@cartanova/qgrid-ai-sdk 2.9.10 → 2.10.1

package/README.ko.md

@@ -212,7 +212,7 @@ const { text } = await generateText({
 | `timeoutMs` | 양의 정수, 최대 `1_800_000` | Anthropic 전용 | 서버의 Claude Code 프로세스 제한시간(ms). SDK의 non-stream HTTP 제한은 이 값보다 60초 길게 설정. 기본값은 240초 |
 | `imageGeneration` | `boolean` | OpenAI 전용, non-stream | codex 내장 `image_generation` tool 활성화 ([아래](#image-generation) 참조) |
 | `imageGenerationOptions` | `{ quality?, size?, background? }` | OpenAI 전용 | 이미지 생성 옵션. `background: "transparent"`는 전용 경로 사용. 실제 크기·품질은 응답 metadata 확인 ([아래](#image-generation)). |
-| `fallbackModels` | `string[]` | 예약 | 향후 qgrid 서버 fallback routing용 예약 필드. 현재 동작하지 않으며 Claude Code의 Fable refusal fallback과 무관 |
+| `fallbackModels` | `string[]` | deprecated | 호환용으로만 남긴 무시되는 필드. 사용하면 unsupported-option 경고를 반환하므로 호출부에서 제거하세요. Claude Code의 Fable refusal fallback과 무관 |
 
 ```typescript
 await generateText({
@@ -300,6 +300,8 @@ const result = await generateText({
 
 ## Telemetry Logger
 
+서버와 AI SDK 2.10.0부터 qgrid provider의 `generateText` 결과에서는 `result.providerMetadata?.qgrid?.requestLogId`로 요청 로그 ID를 확인할 수 있습니다. `streamText`는 스트림 소비 후 `(await result.providerMetadata)?.qgrid?.requestLogId`로 확인합니다. 직접 API를 호출하면 `query` 응답과 스트림 `done` 이벤트의 최상위 `requestLogId`에 같은 값이 담깁니다. `logger: false` 또는 구버전 서버에서는 이 필드가 없으며, 툴 호출 후속 단계는 같은 요청 로그 ID를 공유합니다. 이 ID는 후속 요청 연결 상태가 아니며, 아래 telemetry logger를 쓰는 외부 provider에는 이 metadata가 추가되지 않습니다.
+
 qgrid provider가 아닌 모델(google, openai 직접 호출)에서도 같은 request log 대시보드를 사용하려면 `createQgridLogger`를 `experimental_telemetry`에 넣으면 됩니다.
 
