@oh-my-pi/pi-ai 15.11.6 → 15.11.7

package/CHANGELOG.md

@@ -2,6 +2,22 @@
 
 ## [Unreleased]
 
+## [15.11.7] - 2026-06-12
+
+### Added
+
+- Added `requestModelId` and `thinking.suppress` options to `google-gemini-cli` so collapsed effort-tier variants serialize their per-effort upstream wire id, and thinking-off requests on models with `thinking.suppressWhenOff` send an explicit `thinkingConfig` (`includeThoughts: false` with `thinkingLevel: "MINIMAL"` or `thinkingBudget: 0`) — Cloud Code Assist re-applies the per-id baked server default when the config is omitted, silently thinking and billing the tokens
+- Added mandatory-reasoning clamping: models baked with `thinking.requiresEffort` floor omitted or disabled reasoning to the lowest supported effort in every api mapping, and `disableReasoning` no longer emits OpenRouter `reasoning: { enabled: false }` for them — fixes `omp bench` and utility requests 400ing with "Reasoning is mandatory for this endpoint and cannot be disabled" on OpenRouter Gemini 3.x
+
+### Changed
+
+- Changed `google-gemini-cli` request mapping to route per-request wire ids via `resolveWireModelId`: the session effort picks the backing variant id (collapsed `gemini-3.5-flash` at high → `gemini-3.5-flash-low`; claude pairs route off → bare id, efforts → `-thinking`) while `AssistantMessage.model` and usage attribution stay on the logical id. A thinking budget clamped to zero now falls through to the thinking-off path (off routing plus suppression) instead of only disabling thinking
+- Changed `openai-completions` and `anthropic-messages` to serialize per-request wire ids via `resolveWireModelId`, so collapsed `X`/`X-thinking` pairs on aggregators and custom providers switch to the thinking SKU when reasoning is enabled (previously only `google-gemini-cli` routed effort-tier variants)
+
+### Fixed
+
+- Fixed `google-gemini-cli` ignoring `Model.requestModelId` when serializing the request model id
+
 ## [15.11.5] - 2026-06-12
 
 ### Added

package/dist/types/providers/anthropic.d.ts

@@ -83,6 +83,11 @@ export interface AnthropicOptions extends StreamOptions {
      * Ignored for adaptive-capable models.
      */
     thinkingBudgetTokens?: number;
+    /**
+     * Upstream wire model id override for collapsed effort-tier variants.
+     * Serialized as `requestModelId ?? model.requestModelId ?? model.id`.
+     */
+    requestModelId?: string;
     /**
      * Effort level for adaptive thinking.
      * Controls how much thinking Claude allocates:

package/dist/types/providers/google-gemini-cli.d.ts

@@ -34,7 +34,23 @@ export interface GoogleGeminiCliOptions extends StreamOptions {
         budgetTokens?: number;
         /** Thinking level. Use for Gemini 3 models (LOW/HIGH for Pro, MINIMAL/LOW/MEDIUM/HIGH for Flash). */
         level?: GoogleThinkingLevel;
+        /**
+         * Explicit wire suppression when `enabled` is false. Cloud Code Assist
+         * re-applies the per-id baked server default when thinkingConfig is
+         * omitted, so models with `thinking.suppressWhenOff` must send
+         * `includeThoughts: false` plus a MINIMAL level (or zero budget).
+         */
+        suppress?: {
+            level: GoogleThinkingLevel;
+        } | {
+            budget: number;
+        };
     };
+    /**
+     * Upstream wire model id override for collapsed effort-tier variants.
+     * Serialized as `requestModelId ?? model.requestModelId ?? model.id`.
+     */
+    requestModelId?: string;
     projectId?: string;
 }
 export { ANTIGRAVITY_SYSTEM_INSTRUCTION, getAntigravityUserAgent, getGeminiCliHeaders, getGeminiCliUserAgent, } from "@oh-my-pi/pi-catalog/wire/gemini-headers";

