Add GPT-5.6 max reasoning effort

README.md

@@ -7,7 +7,7 @@ Use dreb if you want a coding agent that can run against direct APIs, coding sub
 ## Why choose dreb?
 
 - **Every session, on every device.** The [web dashboard](#web-dashboard) is a first-party browser UI for the same sessions the terminal runs: a fleet overview of all live and past sessions across projects, full chat with steering, live subagent observability, host file access, dreb memory management, and settings — one synchronized state on desktop and mobile. Local-only by default; remote access is Tailscale-gated with device pairing.
-- **Model and provider freedom.** Authenticate with API keys or `/login` subscriptions, switch models at runtime with `/model`, scope model sets, tune thinking levels, route built-in providers through proxies, use cloud providers such as Bedrock/Vertex/Azure, or add local/proxy/custom models through [Custom Models](packages/coding-agent/docs/models.md) and [Custom Providers](packages/coding-agent/docs/custom-provider.md). See [Providers](packages/coding-agent/docs/providers.md) for the current setup list.
+- **Model and provider freedom.** Authenticate with API keys or `/login` subscriptions, switch models at runtime with `/model`, scope model sets, and tune model-aware thinking levels through `xhigh` plus a model-aware `max` tier. Codex `ultra` is orchestration (`max` plus local multi-agent work), not a raw provider effort. Route built-in providers through proxies, use cloud providers such as Bedrock/Vertex/Azure, or add local/proxy/custom models through [Custom Models](packages/coding-agent/docs/models.md) and [Custom Providers](packages/coding-agent/docs/custom-provider.md). See [Providers](packages/coding-agent/docs/providers.md) for the current setup list.
 - **A real development workflow.** [mach6](packages/coding-agent/docs/mach6.md) is a built-in issue-to-merge workflow: assess issues, plan work, open draft PRs, implement, push progress, run multi-agent reviews, independently assess findings, fix CI or review items, and publish. Plans, reviews, and progress live on GitHub as shared memory.
 - **Composable agent building blocks.** [Skills](packages/coding-agent/docs/skills.md) are markdown workflows loaded on demand; [extensions](packages/coding-agent/docs/extensions.md) are TypeScript modules for custom tools, commands, event hooks, UI components, renderers, keybindings, provider registration, permission gates, and workflow automation; [packages](packages/coding-agent/docs/packages.md) bundle skills, extensions, prompts, and themes for npm, git, or local sharing.
 - **Parallel and specialized agents.** The optional `subagent` tool runs role-matched work in independent child agents using single, parallel, or chain mode. Omitting the agent type selects `Explore`, which retrieves concrete evidence such as files, symbols, documentation, call sites, exact snippets, and explicit data flows; the primary agent retains root-cause diagnosis, requirements interpretation, design, implementation recommendations, planning, synthesis, and final conclusions. Parallel and chain modes do not relax that boundary, while specialized agents continue to perform the broader work in their own definitions. Custom agent definitions can inherit models, record child-session metadata for audit trails, and power workflows such as mach6's specialized reviewers. Per-agent models and per-request thinking remain explicit controls. The built-in [`model-routing-guide` skill](packages/coding-agent/docs/skills.md#model-routing-guide) researches scoped provider/model candidates and local child history, and its `update` mode preserves retained entries while removing stale scope and researching newly added models; the optional global-only [Dispatch Arbiter](packages/coding-agent/docs/agent-models.md#dispatch-arbiter) consumes that guide in a fully headless, tool-less call before every child spawn and may change only agent, scoped canonical model, and supported thinking. Its bounded rolling parent activity follows the title setter and includes useful tool outputs, with existing secret scrubbing applied before inference. It is disabled by default and fails closed—bad configuration, guide, inference, or decisions prevent the child from spawning rather than silently keeping the original route. TUI `/settings` and dashboard Settings expose its enable toggle, exact model, thinking, guide path, and validation/readiness feedback. Typed decisions are persisted and visible in the TUI, JSON/RPC, and dashboard. The same settings surfaces expose `backgroundAgents.maxConcurrentSubagents` (default `4`); `0` starts new parent sessions without the subagent tool and explicitly tells the parent model to perform normally delegated work itself. While background subagents run, a separate guardrail pauses the parent after a few turns, configurable via [`backgroundAgents`](packages/coding-agent/docs/settings.md#background-agents).

