@oh-my-pi/pi-ai 18.0.7 → 18.0.8

package/CHANGELOG.md

@@ -2,6 +2,17 @@
 
 ## [Unreleased]
 
+## [18.0.8] - 2026-08-27
+
+### Added
+
+- Added Z.AI GLM Coding Plan usage tracking: credit-based `CREDIT_LIMIT` windows (5h + weekly) now surface in `omp usage` and the status line with the plan tier (`plan: lite/pro/max`).
+
+### Fixed
+
+- Fixed Amazon Bedrock requests to OpenAI-schema models (the `gpt-5.x` SKUs) failing with HTTP 400 `unknown_parameter: 'thinking'` when reasoning was enabled, by sending `reasoning.effort` instead of Anthropic's `thinking` budget block for models the catalog marks as effort-controlled.
+- Fixed Cursor replay rejecting sessions with orphaned tool results while preserving their output as assistant context.
+
 ## [18.0.7] - 2026-08-26
 
 ### Added

package/dist/types/types.d.ts

@@ -698,6 +698,10 @@ export interface DeveloperMessage {
     content: string | (TextContent | ImageContent)[];
     /** Who initiated this message for billing/attribution semantics. */
     attribution?: MessageAttribution;
+    /** True if the message was injected by the system (e.g., auto-continue) and initiates a fresh run rather than continuing the current one. */
+    synthetic?: boolean;
+    /** True when the synthetic prompt was a deliberate operator action (`.`, `c` continue shortcut) rather than an automatic continuation — its timestamp is the turn's prompt time. */
+    userInitiated?: boolean;
     /** Provider-specific opaque payload used to reconstruct transport-native history. */
     providerPayload?: ProviderPayload;
     timestamp: number;
@@ -781,6 +785,8 @@ export interface AssistantMessage {
     timestamp: number;
     duration?: number;
     ttft?: number;
+    /** Local wall-clock time the response finished streaming (ms since epoch); stamped by the session at message_end so prompt→yield timing never depends on provider-reported duration. */
+    completedAt?: number;
 }
 export interface ToolResultMessage<TDetails = unknown> {
     role: "toolResult";

package/dist/types/usage.d.ts

@@ -1,5 +1,5 @@
 import type { FetchImpl, Provider } from "./types.js";
-export type UsageUnit = "percent" | "tokens" | "requests" | "usd" | "minutes" | "bytes" | "unknown";
+export type UsageUnit = "percent" | "tokens" | "requests" | "credits" | "usd" | "minutes" | "bytes" | "unknown";
 export type UsageStatus = "ok" | "warning" | "exhausted" | "unknown";
 /** Time window for a limit (e.g. 5h, 7d, monthly). */
 export interface UsageWindow {
@@ -204,7 +204,7 @@ export interface ClientUsageClientSummary {
 export interface ClientUsageSummary {
     clients: ClientUsageClientSummary[];
 }
-export declare const usageUnitSchema: import("@oh-my-pi/omptype").FluentType<"bytes" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd", "bytes" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd">;
+export declare const usageUnitSchema: import("@oh-my-pi/omptype").FluentType<"bytes" | "credits" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd", "bytes" | "credits" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd">;
 export declare const usageStatusSchema: import("@oh-my-pi/omptype").FluentType<"exhausted" | "ok" | "unknown" | "warning", "exhausted" | "ok" | "unknown" | "warning">;
 export declare const usageWindowSchema: import("@oh-my-pi/omptype").FluentType<{
     durationMs?: number | undefined;
@@ -223,14 +223,14 @@ export declare const usageAmountSchema: import("@oh-my-pi/omptype").FluentType<{
     limit?: number | undefined;
     remaining?: number | undefined;
     remainingFraction?: number | undefined;
-    unit: "bytes" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd";
+    unit: "bytes" | "credits" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd";
     used?: number | undefined;
     usedFraction?: number | undefined;
 }, {
     limit?: number | undefined;
     remaining?: number | undefined;
     remainingFraction?: number | undefined;
-    unit: "bytes" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd";
+    unit: "bytes" | "credits" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd";
     used?: number | undefined;
     usedFraction?: number | undefined;
 }>;
@@ -258,7 +258,7 @@ export declare const usageLimitSchema: import("@oh-my-pi/omptype").FluentType<{
         limit?: number | undefined;
         remaining?: number | undefined;
         remainingFraction?: number | undefined;
-        unit: "bytes" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd";
+        unit: "bytes" | "credits" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd";
         used?: number | undefined;
         usedFraction?: number | undefined;
     };
@@ -288,7 +288,7 @@ export declare const usageLimitSchema: import("@oh-my-pi/omptype").FluentType<{
         limit?: number | undefined;
         remaining?: number | undefined;
         remainingFraction?: number | undefined;
-        unit: "bytes" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd";
+        unit: "bytes" | "credits" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd";
         used?: number | undefined;
         usedFraction?: number | undefined;
     };
@@ -345,7 +345,7 @@ export declare const usageReportSchema: import("@oh-my-pi/omptype").FluentType<{
             limit?: number | undefined;
             remaining?: number | undefined;
             remainingFraction?: number | undefined;
-            unit: "bytes" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd";
+            unit: "bytes" | "credits" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd";
             used?: number | undefined;
             usedFraction?: number | undefined;
         };
@@ -390,7 +390,7 @@ export declare const usageReportSchema: import("@oh-my-pi/omptype").FluentType<{
             limit?: number | undefined;
             remaining?: number | undefined;
             remainingFraction?: number | undefined;
-            unit: "bytes" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd";
+            unit: "bytes" | "credits" | "minutes" | "percent" | "requests" | "tokens" | "unknown" | "usd";
             used?: number | undefined;
             usedFraction?: number | undefined;
         };

