@oh-my-pi/pi-ai 14.7.5 → 14.7.6

package/CHANGELOG.md

@@ -2,6 +2,15 @@
 
 ## [Unreleased]
 
+## [14.7.6] - 2026-05-07
+
+### Added
+
+- Added `hideThinkingSummary` option to `SimpleStreamOptions`. When true, `streamSimple` requests that the underlying provider omit reasoning/thinking summaries: Anthropic receives `thinking.display = "omitted"` (where supported), and OpenAI Responses / Azure / Codex providers leave `reasoning.summary` unset so the server skips emitting the human-readable summary stream entirely.
+
+### Changed
+
+- Changed OpenAI Responses, Azure OpenAI Responses, and OpenAI Codex providers to omit `reasoning.summary` from requests when `reasoningSummary` is explicitly `null` (previously fell back to `"auto"`).
 ## [14.7.5] - 2026-05-07
 
 ### Added

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-ai",
-	"version": "14.7.5",
+	"version": "14.7.6",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"homepage": "https://github.com/can1357/oh-my-pi",
 	"author": "Can Boluk",
@@ -46,8 +46,8 @@
 		"@aws-sdk/credential-provider-node": "^3.972.39",
 		"@bufbuild/protobuf": "^2.12.0",
 		"@google/genai": "^1.52.0",
-		"@oh-my-pi/pi-natives": "14.7.5",
-		"@oh-my-pi/pi-utils": "14.7.5",
+		"@oh-my-pi/pi-natives": "14.7.6",
+		"@oh-my-pi/pi-utils": "14.7.6",
 		"@sinclair/typebox": "^0.34.49",
 		"@smithy/node-http-handler": "^4.6.1",
 		"ajv": "^8.20.0",

package/src/providers/azure-openai-responses.ts

@@ -317,11 +317,14 @@ function buildParams(
 		// See: https://github.com/can1357/oh-my-pi/issues/41
 		params.include = ["reasoning.encrypted_content"];
 
-		if (options?.reasoning || options?.reasoningSummary) {
-			params.reasoning = {
+		if (options?.reasoning || options?.reasoningSummary !== undefined) {
+			const reasoningParams: NonNullable<typeof params.reasoning> = {
 				effort: options?.reasoning || "medium",
-				summary: options?.reasoningSummary || "auto",
 			};
+			if (options?.reasoningSummary !== null) {
+				reasoningParams.summary = options?.reasoningSummary || "auto";
+			}
+			params.reasoning = reasoningParams;
 		} else {
 			if (model.name.toLowerCase().startsWith("gpt-5")) {
 				// Jesus Christ, see https://community.openai.com/t/need-reasoning-false-option-for-gpt-5/1351588/7

package/src/providers/openai-codex/request-transformer.ts

@@ -4,7 +4,7 @@ import type { Api, Model } from "../../types";
 
 export interface ReasoningConfig {
 	effort: "none" | "minimal" | "low" | "medium" | "high" | "xhigh";
-	summary: "auto" | "concise" | "detailed" | null;
+	summary?: "auto" | "concise" | "detailed";
 }
 
 export interface CodexRequestOptions {
@@ -52,11 +52,14 @@ export interface RequestBody {
 }
 
 function getReasoningConfig(model: Model<Api>, options: CodexRequestOptions): ReasoningConfig {
-	return {
+	const config: ReasoningConfig = {
 		effort:
 			options.reasoningEffort === "none" ? "none" : requireSupportedEffort(model, options.reasoningEffort as Effort),
-		summary: options.reasoningSummary ?? "detailed",
 	};
+	if (options.reasoningSummary !== null) {
+		config.summary = options.reasoningSummary ?? "detailed";
+	}
+	return config;
 }
 
 function filterInput(input: InputItem[] | undefined): InputItem[] | undefined {

package/src/providers/openai-responses.ts

@@ -430,13 +430,16 @@ function buildParams(
 		// See: https://github.com/can1357/oh-my-pi/issues/41
 		params.include = ["reasoning.encrypted_content"];
 