 ```typescript
@@ -398,6 +400,7 @@ type QgridSupportedModel =
   | "anthropic/claude-opus-5"
   | "anthropic/claude-opus-5-5"
   | "anthropic/claude-sonnet-5-5"
+  | "anthropic/claude-haiku-5-5"
 ```
 
 `openai/gpt-5.4`, `openai/gpt-5.4-mini`, `openai/gpt-5.2`, `openai/gpt-5.3-codex`는 하위 호환을 위해 타입에 남아 있지만, qgrid가 사용하는 ChatGPT 구독 Codex 경로에서는 더 이상 제공되지 않습니다. `gpt-5.4`와 `gpt-5.4-mini`는 2026-08-31에 retire되었고(대체: `openai/gpt-5.6-terra`, `openai/gpt-5.6-luna`), `gpt-5.2`와 `gpt-5.3-codex`는 그보다 먼저 해당 경로에서 제거되었습니다. 이 id로 요청하면 백엔드에서 실패합니다.
@@ -433,7 +436,9 @@ Qgrid 비용 추정에는 [표준 API 단가](https://developers.openai.com/api/
 
 `anthropic/claude-opus-5-5`(2026-09-22 출시)는 1M context와 128K 최대 출력을 지원합니다. 1M tokens당 단가는 input $4, cache read $0.20(input의 0.05x), 5분 cache write $5, 1시간 cache write $8, output $20입니다. adaptive thinking이 항상 켜져 있고 끌 수 없으므로 qgrid는 Fable과 같이 이를 보존하고 `effort`로 추론 깊이를 조절합니다. Opus 5.5의 API 수준 파괴적 변경(forced `tool_choice` 거부, 모델에 귀속된 thinking 블록)은 Fable 5.1과 같은 이유로 qgrid에 영향을 주지 않습니다. Opus 5 단가는 변동이 없습니다.
 
-`anthropic/claude-sonnet-5-5`(2026-09-29 출시)는 1M context와 128K 최대 출력을 지원하며 Sonnet 5와 같은 단가입니다. 1M tokens당 input $2, cache read $0.20, 5분 cache write $2.50, 1시간 cache write $4, output $10입니다. thinking 비활성화를 거부하므로 qgrid는 Opus 5.5와 같이 adaptive thinking을 유지하고 `effort`로 추론 깊이를 조절합니다.
+`anthropic/claude-sonnet-5-5`(2026-09-28 출시)는 1M context와 128K 최대 출력을 지원하며 Sonnet 5와 같은 기본 단가입니다. 1M tokens당 input $2, 5분 cache write $2.50, 1시간 cache write $4, output $10이고, cache read는 2026-10-07에 $0.20에서 절반으로 인하되어 Opus 5.5와 같은 0.05x인 $0.10입니다. qgrid는 요청 시점 단가로 비용을 기록하므로 그 이전 로그는 $0.20 추정치를 유지합니다. thinking 비활성화를 거부하므로 qgrid는 Opus 5.5와 같이 adaptive thinking을 유지하고 `effort`로 추론 깊이를 조절합니다.
+
+`anthropic/claude-haiku-5-5`(2026-10-07 출시)는 1M context와 128K 최대 출력을 지원하며, 현행 Claude 모델 중 유일하게 프롬프트 길이별 단가를 씁니다. 100K 토큰 이하 프롬프트는 1M tokens당 input $0.10, cache read $0.01, 5분 cache write $0.125, 1시간 cache write $0.20, output $0.50이고, 프롬프트가 100K 토큰을 넘으면 모든 단가가 5배($0.50 / $0.05 / $0.625 / $1 / $2.50)로 요청 전체에 적용됩니다. qgrid는 GPT-6에 쓰는 요청 전체 long-context 규칙으로 이를 계산합니다. adaptive thinking이 기본으로 켜져 있고 API 기본 effort는 `medium`이며, API는 `xhigh`/`max`에서 thinking 비활성화를 거부하고 Claude Code 2.1.293 카탈로그도 `rejects_disabled_thinking`으로 표시하므로 qgrid는 adaptive thinking을 유지하고 `effort`(SDK 기본 `low`)로 추론 깊이를 조절합니다. Haiku 5.5는 Claude 4.7+ 토크나이저를 쓰므로 같은 텍스트가 Haiku 4.5보다 약 30% 더 많은 토큰으로 계산됩니다.
 