package.json

@@ -39,7 +39,7 @@
 		"node": "22.x"
 	},
 	"packageManager": "npm@11.5.1",
-	"version": "2.61.0",
+	"version": "2.61.1",
 	"dependencies": {
 		"@dreb/coding-agent": "*",
 		"@mariozechner/jiti": "^2.6.5",

packages/agent/README.md

@@ -150,7 +150,7 @@ const agent = new Agent({
   initialState: {
     systemPrompt: string,
     model: Model<any>,
-    thinkingLevel: "off" | "minimal" | "low" | "medium" | "high" | "xhigh",
+    thinkingLevel: "off" | "minimal" | "low" | "medium" | "high" | "xhigh" | "max",
     tools: AgentTool<any>[],
     messages: AgentMessage[],
   },

packages/agent/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@dreb/agent-core",
-	"version": "2.61.0",
+	"version": "2.61.1",
 	"description": "General-purpose agent with transport abstraction, state management, and attachment support",
 	"type": "module",
 	"main": "./dist/index.js",

packages/agent/src/agent-loop.ts

@@ -9,6 +9,7 @@ import {
 	DEFAULT_MAX_OUTPUT_TOKENS,
 	EventStream,
 	streamSimple,
+	supportsMax,
 	supportsXhigh,
 	type ToolResultMessage,
 	validateToolArguments,
@@ -30,6 +31,7 @@ export type AgentEventSink = (event: AgentEvent) => Promise<void> | void;
 function getEffectiveThinkingLevel(config: AgentLoopConfig): ThinkingLevel {
 	const requested = config.reasoning ?? "off";
 	if (!config.model.reasoning) return "off";
+	if (requested === "max" && !supportsMax(config.model)) return supportsXhigh(config.model) ? "xhigh" : "high";
 	return requested === "xhigh" && !supportsXhigh(config.model) ? "high" : requested;
 }
 

packages/agent/src/types.ts

@@ -263,7 +263,7 @@ export interface AgentLoopConfig extends SimpleStreamOptions {
  * Thinking/reasoning level for models that support it.
  * Note: "xhigh" is only supported by OpenAI reasoning models that explicitly advertise extended reasoning (for example GPT-5.2+ and Codex GPT-5.3+).
  */
-export type ThinkingLevel = "off" | "minimal" | "low" | "medium" | "high" | "xhigh";
+export type ThinkingLevel = "off" | "minimal" | "low" | "medium" | "high" | "xhigh" | "max";
 
 /**
  * Extensible interface for custom app messages.

packages/ai/README.md

@@ -443,7 +443,8 @@ if (model.reasoning) {
 const response = await completeSimple(model, {
   messages: [{ role: 'user', content: 'Solve: 2x + 5 = 13' }]
 }, {
-  reasoning: 'medium'  // 'minimal' | 'low' | 'medium' | 'high' | 'xhigh' (xhigh maps to 'max' on Claude Opus 4.6–4.x and Claude 5 families; 'high' on other Anthropic models)
+  reasoning: 'medium'  // 'minimal' | 'low' | 'medium' | 'high' | 'xhigh' | 'max'
+  // Native normalized 'max' is model-aware (currently GPT-5.6). Claude's established xhigh mapping still emits provider-native 'max'.
 });
 
 // Access thinking and text blocks

packages/ai/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@dreb/ai",
-	"version": "2.61.0",
+	"version": "2.61.1",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"type": "module",
 	"main": "./dist/index.js",
@@ -83,7 +83,7 @@
 		"ajv": "^8.17.1",
 		"ajv-formats": "^3.0.1",
 		"chalk": "^5.6.2",
-		"openai": "6.26.0",
+		"openai": "6.49.0",
 		"partial-json": "^0.1.7",
 		"proxy-agent": "^6.5.0",
 		"undici": "^7.19.1",

packages/ai/src/models.ts

@@ -187,6 +187,17 @@ export function isQwen38OrLater(modelId: string): boolean {
 	return qwen.major > 3 || (qwen.major === 3 && qwen.minor >= 8);
 }
 