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-ai",
-	"version": "18.0.7",
+	"version": "18.0.8",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"homepage": "https://omp.sh",
 	"author": "Stencil Labs, Inc.",
@@ -37,10 +37,10 @@
 		"fmt": "biome format --write ."
 	},
 	"dependencies": {
-		"@oh-my-pi/omptype": "18.0.7",
-		"@oh-my-pi/pi-catalog": "18.0.7",
-		"@oh-my-pi/pi-utils": "18.0.7",
-		"@oh-my-pi/pi-wire": "18.0.7"
+		"@oh-my-pi/omptype": "18.0.8",
+		"@oh-my-pi/pi-catalog": "18.0.8",
+		"@oh-my-pi/pi-utils": "18.0.8",
+		"@oh-my-pi/pi-wire": "18.0.8"
 	},
 	"devDependencies": {
 		"@types/bun": "^1.3.14"

package/src/auth-broker/wire-schemas.ts

@@ -208,7 +208,7 @@ const usageAmountSchema = type({
 	"remaining?": "number",
 	"usedFraction?": "number",
 	"remainingFraction?": "number",
-	unit: "'percent' | 'tokens' | 'requests' | 'usd' | 'minutes' | 'bytes' | 'unknown'",
+	unit: "'percent' | 'tokens' | 'requests' | 'credits' | 'usd' | 'minutes' | 'bytes' | 'unknown'",
 });
 
 const usageScopeSchema = type({

package/src/providers/amazon-bedrock.ts

@@ -1089,6 +1089,15 @@ function buildAdditionalModelRequestFields(
 		};
 	}
 
+	if (mode === "effort") {
+		// OpenAI-schema models on Bedrock (the GPT-5.x SKUs) reject the
+		// Anthropic budget block with `unknown_parameter: 'thinking'` and take
+		// `reasoning.effort` instead — same effort vocabulary the catalog
+		// already bakes (low/medium/high/xhigh/max).
+		const level = requireSupportedEffort(model, reasoning);
+		return { reasoning: { effort: model.thinking?.effortMap?.[level] ?? level } };
+	}
+
 	const level = requireSupportedEffort(model, reasoning);
 	const defaultBudgets: Record<Effort, number> = {
 		minimal: 1024,

package/src/providers/cursor.ts

@@ -4844,6 +4844,27 @@ export function buildCursorSystemPromptJsons(systemPrompt: readonly string[] | u
 	return systemPrompts.map(content => JSON.stringify({ role: "system", content }));
 }
 