 Claude Code는 Fable의 safety refusal을 다른 Opus 모델로 자동 재시도할 수 있습니다. 현재 CLI는 refusal 카테고리에 따라 Opus 5 또는 Opus 4.8을 고릅니다. 이 경우 AI SDK 응답의 `response.modelId`와 `providerMetadata.qgrid.model`은 실제 serving 모델인 Opus를 가리킵니다. `providerMetadata.qgrid.requestedModel`은 Fable로 유지되고, `providerMetadata.qgrid.modelFallbacks`에 refusal fallback 이력이 담깁니다. 같은 metadata에서 `costSource`와 5분/1시간 cache-write 토큰 분해도 확인할 수 있습니다.
 
@@ -489,6 +494,21 @@ qgrid(modelId, {
   한도인 64 KiB도 넘지 않아야 합니다.
 - AI SDK의 `toolChoice`는 현재 qgrid에서 지원하지 않습니다.
 
+## 저장소 테스트
+
+`mise run test`는 provider 계정·DB 없이 API·SDK·CLI 테스트를 실행합니다. OAuth relay 테스트는 로컬 loopback 포트를 사용합니다. DB 통합 검증은 PostgreSQL 테스트 템플릿을 준비한 뒤 `mise exec -- pnpm --dir packages/api test:db`로 별도 실행합니다. API의 `test` 명령은 두 suite를 모두 실행합니다.
+
+실제 provider E2E는 토큰이 등록된 격리 qgrid 서버와 명시적인 opt-in이 필요합니다.
+
+```bash
+QGRID_REAL_PROVIDER_ACCEPTANCE=1 QGRID_URL=http://localhost:44900 QGRID_MODEL=openai/gpt-6.1-sol \
+  mise exec -- pnpm --dir packages/ai-sdk e2e
+```
+
+기본 suite는 일반 응답·스트리밍·툴+구조화 출력과 반환된 ID의 요청 로그를 확인합니다. Anthropic 경로는 해당 모델을 지정하고, 두 경로를 모두 검사하려면 모델을 쉼표로 나열합니다. `e2e:tools-output`은 구조화 출력의 집중 회귀 검증, `e2e:logger`는 telemetry가 native 로그를 중복 생성하지 않는지 확인합니다. 모두 같은 opt-in·서버·모델 설정을 요구하며 실제 쿼터를 사용합니다. 기본 CI에는 포함하지 않고, 실패를 자동 재시도로 감추거나 서버 로그를 삭제하지 않습니다.
+
+테스트는 외부에서 관측되는 계약이나 구체적인 회귀 위험에 한해 추가합니다. 실패 경로·비용·데이터 무결성 검증은 유지하고, 중복 정상 경로를 통합합니다. 새 범용 harness는 만들지 않습니다. DB·provider 환경이 없으면 성공이 아니라 미검증으로 기록합니다.
+
 ## 요구사항
 
 - Node.js >= 20

package/README.md

@@ -221,7 +221,7 @@ const { text } = await generateText({
 | `timeoutMs` | positive integer, max `1_800_000` | Anthropic only | Server-side Claude Code process timeout in milliseconds. The SDK's non-stream HTTP budget is 60 seconds longer. Defaults to 240 seconds |
 | `imageGeneration` | `boolean` | OpenAI only, non-stream | Enables codex's built-in `image_generation` tool (see [below](#image-generation)) |
 | `imageGenerationOptions` | `{ quality?, size?, background?, outputFormat? }` | OpenAI only | Image controls; `outputFormat` selects PNG/JPEG/WebP on Responses. `background: "transparent"` selects standalone Images (PNG only). Read actual size/quality from response metadata ([below](#image-generation)). |
-| `fallbackModels` | `string[]` | reserved | Reserved for future qgrid server-side fallback routing. Not functional yet and unrelated to Claude Code's Fable refusal fallback |
+| `fallbackModels` | `string[]` | deprecated | Ignored compatibility field; emits an unsupported-option warning. Remove it from callers. Unrelated to Claude Code's Fable refusal fallback |
 
