proc.stderr.on("data", (data) => {
+      stderr += data.toString();
+      emit();
+    });
+
+    let closed = false;
+    let killTimer: ReturnType<typeof setTimeout> | undefined;
+    let abortListener: (() => void) | undefined;
+
+    const childHasExited = () => proc.exitCode !== null || proc.signalCode !== null;
+    const clearEscalation = () => {
+      if (killTimer) {
+        clearTimeout(killTimer);
+        killTimer = undefined;
+      }
+    };
+    const cleanup = () => {
+      clearEscalation();
+      if (signal && abortListener) {
+        signal.removeEventListener("abort", abortListener);
+        abortListener = undefined;
+      }
+    };
+
+    proc.on("exit", () => {
+      clearEscalation();
+    });
+
+    proc.on("close", (code) => {
+      closed = true;
+      cleanup();
+      if (buffer.trim()) processLine(buffer);
+      resolve(code ?? 0);
+    });
+
+    proc.on("error", (error) => {
+      cleanup();
+      stderr += `${error instanceof Error ? error.message : String(error)}\n`;
+      resolve(1);
+    });
+
+    if (signal) {
+      abortListener = () => {
+        wasAborted = true;
+        proc.kill("SIGTERM");
+        if (closed || childHasExited()) return;
+        killTimer = setTimeout(() => {
+          if (!closed && !childHasExited()) proc.kill("SIGKILL");
+        }, 5000);
+      };
+      if (signal.aborted) abortListener();
+      else signal.addEventListener("abort", abortListener, { once: true });
+    }
+  });
+
+  details.stderr = stderr.trim();
+  details.exitCode = exitCode;
+  details.durationMs = Date.now() - startedAt;
+
+  if (wasAborted) {
+    if (details.exitCode === 0) details.exitCode = 1;
+    return { ok: false, error: "Oracle was aborted.", details };
+  }
+
+  // The pi subprocess in `--mode json` does not promote an errored assistant
+  // turn to a non-zero exit code or stderr; the error is only carried on the
+  // streamed assistant message via stopReason/errorMessage. Surface that here
+  // so callers can distinguish transient provider errors from a genuinely
+  // empty response.
+  if (lastStopReason === "error" || lastStopReason === "aborted") {
+    if (details.exitCode === 0) details.exitCode = 1;
+    return {
+      ok: false,
+      error: formatOracleModelError(lastStopReason, lastErrorMessage),
+      details,
+    };
+  }
+
+  if (exitCode !== 0) {
+    return {
+      ok: false,
+      error: details.stderr || finalOutput || "Oracle subprocess failed.",
+      details,
+    };
+  }
+
+  if (!finalOutput.trim()) {
+    if (details.exitCode === 0) details.exitCode = 1;
+    return {
+      ok: false,
+      error: details.stderr || "Oracle finished without returning any text.",
+      details,
+    };
+  }
+
+  return {
+    ok: true,
+    output: finalOutput.trim(),
+    details,
+  };
 }
 
 function renderCollapsedText(text: string, lineLimit = COLLAPSED_LINE_LIMIT): string {
-	const lines = text.trim().split("\n");
-	if (lines.length <= lineLimit) return lines.join("\n");
-	return [...lines.slice(0, lineLimit), `... (${lines.length - lineLimit} more lines)`].join("\n");
+  const lines = text.trim().split("\n");
+  if (lines.length <= lineLimit) return lines.join("\n");
+  return [...lines.slice(0, lineLimit), `... (${lines.length - lineLimit} more lines)`].join("\n");
 }
 
 type OracleExtensionDeps = {
-	spawnImpl?: typeof spawn;
+  spawnImpl?: typeof spawn;
 };
 
 export const __test__ = {
-	buildSessionFallbackSelection,
-	createOracleExtension,
-	findAvailableModel,
-	isModelAvailabilityError,
-	parseModelPreference,
-	runOracle,
-	resolveThinkingLevel,
-	selectOracleModel,
+  buildSessionFallbackSelection,
+  createOracleExtension,
+  findAvailableModel,
+  isModelAvailabilityError,
+  parseModelPreference,
+  runOracle,
+  resolveThinkingLevel,
+  selectOracleModel,
 };
 
 export default function oracleExtension(pi: ExtensionAPI) {
-	createOracleExtension(pi);
+  createOracleExtension(pi);
 }
 
 function createOracleExtension(pi: ExtensionAPI, deps: OracleExtensionDeps = {}) {
-	const spawnImpl = deps.spawnImpl ?? spawn;
-	const activeRuns = new Map<string, OracleUiRun>();
-	let preferences: OraclePreferences = {};
-
-	pi.on("session_start", async (_event, ctx) => {
-		preferences = await readOraclePreferences();
-		activeRuns.clear();
-		updateOracleUi(ctx, activeRuns);
-	});
-
-	pi.registerCommand("oracle", {
-		description: "Configure Oracle default model and thinking level for future oracle tool calls",
-		getArgumentCompletions: (prefix) => {
-			const parts = prefix.trim().toLowerCase().split(/\s+/).filter(Boolean);
-			const first = parts[0] ?? "";
-			if (parts.length <= 1) {
-				const commands = ["status", "model", "thinking", "clear"];
-				const matches = commands.filter((command) => command.startsWith(first));