-		if (options?.reasoning || options?.reasoningSummary) {
-			params.reasoning = {
+		if (options?.reasoning || options?.reasoningSummary !== undefined) {
+			const reasoningParams: NonNullable<typeof params.reasoning> = {
 				effort: mapReasoningEffort(options?.reasoning || "medium", model.compat?.reasoningEffortMap) as NonNullable<
 					OpenAIResponsesSamplingParams["reasoning"]
 				>["effort"],
-				summary: options?.reasoningSummary || "auto",
 			};
+			if (options?.reasoningSummary !== null) {
+				reasoningParams.summary = options?.reasoningSummary || "auto";
+			}
+			params.reasoning = reasoningParams;
 		} else if (model.name.startsWith("gpt-5")) {
 			// Jesus Christ, see https://community.openai.com/t/need-reasoning-false-option-for-gpt-5/1351588/7
 			messages.push({

package/src/stream.ts

@@ -462,6 +462,7 @@ function mapOptionsForApi<TApi extends Api>(
 					...base,
 					thinkingEnabled: false,
 					toolChoice: mapAnthropicToolChoice(options?.toolChoice),
+					thinkingDisplay: options?.hideThinkingSummary ? "omitted" : undefined,
 				});
 			}
 
@@ -471,6 +472,7 @@ function mapOptionsForApi<TApi extends Api>(
 					...base,
 					thinkingEnabled: false,
 					toolChoice: mapAnthropicToolChoice(options?.toolChoice),
+					thinkingDisplay: options?.hideThinkingSummary ? "omitted" : undefined,
 				});
 			}
 
@@ -483,6 +485,7 @@ function mapOptionsForApi<TApi extends Api>(
 					thinkingEnabled: true,
 					effort,
 					toolChoice: mapAnthropicToolChoice(options?.toolChoice),
+					thinkingDisplay: options?.hideThinkingSummary ? "omitted" : undefined,
 				});
 			}
 
@@ -492,6 +495,7 @@ function mapOptionsForApi<TApi extends Api>(
 					thinkingEnabled: true,
 					thinkingBudgetTokens: thinkingBudget,
 					toolChoice: mapAnthropicToolChoice(options?.toolChoice),
+					thinkingDisplay: options?.hideThinkingSummary ? "omitted" : undefined,
 				});
 			}
 
@@ -509,6 +513,7 @@ function mapOptionsForApi<TApi extends Api>(
 					...base,
 					thinkingEnabled: false,
 					toolChoice: mapAnthropicToolChoice(options?.toolChoice),
+					thinkingDisplay: options?.hideThinkingSummary ? "omitted" : undefined,
 				});
 			} else {
 				return castApi<"anthropic-messages">({
@@ -517,6 +522,7 @@ function mapOptionsForApi<TApi extends Api>(
 					thinkingEnabled: true,
 					thinkingBudgetTokens: thinkingBudget,
 					toolChoice: mapAnthropicToolChoice(options?.toolChoice),
+					thinkingDisplay: options?.hideThinkingSummary ? "omitted" : undefined,
 				});
 			}
 		}
@@ -564,6 +570,7 @@ function mapOptionsForApi<TApi extends Api>(
 				reasoning: resolveOpenAiReasoningEffort(model, options),
 				toolChoice: mapOpenAiToolChoice(options?.toolChoice),
 				serviceTier: options?.serviceTier,
+				reasoningSummary: options?.hideThinkingSummary ? null : undefined,
 			});
 
 		case "azure-openai-responses":
@@ -572,6 +579,7 @@ function mapOptionsForApi<TApi extends Api>(
 				reasoning: resolveOpenAiReasoningEffort(model, options),
 				toolChoice: mapOpenAiToolChoice(options?.toolChoice),
 				serviceTier: options?.serviceTier,
+				reasoningSummary: options?.hideThinkingSummary ? null : undefined,
 			});
 
 		case "openai-codex-responses":
@@ -581,6 +589,7 @@ function mapOptionsForApi<TApi extends Api>(
 				toolChoice: mapOpenAiToolChoice(options?.toolChoice),
 				serviceTier: options?.serviceTier,
 				preferWebsockets: options?.preferWebsockets,
+				reasoningSummary: options?.hideThinkingSummary ? null : undefined,
 			});
 
 		case "google-generative-ai": {

package/src/types.ts

@@ -255,6 +255,14 @@ export interface SimpleStreamOptions extends StreamOptions {
 	 * this way when `reasoning` is undefined.
 	 */
 	disableReasoning?: boolean;
+	/**
+	 * If true, request that the provider omit thinking/reasoning summaries
+	 * from the response (e.g. Anthropic `thinking.display = "omitted"`,
+	 * OpenAI Responses `reasoning.summary` left unset). The model still
+	 * reasons internally; only the human-readable summary stream is dropped.
+	 * Useful when the UI hides thinking blocks anyway and the summary is wasted bandwidth.
+	 */
+	hideThinkingSummary?: boolean;
 	/** Custom token budgets for thinking levels (token-based providers only) */
 	thinkingBudgets?: ThinkingBudgets;
 	/** Cursor exec handlers for local tool execution */