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-ai",
-	"version": "15.11.6",
+	"version": "15.11.7",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
@@ -38,8 +38,8 @@
 	},
 	"dependencies": {
 		"@bufbuild/protobuf": "^2.12.0",
-		"@oh-my-pi/pi-catalog": "15.11.6",
-		"@oh-my-pi/pi-utils": "15.11.6",
+		"@oh-my-pi/pi-catalog": "15.11.7",
+		"@oh-my-pi/pi-utils": "15.11.7",
 		"openai": "^6.39.0",
 		"partial-json": "^0.1.7",
 		"zod": "4.4.3"

package/src/providers/anthropic.ts

@@ -973,6 +973,11 @@ export interface AnthropicOptions extends StreamOptions {
 	 * Ignored for adaptive-capable models.
 	 */
 	thinkingBudgetTokens?: number;
+	/**
+	 * Upstream wire model id override for collapsed effort-tier variants.
+	 * Serialized as `requestModelId ?? model.requestModelId ?? model.id`.
+	 */
+	requestModelId?: string;
 	/**
 	 * Effort level for adaptive thinking.
 	 * Controls how much thinking Claude allocates:
@@ -2816,7 +2821,7 @@ function buildParams(
 	// Build params in the canonical field order: model → messages → system → tools →
 	// metadata → max_tokens → thinking → context_management → output_config → stream.
 	const params: MessageCreateParamsStreaming = {
-		model: model.requestModelId ?? model.id,
+		model: options?.requestModelId ?? model.requestModelId ?? model.id,
 		messages: convertAnthropicMessages(context.messages, model, isOAuthToken),
 		...(systemBlocks && { system: systemBlocks }),
 		...(tools !== undefined && { tools }),

package/src/providers/google-gemini-cli.ts

@@ -76,7 +76,19 @@ export interface GoogleGeminiCliOptions extends StreamOptions {
 		budgetTokens?: number;
 		/** Thinking level. Use for Gemini 3 models (LOW/HIGH for Pro, MINIMAL/LOW/MEDIUM/HIGH for Flash). */
 		level?: GoogleThinkingLevel;
+		/**
+		 * Explicit wire suppression when `enabled` is false. Cloud Code Assist
+		 * re-applies the per-id baked server default when thinkingConfig is
+		 * omitted, so models with `thinking.suppressWhenOff` must send
+		 * `includeThoughts: false` plus a MINIMAL level (or zero budget).
+		 */
+		suppress?: { level: GoogleThinkingLevel } | { budget: number };
 	};
+	/**
+	 * Upstream wire model id override for collapsed effort-tier variants.
+	 * Serialized as `requestModelId ?? model.requestModelId ?? model.id`.
+	 */
+	requestModelId?: string;
 	projectId?: string;
 }
 
@@ -770,6 +782,17 @@ export function buildRequest(
 		} else if (options.thinking.budgetTokens !== undefined) {
 			generationConfig.thinkingConfig.thinkingBudget = options.thinking.budgetTokens;
 		}
+	} else if (options.thinking?.suppress && model.reasoning) {
+		// Explicit off: omitting thinkingConfig re-applies the per-id baked
+		// server default (the model silently thinks and bills the tokens).
+		const suppress = options.thinking.suppress;
+		generationConfig.thinkingConfig = { includeThoughts: false };
+		if ("level" in suppress) {
+			// Cast to any since our GoogleThinkingLevel mirrors Google's ThinkingLevel enum values
+			generationConfig.thinkingConfig.thinkingLevel = suppress.level as any;
+		} else {
+			generationConfig.thinkingConfig.thinkingBudget = suppress.budget;
+		}
 	}
 
 	const request: CloudCodeAssistRequest["request"] = {
@@ -840,7 +863,7 @@ export function buildRequest(
 
 	return {
 		project: projectId,
-		model: model.id,
+		model: options.requestModelId ?? model.requestModelId ?? model.id,
 		request,
 		...(isAntigravity
 			? {