-				return matches.length > 0 ? matches.map((value) => ({ value, label: value })) : null;
-			}
-			if (parts[0] === "thinking") {
-				const query = parts[1] ?? "";
-				const values = ["auto", ...THINKING_LEVELS];
-				const matches = values.filter((value) => value.startsWith(query));
-				return matches.length > 0 ? matches.map((value) => ({ value, label: value })) : null;
-			}
-			if (parts[0] === "clear") {
-				const query = parts[1] ?? "";
-				const values = ["all", "model", "thinking"];
-				const matches = values.filter((value) => value.startsWith(query));
-				return matches.length > 0 ? matches.map((value) => ({ value, label: value })) : null;
-			}
-			return null;
-		},
-		handler: async (args, ctx) => {
-			const raw = args.trim();
-			const tokens = raw ? raw.split(/\s+/) : [];
-			const [command = "status", ...rest] = tokens;
-			const action = command.toLowerCase();
-			const configPath = getOracleConfigPath();
-
-			const save = async (next: OraclePreferences): Promise<string | undefined> => {
-				preferences = next;
-				try {
-					await writeOraclePreferences(preferences);
-					return undefined;
-				} catch (error) {
-					const message = error instanceof Error ? error.message : String(error);
-					return `Preference changed for this process, but could not save ${configPath}: ${message}`;
-				}
-			};
-
-			if (action === "status" || action === "show") {
-				notifyCommand(ctx, formatOraclePreferences(preferences));
-				return;
-			}
-
-			if (action === "model") {
-				const model = rest.join(" ").trim();
-				const normalizedModelAction = model.toLowerCase();
-				if (!model) {
-					notifyCommand(ctx, "Usage: /oracle model <provider/model|auto>", "warning");
-					return;
-				}
-				if (normalizedModelAction === "auto" || normalizedModelAction === "clear" || normalizedModelAction === "default") {
-					const warning = await save({ ...preferences, model: undefined });
-					notifyCommand(ctx, `Oracle default model cleared; future oracle calls will auto-select. ${warning ?? formatOraclePreferences(preferences)}`, warning ? "warning" : "info");
-					return;
-				}
-				const parsedModel = parseModelPreference(model);
-				const next = {
-					...preferences,
-					model: parsedModel.model,
-					thinkingLevel: parsedModel.thinkingLevel ?? preferences.thinkingLevel,
-				};
-				const warning = await save(next);
-				notifyCommand(ctx, `Oracle default model set to ${parsedModel.model}. ${warning ?? formatOraclePreferences(preferences)}`, warning ? "warning" : "info");
-				return;
-			}
-
-			if (action === "thinking" || action === "think" || action === "thinking-level") {
-				const value = rest[0]?.trim().toLowerCase();
-				if (!value) {
-					notifyCommand(ctx, `Usage: /oracle thinking ${THINKING_LEVELS.join(" | ")} | auto`, "warning");
-					return;
-				}
-				if (value === "auto" || value === "clear" || value === "default") {
-					const warning = await save({ ...preferences, thinkingLevel: undefined });
-					notifyCommand(ctx, `Oracle default thinking level cleared; future oracle calls will use built-in defaults. ${warning ?? formatOraclePreferences(preferences)}`, warning ? "warning" : "info");
-					return;
-				}
-				const thinkingLevel = parseThinkingLevel(value);
-				if (!thinkingLevel) {
-					notifyCommand(ctx, `Usage: /oracle thinking ${THINKING_LEVELS.join(" | ")} | auto`, "warning");
-					return;
-				}
-				const warning = await save({ ...preferences, thinkingLevel });
-				notifyCommand(ctx, `Oracle default thinking level set to ${thinkingLevel}. ${warning ?? formatOraclePreferences(preferences)}`, warning ? "warning" : "info");
-				return;
-			}
-
-			if (action === "clear" || action === "reset") {
-				const target = rest[0]?.trim().toLowerCase() || "all";
-				let next: OraclePreferences;
-				if (target === "all") next = {};
-				else if (target === "model") next = { ...preferences, model: undefined };
-				else if (target === "thinking" || target === "thinking-level") next = { ...preferences, thinkingLevel: undefined };
-				else {
-					notifyCommand(ctx, "Usage: /oracle clear [all|model|thinking]", "warning");
-					return;
-				}
-				const warning = await save(next);
-				notifyCommand(ctx, `Oracle defaults cleared (${target}). ${warning ?? formatOraclePreferences(preferences)}`, warning ? "warning" : "info");
-				return;
-			}
-
-			notifyCommand(ctx, "Usage: /oracle status | model <provider/model|auto> | thinking <off|minimal|low|medium|high|xhigh|max|auto> | clear [all|model|thinking]", "warning");
-		},
-	});
-
-	pi.registerCommand("oracle-model", {
-		description: "Show which model the oracle would use right now",
-		handler: async (_args, ctx) => {
-			const defaultModel = parseModelPreference(preferences.model);
-			if (defaultModel.model) {
-				const matched = await findAvailableModel(ctx, defaultModel.model);