 ```typescript
 await generateText({
@@ -335,6 +335,8 @@ Remove green-screen or other background instructions when requesting transparenc
 
 ## Telemetry Logger
 
+With qgrid server and SDK 2.10.0+, read the parent request-log ID from `result.providerMetadata?.qgrid?.requestLogId` after `generateText`. For `streamText`, consume the stream first, then read `(await result.providerMetadata)?.qgrid?.requestLogId`. Raw API query responses and stream `done` events expose the same ID as top-level `requestLogId`. The field is omitted with `logger: false` or an older server. Tool-run steps share the parent ID; it is not continuation state. This metadata applies to the qgrid provider, not external providers using the telemetry logger below.
+
 To use the same request log dashboard with models that don't go through the qgrid provider (direct google/openai calls), pass `createQgridLogger` to `experimental_telemetry`.
 
 ```typescript
@@ -434,6 +436,7 @@ type QgridSupportedModel =
   | "anthropic/claude-opus-5"
   | "anthropic/claude-opus-5-5"
   | "anthropic/claude-sonnet-5-5"
+  | "anthropic/claude-haiku-5-5"
 ```
 
 `openai/gpt-5.4`, `openai/gpt-5.4-mini`, `openai/gpt-5.2`, and `openai/gpt-5.3-codex` remain in the type for backward compatibility, but the ChatGPT-subscription Codex route that qgrid uses no longer serves them. `gpt-5.4` and `gpt-5.4-mini` retired on 2026-08-31 (replacements: `openai/gpt-5.6-terra` and `openai/gpt-5.6-luna`); `gpt-5.2` and `gpt-5.3-codex` were removed from that route earlier. Requests for these ids fail at the backend.
@@ -469,7 +472,9 @@ Qgrid uses [Standard API pricing](https://developers.openai.com/api/docs/pricing
 
 `anthropic/claude-opus-5-5` (released 2026-09-22) has a 1M context window and 128K max output. Its prices per 1M tokens are $4 input, $0.20 cache read (0.05x the input price), $5 five-minute cache write, $8 one-hour cache write, and $20 output. Adaptive thinking is always on and cannot be disabled, so qgrid preserves it and uses `effort` to control depth, as it does for Fable. Opus 5.5's API-level breaking changes (forced `tool_choice` rejection, model-bound thinking blocks) do not affect qgrid for the same reason as Fable 5.1. Opus 5 pricing did not change.
 
-`anthropic/claude-sonnet-5-5` (released 2026-09-29) has a 1M context window and 128K max output at Sonnet 5's prices: $2 input, $0.20 cache read, $2.50 five-minute cache write, $4 one-hour cache write, and $10 output per 1M tokens. It rejects disabled thinking, so qgrid keeps adaptive thinking on and uses `effort` to control depth, as it does for Opus 5.5.
+`anthropic/claude-sonnet-5-5` (released 2026-09-28) has a 1M context window and 128K max output at Sonnet 5's base prices: $2 input, $2.50 five-minute cache write, $4 one-hour cache write, and $10 output per 1M tokens. Its cache reads cost $0.10 (0.05x the input price, the same multiplier as Opus 5.5) since Anthropic halved them from $0.20 on 2026-10-07; qgrid prices each request at the rate current when it runs, so logs from before that date keep the $0.20 estimate. It rejects disabled thinking, so qgrid keeps adaptive thinking on and uses `effort` to control depth, as it does for Opus 5.5.
+
+`anthropic/claude-haiku-5-5` (released 2026-10-07) has a 1M context window and 128K max output, and is the only current Claude model priced by prompt length. For prompts up to 100K tokens it costs $0.10 input, $0.01 cache read, $0.125 five-minute cache write, $0.20 one-hour cache write, and $0.50 output per 1M tokens; once the prompt exceeds 100K tokens, every rate is 5x ($0.50 / $0.05 / $0.625 / $1 / $2.50) and applies to the whole request, which qgrid models with the same whole-request long-context rule it uses for GPT-6. Adaptive thinking is on by default with a `medium` API default effort; the API rejects disabled thinking at `xhigh` and `max`, and Claude Code 2.1.293 marks the model `rejects_disabled_thinking`, so qgrid keeps adaptive thinking on and uses `effort` (SDK default `low`) to control depth. Haiku 5.5 uses the Claude 4.7+ tokenizer, so the same text counts roughly 30% more tokens than on Haiku 4.5.
 
 Claude Code may automatically retry a Fable safety refusal on another Opus model; the current CLI picks Opus 5 or Opus 4.8 by refusal category. In that case, the AI SDK response's `response.modelId` and `providerMetadata.qgrid.model` identify Opus as the actual serving model. `providerMetadata.qgrid.requestedModel` remains Fable, and `providerMetadata.qgrid.modelFallbacks` contains the refusal fallback history. The metadata also exposes `costSource` and the 5m/1h cache-write token split.
 
@@ -528,6 +533,21 @@ If multiple projects/workflows share one qgrid server, set `QGRID_PROJECT_NAME`.
   single-argument budget used by the Claude Code transport.
 - AI SDK `toolChoice` is not currently supported by qgrid.
 