package/src/providers/openai-completions.ts

@@ -1,7 +1,7 @@
 import type { Effort } from "@oh-my-pi/pi-catalog/effort";
 import { toFirepassWireModelId, toFireworksWireModelId } from "@oh-my-pi/pi-catalog/fireworks-model-id";
 import { isDeepseekModelIdOrName } from "@oh-my-pi/pi-catalog/identity";
-import { getSupportedEfforts } from "@oh-my-pi/pi-catalog/model-thinking";
+import { getSupportedEfforts, resolveWireModelId } from "@oh-my-pi/pi-catalog/model-thinking";
 import { calculateCost } from "@oh-my-pi/pi-catalog/models";
 import type { ResolvedOpenAICompat } from "@oh-my-pi/pi-catalog/types";
 import { parseGitHubCopilotApiKey } from "@oh-my-pi/pi-catalog/wire/github-copilot";
@@ -113,12 +113,16 @@ function resolveOpenAICompletionsModelId(
 	model: Model<"openai-completions">,
 	options: OpenAICompletionsOptions | undefined,
 ): string {
-	// Catalog variants (e.g. Copilot long-context `-1m` entries) pin the wire id.
-	if (model.requestModelId) return model.requestModelId;
-	if (model.provider === "firepass") return toFirepassWireModelId(model.id);
-	if (model.provider === "fireworks") return toFireworksWireModelId(model.id);
-	if (model.provider === "openrouter") return applyOpenRouterRoutingVariant(model.id, options?.openrouterVariant);
-	return model.id;
+	// Effort-tier variants route per request effort (off → bare id, efforts →
+	// the thinking backing id); catalog variants (Copilot long-context `-1m`
+	// entries) pin via `requestModelId`; everything else serializes `model.id`.
+	const effort =
+		options?.reasoning && !options.disableReasoning && model.reasoning ? (options.reasoning as Effort) : undefined;
+	const wireId = resolveWireModelId(model, effort);
+	if (model.provider === "firepass") return toFirepassWireModelId(wireId);
+	if (model.provider === "fireworks") return toFireworksWireModelId(wireId);
+	if (model.provider === "openrouter") return applyOpenRouterRoutingVariant(wireId, options?.openrouterVariant);
+	return wireId;
 }
 
 /**

package/src/stream.ts

@@ -3,7 +3,9 @@ import { isVertexExpressOpenAIUrl, isVertexRawPredictUrl } from "@oh-my-pi/pi-ca
 import {
 	mapEffortToAnthropicAdaptiveEffort,
 	mapEffortToGoogleThinkingLevel,
+	minimumSupportedEffort,
 	requireSupportedEffort,
+	resolveWireModelId,
 } from "@oh-my-pi/pi-catalog/model-thinking";
 import { CATALOG_PROVIDERS, type ProviderCatalogEntry } from "@oh-my-pi/pi-catalog/provider-models";
 import { $env, $pickenv, extractHttpStatusFromError } from "@oh-my-pi/pi-utils";
@@ -667,11 +669,39 @@ function resolveOpenAiReasoningEffort<TApi extends Api>(
 
 const castApi = <TApi extends Api>(api: OptionsForApi<TApi>): OptionsForApi<Api> => api as OptionsForApi<Api>;
 