+function collectCursorToolHistory(messages: Message[], historyEnd: number) {
+	const toolResults = new Map<string, ToolResultMessage>();
+	const pairedToolCallIds = new Set<string>();
+	for (let index = 0; index < historyEnd; index++) {
+		const message = messages[index];
+		if (message.role === "toolResult") {
+			toolResults.set(message.toolCallId, message);
+		} else if (message.role === "assistant") {
+			for (const item of message.content) {
+				if (item.type === "toolCall") pairedToolCallIds.add(item.id);
+			}
+		}
+	}
+	return { toolResults, pairedToolCallIds };
+}
+
+function cursorOrphanToolResultText(result: ToolResultMessage): string {
+	const prefix = result.isError ? "[Tool Error]" : "[Tool Result]";
+	return `${prefix}\n${toolResultToText(result) || "(empty result)"}`;
+}
+
 function buildRootPromptMessagesJson(
 	messages: Message[],
 	systemPromptIds: Uint8Array[],
@@ -4852,6 +4873,8 @@ function buildRootPromptMessagesJson(
 	targetModelId?: string,
 ): Uint8Array[] {
 	assertCursorKimiK3HistoryReplayable(messages, activeUserMessageIndex, targetModelId);
+	const historyEnd = activeUserMessageIndex >= 0 ? activeUserMessageIndex : messages.length;
+	const { pairedToolCallIds } = collectCursorToolHistory(messages, historyEnd);
 	const entries: Uint8Array[] = [...systemPromptIds];
 	const pushJson = (obj: unknown) => {
 		const bytes = new TextEncoder().encode(JSON.stringify(obj));
@@ -4870,6 +4893,13 @@ function buildRootPromptMessagesJson(
 			if (content.length === 0) continue;
 			pushJson({ role: "assistant", content });
 		} else if (msg.role === "toolResult") {
+			if (!pairedToolCallIds.has(msg.toolCallId)) {
+				pushJson({
+					role: "assistant",
+					content: [{ type: "text", text: cursorOrphanToolResultText(msg) }],
+				});
+				continue;
+			}
 			// Emit even when the result text is empty: the assistant `tool-call` is
 			// already in history, so dropping the pair would replay an orphaned call.
 			const toolCallId = normalizeToolCallId(msg.toolCallId);
@@ -5008,18 +5038,7 @@ function buildConversationTurns(
 ): Uint8Array[] {
 	const turns: Uint8Array[] = [];
 	const historyEnd = activeUserMessageIndex >= 0 ? activeUserMessageIndex : messages.length;
-	const toolResults = new Map<string, ToolResultMessage>();
-	const pairedToolCallIds = new Set<string>();
-	for (let index = 0; index < historyEnd; index++) {
-		const message = messages[index];
-		if (message.role === "toolResult") {
-			toolResults.set(message.toolCallId, message);
-		} else if (message.role === "assistant") {
-			for (const item of message.content) {
-				if (item.type === "toolCall") pairedToolCallIds.add(item.id);
-			}
-		}
-	}
+	const { toolResults, pairedToolCallIds } = collectCursorToolHistory(messages, historyEnd);
 
 	let i = 0;
 	while (i < messages.length) {
@@ -5077,17 +5096,13 @@ function buildConversationTurns(
 					stepBlobIds.push(storeCursorBlob(blobStore, toBinary(ConversationStepSchema, step)));
 				}
 			} else if (stepMsg.role === "toolResult" && !pairedToolCallIds.has(stepMsg.toolCallId)) {
-				const text = toolResultToText(stepMsg);
-				if (text) {
-					const prefix = stepMsg.isError ? "[Tool Error]" : "[Tool Result]";
-					const step = create(ConversationStepSchema, {
-						message: {
-							case: "assistantMessage",
-							value: create(AssistantMessageSchema, { text: `${prefix}\n${text}` }),
-						},
-					});
-					stepBlobIds.push(storeCursorBlob(blobStore, toBinary(ConversationStepSchema, step)));
-				}
+				const step = create(ConversationStepSchema, {
+					message: {
+						case: "assistantMessage",
+						value: create(AssistantMessageSchema, { text: cursorOrphanToolResultText(stepMsg) }),
+					},
+				});
+				stepBlobIds.push(storeCursorBlob(blobStore, toBinary(ConversationStepSchema, step)));
 			}
 			i++;
 		}