-				const thinking = resolveThinkingLevel(matched, preferences.thinkingLevel ?? defaultModel.thinkingLevel);
-				const modelRef = matched ? `${matched.provider}/${matched.id}` : defaultModel.model;
-				const suffix = matched
-					? appendThinkingLevelClampReason("Configured default oracle model is active.", thinking)
-					: "Configured default oracle model is active, but it was not matched against the authenticated model list.";
-				notifyCommand(ctx, `Oracle: ${modelRef} (${thinking.effective}) — ${suffix}`);
-				return;
-			}
-			const selectionResult = await selectOracleModel(ctx, preferences.thinkingLevel);
-			if (!selectionResult.ok) {
-				notifyCommand(ctx, selectionResult.error, "error");
-				return;
-			}
-
-			const { selection } = selectionResult;
-			const message = `Oracle: ${selection.modelRef} (${selection.thinkingLevel}) — ${selection.selectionReason}`;
-			notifyCommand(ctx, message);
-		},
-	});
-
-	pi.on("tool_result", async (event) => {
-		if (event.toolName !== "oracle") return undefined;
-		const details = event.details as OracleDetails | undefined;
-		if ((details?.exitCode ?? 0) === 0) return undefined;
-		return { isError: true };
-	});
-
-	pi.registerTool({
-		name: "oracle",
-		label: "Oracle",
-		description:
-			"Consult a separate read-only oracle subprocess for deep analysis, code review, debugging, planning, and second opinions.",
-		promptSnippet:
-			"Consult a read-only oracle that auto-selects the strongest reasoning model on the current provider/subscription.",
-		promptGuidelines: [
-			"Use oracle sparingly when you want a second opinion, deeper analysis, code review, debugging help, or a higher-reasoning pass.",
-			"Do not use oracle for routine low-value work; oracle is slower than the main agent.",
-			"The oracle tool is read-only by default and only exposes read, grep, find, and ls unless oracle includeBash is enabled.",
-			"Set oracle includeBash only when the extra bash inspection tool is genuinely useful; keep oracle relying on read, grep, find, and ls otherwise.",
-			"The oracle tool requests xhigh by default for reasoning models; oracle defaults and explicit thinkingLevel overrides are clamped to the effective model-supported level when the model is matched.",
-		],
-		parameters: OracleParams,
-
-		async execute(toolCallId, params, signal, onUpdate, ctx) {
-			const rawTask = (params as { task?: unknown }).task;
-			const task = typeof rawTask === "string" ? rawTask.trim() : "";
-			if (!task) throw new Error("Invalid parameters: expected task to be a non-empty string.");
-
-			const explicitModel = parseModelPreference(params.model);
-			const defaultModel = parseModelPreference(preferences.model);
-			const configuredModel = explicitModel.model ?? defaultModel.model;
-			const configuredThinkingLevel =
-				params.thinkingLevel ?? explicitModel.thinkingLevel ?? preferences.thinkingLevel ?? defaultModel.thinkingLevel;
-			const uiRun: OracleUiRun = {
-				task,
-				includeBash: params.includeBash ?? false,
-				startedAt: Date.now(),
-			};
-			activeRuns.set(toolCallId, uiRun);
-			updateOracleUi(ctx, activeRuns);
-
-			try {
-				const attempts: OracleSelection[] = [];
-				const seen = new Set<string>();
-				const aggregateUsage = createEmptyUsage();
-				const finalizeResult = (
-					result: { content: Array<{ type: "text"; text: string }>; details: OracleDetails },
-				): AgentToolResult<OracleDetails> => ({
-					...result,
-					details: { ...result.details, usage: cloneUsageStats(aggregateUsage) },
-					usage: usageStatsToToolUsage(aggregateUsage),
-				});
-				const pushAttempt = (candidate: OracleSelection | undefined) => {
-					if (!candidate) return;
-					const key = `${candidate.provider}/${candidate.modelId}`.toLowerCase();
-					if (seen.has(key)) return;
-					seen.add(key);
-					attempts.push(candidate);
-				};
-
-				if (configuredModel) {
-					const modelRef = configuredModel;
-					const matched = await findAvailableModel(ctx, modelRef);
-					const provider =
-						matched?.provider ?? (modelRef.includes("/") ? modelRef.split("/")[0] : ctx.model?.provider ?? "unknown");
-					const modelId = matched?.id ?? (modelRef.includes("/") ? modelRef.split("/").slice(1).join("/") : modelRef);
-					const thinking = resolveThinkingLevel(matched, configuredThinkingLevel);
-					const usedToolOverride = !!explicitModel.model;
-					const selectionReason = matched
-						? appendThinkingLevelClampReason(
-							usedToolOverride
-								? "Used the explicit model override provided in the tool call."
-								: "Used the configured default oracle model.",
-							thinking,
-						)
-						: usedToolOverride
-							? "Used the explicit model override provided in the tool call. The model was not matched against the authenticated model list, so the reasoning level fallback was applied."