-function mapOptionsForApi<TApi extends Api>(
+/**
+ * Mandatory-reasoning endpoints (`thinking.requiresEffort`) reject disabled
+ * or omitted thinking ("Reasoning is mandatory for this endpoint and cannot
+ * be disabled") — clamp to the lowest supported effort instead.
+ * `suppressWhenOff` models handle off provider-side via explicit wire
+ * suppression. Collapsed pairs interplay: pair derivation strips member
+ * flags (off routes to a bare SKU that CAN disable), while identity backfill
+ * re-flags pairs whose logical id is itself mandatory (Gemini 3.x) — there
+ * the clamp wins and the floored effort routes to the thinking SKU.
+ */
+function normalizeMandatoryReasoningOptions<TApi extends Api>(
 	model: Model<TApi>,
 	options?: SimpleStreamOptions,
+): SimpleStreamOptions | undefined {
+	if (
+		!model.reasoning ||
+		!model.thinking?.requiresEffort ||
+		model.thinking.suppressWhenOff ||
+		(options?.reasoning !== undefined && !options.disableReasoning)
+	) {
+		return options;
+	}
+	const floor = minimumSupportedEffort(model);
+	if (floor === undefined) return options;
+	return { ...options, reasoning: floor, disableReasoning: undefined };
+}
+
+function mapOptionsForApi<TApi extends Api>(
+	model: Model<TApi>,
+	rawOptions?: SimpleStreamOptions,
 	apiKey?: string,
 ): OptionsForApi<TApi> {
+	const options = normalizeMandatoryReasoningOptions(model, rawOptions);
 	const base = {
 		temperature: options?.temperature,
 		topP: options?.topP,
@@ -707,6 +737,7 @@ function mapOptionsForApi<TApi extends Api>(
 			if (!reasoning || !model.reasoning) {
 				return castApi<"anthropic-messages">({
 					...base,
+					requestModelId: resolveWireModelId(model, undefined),
 					thinkingEnabled: false,
 					toolChoice: mapAnthropicToolChoice(options?.toolChoice),
 					thinkingDisplay: options?.hideThinkingSummary ? "omitted" : undefined,
@@ -718,6 +749,7 @@ function mapOptionsForApi<TApi extends Api>(
 			if (thinkingBudget <= 0) {
 				return castApi<"anthropic-messages">({
 					...base,
+					requestModelId: resolveWireModelId(model, undefined),
 					thinkingEnabled: false,
 					toolChoice: mapAnthropicToolChoice(options?.toolChoice),
 					thinkingDisplay: options?.hideThinkingSummary ? "omitted" : undefined,
@@ -731,6 +763,7 @@ function mapOptionsForApi<TApi extends Api>(
 				const effort = mapEffortToAnthropicAdaptiveEffort(model, reasoning);
 				return castApi<"anthropic-messages">({
 					...base,
+					requestModelId: resolveWireModelId(model, reasoning),
 					thinkingEnabled: true,
 					effort,
 					toolChoice: mapAnthropicToolChoice(options?.toolChoice),
@@ -742,6 +775,7 @@ function mapOptionsForApi<TApi extends Api>(
 			if (ANTHROPIC_USE_INTERLEAVED_THINKING) {
 				return castApi<"anthropic-messages">({
 					...base,
+					requestModelId: resolveWireModelId(model, reasoning),
 					thinkingEnabled: true,
 					thinkingBudgetTokens: thinkingBudget,
 					toolChoice: mapAnthropicToolChoice(options?.toolChoice),
@@ -762,6 +796,7 @@ function mapOptionsForApi<TApi extends Api>(
 			if (thinkingBudget <= 0) {
 				return castApi<"anthropic-messages">({
 					...base,
+					requestModelId: resolveWireModelId(model, undefined),
 					thinkingEnabled: false,
 					toolChoice: mapAnthropicToolChoice(options?.toolChoice),
 					thinkingDisplay: options?.hideThinkingSummary ? "omitted" : undefined,
@@ -771,6 +806,7 @@ function mapOptionsForApi<TApi extends Api>(
 				return castApi<"anthropic-messages">({
 					...base,
 					maxTokens,
+					requestModelId: resolveWireModelId(model, reasoning),
 					thinkingEnabled: true,
 					thinkingBudgetTokens: thinkingBudget,
 					toolChoice: mapAnthropicToolChoice(options?.toolChoice),
@@ -887,53 +923,57 @@ function mapOptionsForApi<TApi extends Api>(
 