package/src/stream.ts

@@ -2080,8 +2080,8 @@ function mapOptionsForApi<TApi extends Api>(
 				guardrailVersion: model.guardrailVersion ?? options?.guardrailVersion,
 				guardrailTrace: model.guardrailTrace ?? options?.guardrailTrace,
 			};
-			// Adaptive mode sends effort directly, no budget_tokens — skip budget inflation.
-			if (model.thinking?.mode === "anthropic-adaptive") {
+			// Effort modes send effort directly, no budget_tokens — skip budget inflation.
+			if (model.thinking?.mode === "effort" || model.thinking?.mode === "anthropic-adaptive") {
 				return castApi<"bedrock-converse-stream">(bedrockBase);
 			}
 			const budgetInfo = resolveBedrockThinkingBudget(model as Model<"bedrock-converse-stream">, options);

package/src/types.ts

@@ -879,6 +879,10 @@ export interface DeveloperMessage {
 	content: string | (TextContent | ImageContent)[];
 	/** Who initiated this message for billing/attribution semantics. */
 	attribution?: MessageAttribution;
+	/** True if the message was injected by the system (e.g., auto-continue) and initiates a fresh run rather than continuing the current one. */
+	synthetic?: boolean;
+	/** True when the synthetic prompt was a deliberate operator action (`.`, `c` continue shortcut) rather than an automatic continuation — its timestamp is the turn's prompt time. */
+	userInitiated?: boolean;
 	/** Provider-specific opaque payload used to reconstruct transport-native history. */
 	providerPayload?: ProviderPayload;
 	timestamp: number; // Unix timestamp in milliseconds
@@ -976,6 +980,8 @@ export interface AssistantMessage {
 	timestamp: number; // Unix timestamp in milliseconds
 	duration?: number; // Request duration in milliseconds
 	ttft?: number; // Time to first token in milliseconds
+	/** Local wall-clock time the response finished streaming (ms since epoch); stamped by the session at message_end so prompt→yield timing never depends on provider-reported duration. */
+	completedAt?: number;
 }
 
 export interface ToolResultMessage<TDetails = unknown> {

package/src/usage/zai.ts

@@ -51,6 +51,8 @@ interface ZaiQuotaPayload {
 	msg?: string;
 	data?: {
 		limits?: ZaiUsageLimitItem[];
+		/** Coding-plan tier (e.g. "lite", "pro", "max") surfaced as the plan label. */
+		level?: string;
 	};
 }
 
@@ -193,17 +195,33 @@ function buildModelUsageUrl(baseUrl: string, now: Date): string {
 }
 
 function getZaiCredentialLimits(report: UsageReport): UsageLimit[] {
-	const limits = report.limits.filter(
-		limit => limit.id.startsWith("zai:requests:") || limit.id.startsWith("zai:tokens:"),
+	return report.limits.filter(
+		limit =>
+			limit.id.startsWith("zai:requests:") ||
+			limit.id.startsWith("zai:tokens:") ||
+			limit.id.startsWith("zai:credits:"),
 	);
-	return limits;
+}
+
+function zaiLimitPressure(limit: UsageLimit): number {
+	const fraction = limit.amount.usedFraction;
+	return typeof fraction === "number" && Number.isFinite(fraction) ? fraction : -1;
 }
 