+/**
+ * Check if a model supports the native `max` reasoning tier.
+ *
+ * Supported today:
+ * - GPT-5.6 model families (Sol, Terra, Luna, and the alias)
+ */
+export function supportsMax<TApi extends Api>(model: Model<TApi>): boolean {
+	if (/(?:^|\/)gpt-5\.6(?:$|[-.])/.test(model.id.toLowerCase())) return true;
+	return false;
+}
+
 /**
  * Check if a model supports xhigh thinking level.
  *

packages/ai/src/providers/amazon-bedrock.ts

@@ -415,6 +415,7 @@ function mapThinkingLevelToEffort(
 		case "high":
 			return "high";
 		case "xhigh":
+		case "max":
 			return supportsXhigh(model) ? "max" : "high";
 		default:
 			return "high";
@@ -732,10 +733,11 @@ export function buildAdditionalModelRequestFields(
 						medium: 8192,
 						high: 16384,
 						xhigh: 16384, // Claude doesn't support xhigh, clamp to high
+						max: 16384, // Normalized max falls back through xhigh to high here
 					};
 
-					// Custom budgets override defaults (xhigh not in ThinkingBudgets, use high)
-					const level = options.reasoning === "xhigh" ? "high" : options.reasoning;
+					// Custom budgets do not define xhigh/max, so both use high.
+					const level = options.reasoning === "xhigh" || options.reasoning === "max" ? "high" : options.reasoning;
 					const budget = options.thinkingBudgets?.[level] ?? defaultBudgets[options.reasoning];
 
 					return {

packages/ai/src/providers/anthropic.ts

@@ -479,6 +479,7 @@ function mapThinkingLevelToEffort(
 		case "high":
 			return "high";
 		case "xhigh":
+		case "max":
 			return supportsXhigh(model) ? "max" : "high";
 		default:
 			return "high";

packages/ai/src/providers/azure-openai-responses.ts

@@ -1,7 +1,6 @@
 import { AzureOpenAI } from "openai";
 import type { ResponseCreateParamsStreaming } from "openai/resources/responses/responses.js";
 import { getEnvApiKey } from "../env-api-keys.js";
-import { supportsXhigh } from "../models.js";
 import type {
 	Api,
 	AssistantMessage,
@@ -13,7 +12,7 @@ import type {
 } from "../types.js";
 import { AssistantMessageEventStream } from "../utils/event-stream.js";
 import { convertResponsesMessages, convertResponsesTools, processResponsesStream } from "./openai-responses-shared.js";
-import { buildBaseOptions, clampReasoning } from "./simple-options.js";
+import { buildBaseOptions, resolveReasoningEffort } from "./simple-options.js";
 
 const DEFAULT_AZURE_API_VERSION = "v1";
 const AZURE_TOOL_CALL_PROVIDERS = new Set(["openai", "openai-codex", "opencode", "azure-openai-responses"]);
@@ -41,7 +40,7 @@ function resolveDeploymentName(model: Model<"azure-openai-responses">, options?:
 
 // Azure OpenAI Responses-specific options
 export interface AzureOpenAIResponsesOptions extends StreamOptions {
-	reasoningEffort?: "minimal" | "low" | "medium" | "high" | "xhigh";
+	reasoningEffort?: "minimal" | "low" | "medium" | "high" | "xhigh" | "max";
 	reasoningSummary?: "auto" | "detailed" | "concise" | null;
 	azureApiVersion?: string;
 	azureResourceName?: string;
@@ -133,7 +132,7 @@ export const streamSimpleAzureOpenAIResponses: StreamFunction<"azure-openai-resp
 	}
 
 	const base = buildBaseOptions(model, options, apiKey);
-	const reasoningEffort = supportsXhigh(model) ? options?.reasoning : clampReasoning(options?.reasoning);
+	const reasoningEffort = resolveReasoningEffort(model, options?.reasoning);
 
 	return streamAzureOpenAIResponses(model, context, {
 		...base,

packages/ai/src/providers/google-gemini-cli.ts

@@ -964,7 +964,7 @@ export function buildRequest(
 	};
 }
 
-type ClampedThinkingLevel = Exclude<ThinkingLevel, "xhigh">;
+type ClampedThinkingLevel = Exclude<ThinkingLevel, "xhigh" | "max">;
 
 function getDisabledThinkingConfig(modelId: string): ThinkingConfig {
 	// Google docs: Gemini 3.1 Pro cannot disable thinking, and Gemini 3 Flash / Flash-Lite