-							: "Used the configured default oracle model. The model was not matched against the authenticated model list, so the reasoning level fallback was applied.";
-					pushAttempt({
-						modelRef: matched ? `${matched.provider}/${matched.id}` : modelRef,
-						provider,
-						modelId,
-						modelName: matched?.name,
-						thinkingLevel: thinking.effective,
-						...(matched && thinking.clamped
-							? { requestedThinkingLevel: thinking.requested, thinkingLevelClamped: true }
-							: {}),
-						autoSelected: false,
-						selectionReason,
-					});
-				} else {
-					const selectionResult = await selectOracleModel(ctx, configuredThinkingLevel);
-					if (!selectionResult.ok) {
-						return finalizeResult({
-							content: [{ type: "text", text: selectionResult.error }],
-							details: {
-								modelRef: "",
-								provider: ctx.model?.provider ?? "unknown",
-								modelId: "",
-								modelName: undefined,
-								thinkingLevel: configuredThinkingLevel ?? DEFAULT_THINKING_LEVEL,
-								autoSelected: true,
-								selectionReason: selectionResult.error,
-								includeBash: params.includeBash ?? false,
-								usage: createEmptyUsage(),
-								stderr: "",
-								exitCode: 1,
-								durationMs: 0,
-								cwd: params.cwd ?? ctx.cwd,
-							},
-						});
-					}
-					for (const candidate of selectionResult.ordered) pushAttempt(candidate);
-				}
-
-				// Keep the known-good session model as a final fallback: the catalog can
-				// advertise models the active provider/subscription cannot actually serve
-				// (legacy snapshots, access-gated tiers), which fail with a not-found/404
-				// error. Falling back lets the run degrade gracefully instead of hard-failing.
-				pushAttempt(buildSessionFallbackSelection(ctx, configuredThinkingLevel));
-
-				const handleUpdate = (partial: { content: Array<{ type: "text"; text: string }>; details: OracleDetails }) => {
-					uiRun.preview = partial.content[0]?.text ?? uiRun.preview;
-					updateOracleUi(ctx, activeRuns);
-					onUpdate?.(partial);
-				};
-
-				let lastErrorResult: { ok: false; error: string; details: OracleDetails } | undefined;
-				for (let index = 0; index < attempts.length; index++) {
-					const previous = index > 0 ? attempts[index - 1] : undefined;
-					const attempt = previous ? withFallbackReason(attempts[index], previous) : attempts[index];
-					uiRun.selection = attempt;
-					updateOracleUi(ctx, activeRuns);
-
-					const result = await runOracle(attempt, { ...params, task }, signal, handleUpdate, ctx.cwd, spawnImpl);
-					mergeUsageStats(aggregateUsage, result.details.usage);
-					if (result.ok) {
-						return finalizeResult({
-							content: [{ type: "text", text: result.output }],
-							details: result.details,
-						});
-					}
-
-					lastErrorResult = result;
-					const canFallBack = index < attempts.length - 1 && isModelAvailabilityError(result.error);
-					if (!canFallBack) {
-						return finalizeResult({
-							content: [{ type: "text", text: result.error }],
-							details: result.details,
-						});
-					}
-				}
-
-				return finalizeResult({
-					content: [
-						{ type: "text", text: lastErrorResult?.error ?? "Oracle could not select an available model." },
-					],
-					details:
-						lastErrorResult?.details ?? {
-							modelRef: "",
-							provider: ctx.model?.provider ?? "unknown",
-							modelId: "",
-							modelName: undefined,
-							thinkingLevel: configuredThinkingLevel ?? DEFAULT_THINKING_LEVEL,
-							autoSelected: true,
-							selectionReason: "No authenticated models were available to run the oracle.",
-							includeBash: params.includeBash ?? false,
-							usage: createEmptyUsage(),
-							stderr: "",
-							exitCode: 1,
-							durationMs: 0,
-							cwd: params.cwd ?? ctx.cwd,
-						},
-				});
-			} finally {
-				activeRuns.delete(toolCallId);
-				updateOracleUi(ctx, activeRuns);
-			}
-		},
-
-		renderCall(args, theme) {
-			const task = args.task.length > 90 ? `${args.task.slice(0, 90)}...` : args.task;
-			const mode = args.includeBash ? "read-only+bash" : "read-only";
-			const override = args.model ? ` ${theme.fg("muted", `[${args.model}]`)}` : "";
-			const thinking = args.thinkingLevel ? ` ${theme.fg("warning", `(${args.thinkingLevel})`)}` : "";
-			return new Text(
-				`${theme.fg("toolTitle", theme.bold("oracle "))}${theme.fg("accent", mode)}${override}${thinking}\n  ${theme.fg("dim", task)}`,
-				0,
-				0,
-			);
-		},
-
-		renderResult(result, { expanded }, theme) {
-			const details = result.details as OracleDetails | undefined;
-			const body = result.content[0]?.type === "text" ? result.content[0].text : "(no output)";
-			if (!details) return new Text(body, 0, 0);
-
-			const isError = (details.exitCode ?? 0) !== 0;
-			const icon = isError ? theme.fg("error", "✗") : theme.fg("success", "✓");
-			const header = `${icon} ${theme.fg("toolTitle", theme.bold("oracle "))}${theme.fg("accent", details.modelRef || "(auto)")}`;
-			const subheader = [
-				details.thinkingLevel !== "off" ? details.thinkingLevel : undefined,
-				details.includeBash ? "read-only+bash" : "read-only",
-				formatDuration(details.durationMs),
-			]
-				.filter(Boolean)
-				.join(" · ");
-			const usage = formatUsage(details.usage);
-
-			if (!expanded) {
-				let text = header;
-				if (subheader) text += `\n${theme.fg("dim", subheader)}`;
-				text += `\n\n${theme.fg("toolOutput", renderCollapsedText(body))}`;
-				if (usage) text += `\n\n${theme.fg("dim", usage)}`;
-				if (isError && details.stderr) text += `\n${theme.fg("error", renderCollapsedText(details.stderr, 4))}`;