 function rankZaiRequestLimits(report: UsageReport): UsageLimit[] {
 	const requestLimits = report.limits.filter(limit => limit.id.startsWith("zai:requests:"));
 	const credentialLimits = getZaiCredentialLimits(report);
 	const limits = requestLimits.length > 0 ? requestLimits : credentialLimits;
-	const ranked = [...limits];
+	// Mixed-meter payloads (tokens + credits on the same plan) can repeat a
+	// window; keep the most-binding limit per window so a second 5h row never
+	// displaces the weekly window when primary/secondary are picked positionally.
+	const byWindow = new Map<number, UsageLimit>();
+	for (const limit of limits) {
+		const durationMs = limit.window?.durationMs ?? Number.POSITIVE_INFINITY;
+		const current = byWindow.get(durationMs);
+		if (!current || zaiLimitPressure(limit) > zaiLimitPressure(current)) byWindow.set(durationMs, limit);
+	}
+	const ranked = [...byWindow.values()];
 	ranked.sort((left, right) => {
 		const leftDuration = left.window?.durationMs ?? Number.POSITIVE_INFINITY;
 		const rightDuration = right.window?.durationMs ?? Number.POSITIVE_INFINITY;
@@ -306,6 +324,33 @@ async function fetchZaiUsage(params: UsageFetchParams, ctx: UsageFetchContext):
 				status: getUsageStatus(amount.usedFraction),
 			});
 		}
+		if (parsed.type === "CREDIT_LIMIT") {
+			// GLM Coding Plan windows (e.g. 12k credits / 5h + 60k credits / week):
+			// `usage` is the plan's credit allotment, `currentValue` the spend.
+			// `percentage` is a server-rounded integer (11 for 1438/12000 ≈ 11.98%),
+			// so prefer the exact ratio and fall back to it only without absolutes.
+			const window = buildZaiWindow(parsed);
+			const hasAbsoluteMeter = parsed.currentValue !== undefined && parsed.usage !== undefined && parsed.usage > 0;
+			const amount = buildUsageAmount({
+				used: parsed.currentValue,
+				limit: parsed.usage,
+				remaining: parsed.remaining,
+				percentage: hasAbsoluteMeter ? undefined : parsed.percentage,
+				unit: "credits",
+			});
+			limits.push({
+				id: `zai:credits:${window.id}`,
+				label: `ZAI ${window.label} Credit Quota`,
+				scope: {
+					provider: params.provider,
+					windowId: window.id,
+					shared: true,
+				},
+				window,
+				amount,
+				status: getUsageStatus(amount.usedFraction),
+			});
+		}
 	}
 
 	if (limits.length === 0) return null;
@@ -318,6 +363,7 @@ async function fetchZaiUsage(params: UsageFetchParams, ctx: UsageFetchContext):
 			endpoint: url,
 			accountId: credential.accountId,
 			email: credential.email,
+			...(typeof payload.data?.level === "string" && payload.data.level ? { planType: payload.data.level } : {}),
 		},
 		raw: payload,
 	};

package/src/usage.ts

@@ -6,7 +6,7 @@
  */
 import { type } from "@oh-my-pi/omptype";
 import type { FetchImpl, Provider } from "./types";
-export type UsageUnit = "percent" | "tokens" | "requests" | "usd" | "minutes" | "bytes" | "unknown";
+export type UsageUnit = "percent" | "tokens" | "requests" | "credits" | "usd" | "minutes" | "bytes" | "unknown";
 
 export type UsageStatus = "ok" | "warning" | "exhausted" | "unknown";
 
@@ -240,7 +240,9 @@ export interface ClientUsageSummary {
 
 // ─── Zod schemas (wire-shape validation for the broker `/v1/usage` endpoint) ─
 
-export const usageUnitSchema = type("'percent' | 'tokens' | 'requests' | 'usd' | 'minutes' | 'bytes' | 'unknown'");
+export const usageUnitSchema = type(
+	"'percent' | 'tokens' | 'requests' | 'credits' | 'usd' | 'minutes' | 'bytes' | 'unknown'",
+);
 export const usageStatusSchema = type("'ok' | 'warning' | 'exhausted' | 'unknown'");
 
 export const usageWindowSchema = type({