 		case "google-gemini-cli": {
 			const reasoning = options?.reasoning;
-			if (!reasoning || !model.reasoning) {
-				return castApi<"google-gemini-cli">({
-					...base,
-					thinking: { enabled: false },
-					toolChoice: mapGoogleToolChoice(options?.toolChoice),
-				});
-			}
-
-			const effort = requireSupportedEffort(model, reasoning);
+			const toolChoice = mapGoogleToolChoice(options?.toolChoice);
+			if (reasoning && model.reasoning) {
+				const effort = requireSupportedEffort(model, reasoning);
+
+				// Gemini 3+ models use thinkingLevel instead of thinkingBudget
+				if (model.thinking?.mode === "google-level") {
+					return castApi<"google-gemini-cli">({
+						...base,
+						requestModelId: resolveWireModelId(model, effort),
+						thinking: {
+							enabled: true,
+							level: mapEffortToGoogleThinkingLevel(effort),
+						},
+						toolChoice,
+					});
+				}
 
-			// Gemini 3+ models use thinkingLevel instead of thinkingBudget
-			if (model.thinking?.mode === "google-level") {
-				return castApi<"google-gemini-cli">({
-					...base,
-					thinking: {
-						enabled: true,
-						level: mapEffortToGoogleThinkingLevel(effort),
-					},
-					toolChoice: mapGoogleToolChoice(options?.toolChoice),
-				});
-			}
+				let thinkingBudget = options.thinkingBudgets?.[effort] ?? GOOGLE_THINKING[effort];
 
-			let thinkingBudget = options.thinkingBudgets?.[effort] ?? GOOGLE_THINKING[effort];
+				// Caller's maxTokens is the desired output; add thinking budget on top, capped at model limit
+				const maxTokens = Math.min((base.maxTokens || 0) + thinkingBudget, model.maxTokens);
 
-			// Caller's maxTokens is the desired output; add thinking budget on top, capped at model limit
-			const maxTokens = Math.min((base.maxTokens || 0) + thinkingBudget, model.maxTokens);
+				// If not enough room for thinking + output, reduce thinking budget
+				if (maxTokens <= thinkingBudget) {
+					thinkingBudget = Math.max(0, maxTokens - MIN_OUTPUT_TOKENS);
+				}
 
-			// If not enough room for thinking + output, reduce thinking budget
-			if (maxTokens <= thinkingBudget) {
-				thinkingBudget = Math.max(0, maxTokens - MIN_OUTPUT_TOKENS) ?? 0;
+				if (thinkingBudget > 0) {
+					return castApi<"google-gemini-cli">({
+						...base,
+						maxTokens,
+						requestModelId: resolveWireModelId(model, effort),
+						thinking: { enabled: true, budgetTokens: thinkingBudget },
+						toolChoice,
+					});
+				}
+				// Budget clamped to zero — fall through to the thinking-off path.
 			}
 
-			// If thinking budget is too low, disable thinking
-			if (thinkingBudget <= 0) {
-				return castApi<"google-gemini-cli">({
-					...base,
-					thinking: { enabled: false },
-					toolChoice: mapGoogleToolChoice(options?.toolChoice),
-				});
-			} else {
-				return castApi<"google-gemini-cli">({
-					...base,
-					maxTokens,
-					thinking: { enabled: true, budgetTokens: thinkingBudget },
-					toolChoice: mapGoogleToolChoice(options?.toolChoice),
-				});
+			const thinking: GoogleGeminiCliOptions["thinking"] = { enabled: false };
+			if (model.reasoning && model.thinking?.suppressWhenOff) {
+				// CCA re-applies the per-id baked server default when the config
+				// is omitted; suppression must be explicit on the wire.
+				thinking.suppress = model.thinking.mode === "google-level" ? { level: "MINIMAL" } : { budget: 0 };
 			}
+			return castApi<"google-gemini-cli">({
+				...base,
+				requestModelId: resolveWireModelId(model, undefined),
+				thinking,
+				toolChoice,
+			});
 		}
 
 		case "google-vertex": {