-				text += `\n${theme.fg("muted", "(Ctrl+O to expand)")}`;
-				return new Text(text, 0, 0);
-			}
-
-			const container = new Container();
-			container.addChild(new Text(header, 0, 0));
-			if (subheader) container.addChild(new Text(theme.fg("dim", subheader), 0, 0));
-			container.addChild(new Spacer(1));
-			container.addChild(new Text(theme.fg("muted", "Selection"), 0, 0));
-			container.addChild(new Text(theme.fg("dim", details.selectionReason), 0, 0));
-			container.addChild(new Spacer(1));
-			container.addChild(new Text(theme.fg("muted", "Output"), 0, 0));
-			container.addChild(new Markdown(body.trim(), 0, 0, getMarkdownTheme()));
-			if (usage) {
-				container.addChild(new Spacer(1));
-				container.addChild(new Text(theme.fg("muted", "Usage"), 0, 0));
-				container.addChild(new Text(theme.fg("dim", usage), 0, 0));
-			}
-			if (details.stderr) {
-				container.addChild(new Spacer(1));
-				container.addChild(new Text(theme.fg("muted", "stderr"), 0, 0));
-				container.addChild(
-					new Text(isError ? theme.fg("error", details.stderr) : theme.fg("dim", details.stderr), 0, 0),
-				);
-			}
-			return container;
-		},
-	});
+  const spawnImpl = deps.spawnImpl ?? spawn;
+  const activeRuns = new Map<string, OracleUiRun>();
+  let preferences: OraclePreferences = {};
+
+  pi.on("session_start", async (_event, ctx) => {
+    preferences = await readOraclePreferences();
+    activeRuns.clear();
+    updateOracleUi(ctx, activeRuns);
+  });
+
+  pi.registerCommand("oracle", {
+    description: "Configure Oracle default model and thinking level for future oracle tool calls",
+    getArgumentCompletions: (prefix) => {
+      const parts = prefix.trim().toLowerCase().split(/\s+/).filter(Boolean);
+      const first = parts[0] ?? "";
+      if (parts.length <= 1) {
+        const commands = ["status", "model", "thinking", "clear"];
+        const matches = commands.filter((command) => command.startsWith(first));
+        return matches.length > 0 ? matches.map((value) => ({ value, label: value })) : null;
+      }
+      if (parts[0] === "thinking") {
+        const query = parts[1] ?? "";
+        const values = ["auto", ...THINKING_LEVELS];
+        const matches = values.filter((value) => value.startsWith(query));
+        return matches.length > 0 ? matches.map((value) => ({ value, label: value })) : null;
+      }
+      if (parts[0] === "clear") {
+        const query = parts[1] ?? "";
+        const values = ["all", "model", "thinking"];
+        const matches = values.filter((value) => value.startsWith(query));
+        return matches.length > 0 ? matches.map((value) => ({ value, label: value })) : null;
+      }
+      return null;
+    },
+    handler: async (args, ctx) => {
+      const raw = args.trim();
+      const tokens = raw ? raw.split(/\s+/) : [];
+      const [command = "status", ...rest] = tokens;
+      const action = command.toLowerCase();
+      const configPath = getOracleConfigPath();
+
+      const save = async (next: OraclePreferences): Promise<string | undefined> => {
+        preferences = next;
+        try {
+          await writeOraclePreferences(preferences);
+          return undefined;
+        } catch (error) {
+          const message = error instanceof Error ? error.message : String(error);
+          return `Preference changed for this process, but could not save ${configPath}: ${message}`;
+        }
+      };
+
+      if (action === "status" || action === "show") {
+        notifyCommand(ctx, formatOraclePreferences(preferences));
+        return;
+      }
+
+      if (action === "model") {
+        const model = rest.join(" ").trim();
+        const normalizedModelAction = model.toLowerCase();
+        if (!model) {
+          notifyCommand(ctx, "Usage: /oracle model <provider/model|auto>", "warning");
+          return;
+        }
+        if (
+          normalizedModelAction === "auto" ||
+          normalizedModelAction === "clear" ||
+          normalizedModelAction === "default"
+        ) {
+          const warning = await save({ ...preferences, model: undefined });
+          notifyCommand(
+            ctx,
+            `Oracle default model cleared; future oracle calls will auto-select. ${warning ?? formatOraclePreferences(preferences)}`,
+            warning ? "warning" : "info",
+          );
+          return;
+        }
+        const parsedModel = parseModelPreference(model);
+        const next = {
+          ...preferences,
+          model: parsedModel.model,
+          thinkingLevel: parsedModel.thinkingLevel ?? preferences.thinkingLevel,
+        };
+        const warning = await save(next);
+        notifyCommand(
+          ctx,
+          `Oracle default model set to ${parsedModel.model}. ${warning ?? formatOraclePreferences(preferences)}`,
+          warning ? "warning" : "info",
+        );
+        return;
+      }
+
+      if (action === "thinking" || action === "think" || action === "thinking-level") {
+        const value = rest[0]?.trim().toLowerCase();
+        if (!value) {
+          notifyCommand(
+            ctx,
+            `Usage: /oracle thinking ${THINKING_LEVELS.join(" | ")} | auto`,
+            "warning",
+          );
+          return;
+        }
+        if (value === "auto" || value === "clear" || value === "default") {
+          const warning = await save({ ...preferences, thinkingLevel: undefined });
+          notifyCommand(
+            ctx,
+            `Oracle default thinking level cleared; future oracle calls will use built-in defaults. ${warning ?? formatOraclePreferences(preferences)}`,