+## Repository testing
+
+Run `mise run test` for DB-free API, SDK, and CLI tests. The OAuth relay test uses a local loopback port; no provider credentials are needed. Run `mise exec -- pnpm --dir packages/api test:db` separately with the prepared PostgreSQL test template. The API's `test` command still runs both suites.
+
+Real-provider E2E is opt-in and requires an isolated qgrid server with registered tokens:
+
+```bash
+QGRID_REAL_PROVIDER_ACCEPTANCE=1 QGRID_URL=http://localhost:44900 QGRID_MODEL=openai/gpt-6.1-sol \
+  mise exec -- pnpm --dir packages/ai-sdk e2e
+```
+
+The default suite checks text, streaming, and tool-plus-structured output, following returned request-log IDs to persisted results. Set an Anthropic model for that route, or use a comma-separated model list to explicitly run both. `e2e:tools-output` is a focused structured-output regression probe; `e2e:logger` checks native logging is not duplicated by telemetry. All live commands consume quota and require the same explicit opt-in/server/model settings. They are not part of ordinary CI, do not retry failures automatically, and do not delete server logs.
+
+Keep new tests focused on observable contracts or concrete regressions. Preserve failure-path, cost, and data-integrity checks; consolidate duplicate happy paths instead of adding more mocks or a new test harness. Missing DB/provider infrastructure means unverified, not passed.
+
 ## Requirements
 
 - Node.js >= 20

package/dist/index.d.ts

@@ -39,8 +39,8 @@ type QgridCommonProviderOptions = {
    */
   logger?: boolean;
   /**
-   * @todo 향후 qgrid 서버 fallback routing에 사용할 후보 모델 목록.
-   * Claude Code가 처리하는 Fable safety-refusal fallback과는 무관하다.
+   * @deprecated Ignored: qgrid does not implement caller-configured model fallback routing.
+   * Retained for source compatibility. Claude Code's safety fallback is independent.
    */
   fallbackModels?: string[];
 };
@@ -96,7 +96,7 @@ type QgridAnthropicProviderOptions = QgridCommonProviderOptions & {
  * `QgridOpenAIProviderOptions` / `QgridAnthropicProviderOptions` 를 직접 쓰는 편이 정확하다.
  */
 type QgridProviderOptions = QgridOpenAIProviderOptions | QgridAnthropicProviderOptions;
-type QgridSupportedModel = "openai/gpt-6-astra" | "openai/gpt-6.1-sol" | "openai/gpt-6-sol" | "openai/gpt-6-luna" | "openai/gpt-5.6-sol" | "openai/gpt-5.6-terra" | "openai/gpt-5.6-luna" | "openai/gpt-5.5" | "openai/gpt-5.4" | "openai/gpt-5.2" | "openai/gpt-5.4-mini" | "openai/gpt-5.3-codex" | "openai/gpt-5.3-codex-spark" | "anthropic/claude-fable-5-1" | "anthropic/claude-fable-5" | "anthropic/claude-haiku-4-5" | "anthropic/claude-sonnet-4" | "anthropic/claude-sonnet-4-5" | "anthropic/claude-sonnet-4-6" | "anthropic/claude-sonnet-4-7" | "anthropic/claude-sonnet-5" | "anthropic/claude-opus-4" | "anthropic/claude-opus-4-1" | "anthropic/claude-opus-4-5" | "anthropic/claude-opus-4-6" | "anthropic/claude-opus-4-7" | "anthropic/claude-opus-4-8" | "anthropic/claude-opus-5" | "anthropic/claude-opus-5-5" | "anthropic/claude-sonnet-5-5";
+type QgridSupportedModel = "openai/gpt-6-astra" | "openai/gpt-6.1-sol" | "openai/gpt-6-sol" | "openai/gpt-6-luna" | "openai/gpt-5.6-sol" | "openai/gpt-5.6-terra" | "openai/gpt-5.6-luna" | "openai/gpt-5.5" | "openai/gpt-5.4" | "openai/gpt-5.2" | "openai/gpt-5.4-mini" | "openai/gpt-5.3-codex" | "openai/gpt-5.3-codex-spark" | "anthropic/claude-fable-5-1" | "anthropic/claude-fable-5" | "anthropic/claude-haiku-4-5" | "anthropic/claude-sonnet-4" | "anthropic/claude-sonnet-4-5" | "anthropic/claude-sonnet-4-6" | "anthropic/claude-sonnet-4-7" | "anthropic/claude-sonnet-5" | "anthropic/claude-opus-4" | "anthropic/claude-opus-4-1" | "anthropic/claude-opus-4-5" | "anthropic/claude-opus-4-6" | "anthropic/claude-opus-4-7" | "anthropic/claude-opus-4-8" | "anthropic/claude-opus-5" | "anthropic/claude-opus-5-5" | "anthropic/claude-sonnet-5-5" | "anthropic/claude-haiku-5-5";
 /** Observed image response values. Usage is response-level and appears on the first image only. */
 type QgridImageGenerationMetadata = {
   route: "codex-images";