+            warning ? "warning" : "info",
+          );
+          return;
+        }
+        const thinkingLevel = parseThinkingLevel(value);
+        if (!thinkingLevel) {
+          notifyCommand(
+            ctx,
+            `Usage: /oracle thinking ${THINKING_LEVELS.join(" | ")} | auto`,
+            "warning",
+          );
+          return;
+        }
+        const warning = await save({ ...preferences, thinkingLevel });
+        notifyCommand(
+          ctx,
+          `Oracle default thinking level set to ${thinkingLevel}. ${warning ?? formatOraclePreferences(preferences)}`,
+          warning ? "warning" : "info",
+        );
+        return;
+      }
+
+      if (action === "clear" || action === "reset") {
+        const target = rest[0]?.trim().toLowerCase() || "all";
+        let next: OraclePreferences;
+        if (target === "all") next = {};
+        else if (target === "model") next = { ...preferences, model: undefined };
+        else if (target === "thinking" || target === "thinking-level")
+          next = { ...preferences, thinkingLevel: undefined };
+        else {
+          notifyCommand(ctx, "Usage: /oracle clear [all|model|thinking]", "warning");
+          return;
+        }
+        const warning = await save(next);
+        notifyCommand(
+          ctx,
+          `Oracle defaults cleared (${target}). ${warning ?? formatOraclePreferences(preferences)}`,
+          warning ? "warning" : "info",
+        );
+        return;
+      }
+
+      notifyCommand(
+        ctx,
+        "Usage: /oracle status | model <provider/model|auto> | thinking <off|minimal|low|medium|high|xhigh|max|auto> | clear [all|model|thinking]",
+        "warning",
+      );
+    },
+  });
+
+  pi.registerCommand("oracle-model", {
+    description: "Show which model the oracle would use right now",
+    handler: async (_args, ctx) => {
+      const defaultModel = parseModelPreference(preferences.model);
+      if (defaultModel.model) {
+        const matched = await findAvailableModel(ctx, defaultModel.model);
+        const thinking = resolveThinkingLevel(
+          matched,
+          preferences.thinkingLevel ?? defaultModel.thinkingLevel,
+        );
+        const modelRef = matched ? `${matched.provider}/${matched.id}` : defaultModel.model;
+        const suffix = matched
+          ? appendThinkingLevelClampReason("Configured default oracle model is active.", thinking)
+          : "Configured default oracle model is active, but it was not matched against the authenticated model list.";
+        notifyCommand(ctx, `Oracle: ${modelRef} (${thinking.effective}) — ${suffix}`);
+        return;
+      }
+      const selectionResult = await selectOracleModel(ctx, preferences.thinkingLevel);
+      if (!selectionResult.ok) {
+        notifyCommand(ctx, selectionResult.error, "error");
+        return;
+      }
+
+      const { selection } = selectionResult;
+      const message = `Oracle: ${selection.modelRef} (${selection.thinkingLevel}) — ${selection.selectionReason}`;
+      notifyCommand(ctx, message);
+    },
+  });
+
+  pi.on("tool_result", async (event) => {
+    if (event.toolName !== "oracle") return undefined;
+    const details = event.details as OracleDetails | undefined;
+    if ((details?.exitCode ?? 0) === 0) return undefined;
+    return { isError: true };
+  });
+
+  pi.registerTool({
+    name: "oracle",
+    label: "Oracle",
+    description:
+      "Consult a separate oracle subprocess with read-only built-in tools and configured MCP tools for deep analysis, code review, debugging, planning, and second opinions.",
+    promptSnippet:
+      "Consult an oracle subprocess that auto-selects the strongest reasoning model on the current provider/subscription; exposes read-only built-in tools plus any configured MCP tools.",
+    promptGuidelines: [
+      "Use oracle sparingly when you want a second opinion, deeper analysis, code review, debugging help, or a higher-reasoning pass.",
+      "Do not use oracle for routine low-value work; oracle is slower than the main agent.",
+      "The oracle tool exposes read-only built-in tools (read, grep, find, and ls) plus any configured MCP tools; oracle includeBash adds the bash inspection tool.",
+      "Set oracle includeBash only when the extra bash inspection tool is genuinely useful; keep oracle relying on read, grep, find, and ls otherwise.",
+      "Configured MCP tools are available to oracle and run without confirmation inside the subagent; they may have side effects — use them only for gathering evidence, not for state-changing actions unless the task explicitly asks. MCP access requires host Pi >=1.0.4; older hosts degrade to no MCP, no error.",
+      "The oracle tool requests xhigh by default for reasoning models; oracle defaults and explicit thinkingLevel overrides are clamped to the effective model-supported level when the model is matched.",
+    ],
+    parameters: OracleParams,
+
+    async execute(toolCallId, params, signal, onUpdate, ctx) {
+      const rawTask = (params as { task?: unknown }).task;
+      const task = typeof rawTask === "string" ? rawTask.trim() : "";
+      if (!task) throw new Error("Invalid parameters: expected task to be a non-empty string.");
+
+      const explicitModel = parseModelPreference(params.model);
+      const defaultModel = parseModelPreference(preferences.model);
+      const configuredModel = explicitModel.model ?? defaultModel.model;
+      const configuredThinkingLevel =
+        params.thinkingLevel ??
+        explicitModel.thinkingLevel ??
+        preferences.thinkingLevel ??
+        defaultModel.thinkingLevel;
+      const uiRun: OracleUiRun = {
+        task,
+        includeBash: params.includeBash ?? false,
+        startedAt: Date.now(),
+      };
+      activeRuns.set(toolCallId, uiRun);
+      updateOracleUi(ctx, activeRuns);
+
+      try {
+        const attempts: OracleSelection[] = [];
+        const seen = new Set<string>();
+        const aggregateUsage = createEmptyUsage();
+        const finalizeResult = (result: {
+          content: Array<{ type: "text"; text: string }>;
+          details: OracleDetails;
+        }): AgentToolResult<OracleDetails> => ({
+          ...result,
+          details: { ...result.details, usage: cloneUsageStats(aggregateUsage) },
+          usage: usageStatsToToolUsage(aggregateUsage),
+        });
+        const pushAttempt = (candidate: OracleSelection | undefined) => {
+          if (!candidate) return;
+          const key = `${candidate.provider}/${candidate.modelId}`.toLowerCase();
+          if (seen.has(key)) return;
+          seen.add(key);
+          attempts.push(candidate);
+        };
+
+        if (configuredModel) {
+          const modelRef = configuredModel;
+          const matched = await findAvailableModel(ctx, modelRef);
+          const provider =
+            matched?.provider ??
+            (modelRef.includes("/") ? modelRef.split("/")[0] : (ctx.model?.provider ?? "unknown"));
+          const modelId =
+            matched?.id ??
+            (modelRef.includes("/") ? modelRef.split("/").slice(1).join("/") : modelRef);
+          const thinking = resolveThinkingLevel(matched, configuredThinkingLevel);
+          const usedToolOverride = !!explicitModel.model;
+          const selectionReason = matched
+            ? appendThinkingLevelClampReason(
+                usedToolOverride
+                  ? "Used the explicit model override provided in the tool call."
+                  : "Used the configured default oracle model.",
+                thinking,
+              )
+            : usedToolOverride
+              ? "Used the explicit model override provided in the tool call. The model was not matched against the authenticated model list, so the reasoning level fallback was applied."
+              : "Used the configured default oracle model. The model was not matched against the authenticated model list, so the reasoning level fallback was applied.";
+          pushAttempt({
+            modelRef: matched ? `${matched.provider}/${matched.id}` : modelRef,
+            provider,
+            modelId,
+            modelName: matched?.name,
+            thinkingLevel: thinking.effective,
+            ...(matched && thinking.clamped
+              ? { requestedThinkingLevel: thinking.requested, thinkingLevelClamped: true }
+              : {}),
+            autoSelected: false,
+            selectionReason,
+          });
+        } else {
+          const selectionResult = await selectOracleModel(ctx, configuredThinkingLevel);
+          if (!selectionResult.ok) {
+            return finalizeResult({
+              content: [{ type: "text", text: selectionResult.error }],
+              details: {
+                modelRef: "",
+                provider: ctx.model?.provider ?? "unknown",
+                modelId: "",
+                modelName: undefined,
+                thinkingLevel: configuredThinkingLevel ?? DEFAULT_THINKING_LEVEL,
+                autoSelected: true,
+                selectionReason: selectionResult.error,
+                includeBash: params.includeBash ?? false,
+                usage: createEmptyUsage(),
+                stderr: "",
+                exitCode: 1,
+                durationMs: 0,
+                cwd: params.cwd ?? ctx.cwd,
+              },
+            });
+          }
+          for (const candidate of selectionResult.ordered) pushAttempt(candidate);
+        }
+
+        pushAttempt(buildSessionFallbackSelection(ctx, configuredThinkingLevel));
+
+        const handleUpdate = (partial: {
+          content: Array<{ type: "text"; text: string }>;
+          details: OracleDetails;
+        }) => {
+          uiRun.preview = partial.content[0]?.text ?? uiRun.preview;
+          updateOracleUi(ctx, activeRuns);
+          onUpdate?.(partial);
+        };
+
+        let lastErrorResult: { ok: false; error: string; details: OracleDetails } | undefined;
+        for (let index = 0; index < attempts.length; index++) {
+          const previous = index > 0 ? attempts[index - 1] : undefined;
+          const attempt = previous
+            ? withFallbackReason(attempts[index], previous)
+            : attempts[index];
+          uiRun.selection = attempt;
+          updateOracleUi(ctx, activeRuns);
+
+          const result = await runOracle(
+            attempt,