package/dist/index.js

@@ -396,27 +396,41 @@ function extractToolResultsFromHistory(messages) {
 	});
 	return out;
 }
-async function* parseSSE(body) {
+async function* parseSSE(body, signal) {
 	const reader = body.getReader();
 	const decoder = new TextDecoder();
 	let buffer = "";
 	let eventType = "";
-	for (;;) {
-		const { done, value } = await reader.read();
-		if (done) break;
-		buffer += decoder.decode(value, { stream: true });
-		const lines = buffer.split("\n");
-		buffer = lines.pop() ?? "";
-		for (const line of lines) if (line.startsWith("event: ")) eventType = line.slice(7).trim();
-		else if (line.startsWith("data: ")) {
-			const raw = line.slice(6);
-			try {
-				yield {
-					type: eventType || "message",
-					data: JSON.parse(raw)
-				};
-			} catch {}
-			eventType = "";
+	const cancel = () => {
+		reader.cancel(signal?.reason).catch(() => void 0);
+	};
+	signal?.addEventListener("abort", cancel, { once: true });
+	try {
+		signal?.throwIfAborted();
+		for (;;) {
+			const { done, value } = await reader.read();
+			if (done) break;
+			buffer += decoder.decode(value, { stream: true });
+			const lines = buffer.split("\n");
+			buffer = lines.pop() ?? "";
+			for (const line of lines) if (line.startsWith("event: ")) eventType = line.slice(7).trim();
+			else if (line.startsWith("data: ")) {
+				const raw = line.slice(6);
+				try {
+					yield {
+						type: eventType || "message",
+						data: JSON.parse(raw)
+					};
+				} catch {}
+				eventType = "";
+			}
+		}
+	} finally {
+		signal?.removeEventListener("abort", cancel);
+		try {
+			await reader.cancel();
+		} catch {} finally {
+			reader.releaseLock();
 		}
 	}
 }
@@ -1029,6 +1043,7 @@ function createQgridLogger(config = {}) {
 //#region src/index.ts
 const threadCoordStore = /* @__PURE__ */ new Map();
 const CACHE_AFFINITY_NAMESPACE = "qgrid-cache-affinity-v1";
+const MAX_THREAD_COORDS = 1e3;
 const MAX_IMAGE_INPUT_DATA_URL_CHARS = 9e6;
 const MAX_IMAGE_INPUT_DATA_URL_CHARS_LABEL = "9M";
 function assertImageInputFitsJsonTransport(imageUrls) {
@@ -1064,10 +1079,21 @@ function getThreadCoord(sessionKey) {
 	return entry.coord;
 }
 function setThreadCoord(sessionKey, coord) {
+	const now = Date.now();
+	for (const [key, entry] of threadCoordStore) if (entry.expiresAt <= now) threadCoordStore.delete(key);
+	threadCoordStore.delete(sessionKey);
 	threadCoordStore.set(sessionKey, {
 		coord,
-		expiresAt: Date.now() + THREAD_COORD_TTL_MS
+		expiresAt: now + THREAD_COORD_TTL_MS
 	});
+	if (threadCoordStore.size > MAX_THREAD_COORDS) threadCoordStore.delete(threadCoordStore.keys().next().value);
+}
+function providerWarnings(options) {
+	return options.fallbackModels === void 0 ? [] : [{
+		type: "unsupported",
+		feature: "providerOptions.qgrid.fallbackModels",
+		details: "fallbackModels is deprecated and ignored; qgrid does not implement model fallback routing."
+	}];
 }
 async function deriveCacheAffinityKey(modelId, sessionKey) {
 	if (!sessionKey || modelId.startsWith("anthropic/")) return void 0;
@@ -1081,6 +1107,7 @@ function qgridProviderMetadata(data) {
 		...item.generation
 	}] : []);
 	return { qgrid: {
+		...data.requestLogId !== void 0 ? { requestLogId: data.requestLogId } : {},
 		model: data.model,
 		requestedModel: data.requestedModel ?? data.model,
 		modelFallbacks: data.modelFallbacks ?? [],
@@ -1263,7 +1290,7 @@ function qgrid(modelId, config) {
 				content,
 				finishReason,