+            { ...params, task },
+            signal,
+            handleUpdate,
+            ctx.cwd,
+            spawnImpl,
+          );
+          mergeUsageStats(aggregateUsage, result.details.usage);
+          if (result.ok) {
+            return finalizeResult({
+              content: [{ type: "text", text: result.output }],
+              details: result.details,
+            });
+          }
+
+          lastErrorResult = result;
+          const canFallBack = index < attempts.length - 1 && isModelAvailabilityError(result.error);
+          if (!canFallBack) {
+            return finalizeResult({
+              content: [{ type: "text", text: result.error }],
+              details: result.details,
+            });
+          }
+        }
+
+        return finalizeResult({
+          content: [
+            {
+              type: "text",
+              text: lastErrorResult?.error ?? "Oracle could not select an available model.",
+            },
+          ],
+          details: lastErrorResult?.details ?? {
+            modelRef: "",
+            provider: ctx.model?.provider ?? "unknown",
+            modelId: "",
+            modelName: undefined,
+            thinkingLevel: configuredThinkingLevel ?? DEFAULT_THINKING_LEVEL,
+            autoSelected: true,
+            selectionReason: "No authenticated models were available to run the oracle.",
+            includeBash: params.includeBash ?? false,
+            usage: createEmptyUsage(),
+            stderr: "",
+            exitCode: 1,
+            durationMs: 0,
+            cwd: params.cwd ?? ctx.cwd,
+          },
+        });
+      } finally {
+        activeRuns.delete(toolCallId);
+        updateOracleUi(ctx, activeRuns);
+      }
+    },
+
+    renderCall(args, theme) {
+      const task = args.task.length > 90 ? `${args.task.slice(0, 90)}...` : args.task;
+      const mode = args.includeBash ? "read-only+bash" : "read-only";
+      const override = args.model ? ` ${theme.fg("muted", `[${args.model}]`)}` : "";
+      const thinking = args.thinkingLevel
+        ? ` ${theme.fg("warning", `(${args.thinkingLevel})`)}`
+        : "";
+      return new Text(
+        `${theme.fg("toolTitle", theme.bold("oracle "))}${theme.fg("accent", mode)}${override}${thinking}\n  ${theme.fg("dim", task)}`,
+        0,
+        0,
+      );
+    },
+
+    renderResult(result, { expanded }, theme) {
+      const details = result.details as OracleDetails | undefined;
+      const body = result.content[0]?.type === "text" ? result.content[0].text : "(no output)";
+      if (!details) return new Text(body, 0, 0);
+
+      const isError = (details.exitCode ?? 0) !== 0;
+      const icon = isError ? theme.fg("error", "✗") : theme.fg("success", "✓");
+      const header = `${icon} ${theme.fg("toolTitle", theme.bold("oracle "))}${theme.fg("accent", details.modelRef || "(auto)")}`;
+      const subheader = [
+        details.thinkingLevel !== "off" ? details.thinkingLevel : undefined,
+        details.includeBash ? "read-only+bash" : "read-only",
+        formatDuration(details.durationMs),
+      ]
+        .filter(Boolean)
+        .join(" · ");
+      const usage = formatUsage(details.usage);
+
+      if (!expanded) {
+        let text = header;
+        if (subheader) text += `\n${theme.fg("dim", subheader)}`;
+        text += `\n\n${theme.fg("toolOutput", renderCollapsedText(body))}`;
+        if (usage) text += `\n\n${theme.fg("dim", usage)}`;
+        if (isError && details.stderr)
+          text += `\n${theme.fg("error", renderCollapsedText(details.stderr, 4))}`;
+        text += `\n${theme.fg("muted", "(Ctrl+O to expand)")}`;
+        return new Text(text, 0, 0);
+      }
+
+      const container = new Container();
+      container.addChild(new Text(header, 0, 0));
+      if (subheader) container.addChild(new Text(theme.fg("dim", subheader), 0, 0));
+      container.addChild(new Spacer(1));
+      container.addChild(new Text(theme.fg("muted", "Selection"), 0, 0));
+      container.addChild(new Text(theme.fg("dim", details.selectionReason), 0, 0));
+      container.addChild(new Spacer(1));
+      container.addChild(new Text(theme.fg("muted", "Output"), 0, 0));
+      container.addChild(new Markdown(body.trim(), 0, 0, getMarkdownTheme()));
+      if (usage) {
+        container.addChild(new Spacer(1));
+        container.addChild(new Text(theme.fg("muted", "Usage"), 0, 0));
+        container.addChild(new Text(theme.fg("dim", usage), 0, 0));
+      }
+      if (details.stderr) {
+        container.addChild(new Spacer(1));
+        container.addChild(new Text(theme.fg("muted", "stderr"), 0, 0));
+        container.addChild(
+          new Text(
+            isError ? theme.fg("error", details.stderr) : theme.fg("dim", details.stderr),
+            0,
+            0,
+          ),
+        );
+      }
+      return container;
+    },
+  });
 }