 				usage: toAiSdkUsage(data.usage),
-				warnings: [],
+				warnings: providerWarnings(qgridOptions),
 				providerMetadata: qgridProviderMetadata(data),
 				response: { modelId: data.model }
 			};
@@ -1294,6 +1321,8 @@ function qgrid(modelId, config) {
 			}
 			const storedCoord = cacheAffinityKey ? getThreadCoord(cacheAffinityKey) : void 0;
 			if (!matchedClientRun && storedCoord) runContext = { threadCoord: storedCoord };
+			const abortController = new AbortController();
+			const signal = options.abortSignal ? AbortSignal.any([options.abortSignal, abortController.signal]) : abortController.signal;
 			const prepRes = await fetch(`${serverUrl}/api/qgrid/prepareStream`, {
 				method: "POST",
 				headers: { "Content-Type": "application/json" },
@@ -1316,14 +1345,14 @@ function qgrid(modelId, config) {
 					...runContext ? { runContext } : {},
 					...toolResultsPayload ? { toolResults: toolResultsPayload } : {}
 				} }),
-				signal: options.abortSignal
+				signal
 			});
 			if (!prepRes.ok) {
 				const text = await prepRes.text().catch(() => "");
 				throw new Error(`qgrid prepareStream ${prepRes.status}: ${text}`);
 			}
 			const { streamId } = await prepRes.json();
-			const streamRes = await fetch(`${serverUrl}/api/qgrid/queryStream?streamId=${streamId}`, { signal: options.abortSignal });
+			const streamRes = await fetch(`${serverUrl}/api/qgrid/queryStream?streamId=${streamId}`, { signal });
 			if (!streamRes.ok || !streamRes.body) {
 				const text = await streamRes.text().catch(() => "");
 				throw new Error(`qgrid stream ${streamRes.status}: ${text}`);
@@ -1332,107 +1361,131 @@ function qgrid(modelId, config) {
 			let textStarted = false;
 			let deltaTextEmitted = false;
 			const envelopeParser = hasTools ? createEnvelopeStreamParser(jsonSchema ? "json" : "text") : void 0;
-			return { stream: new ReadableStream({ async start(controller) {
+			let cancelled = false;
+			let processing;
+			const stream = new ReadableStream({
+				start(controller) {
+					processing = processEvents(controller);
+				},
+				async cancel(reason) {
+					cancelled = true;
+					abortController.abort(reason);
+					await processing;
+				}
+			});
+			async function processEvents(controller) {
 				try {
+					controller.enqueue({
+						type: "stream-start",
+						warnings: providerWarnings(qgridOptions)
+					});
 					let streamCompleted = false;
-					for await (const event of parseSSE(streamRes.body)) if (event.type === "delta") {
-						const raw = event.data.text;
-						const text = envelopeParser ? envelopeParser.push(raw) : raw;
-						if (text !== "") {
-							if (!textStarted) {
+					for await (const event of parseSSE(streamRes.body, signal)) {
+						if (cancelled) return;
+						signal.throwIfAborted();
+						if (event.type === "delta") {
+							const raw = event.data.text;
+							const text = envelopeParser ? envelopeParser.push(raw) : raw;
+							if (text !== "") {
+								if (!textStarted) {
+									controller.enqueue({
+										type: "text-start",
+										id: textId
+									});
+									textStarted = true;
+								}
 								controller.enqueue({
-									type: "text-start",
+									type: "text-delta",
+									id: textId,
+									delta: text
+								});
+								deltaTextEmitted = true;
+							}
+						} else if (event.type === "done") {
+							if (textStarted) {
+								controller.enqueue({
+									type: "text-end",
 									id: textId
 								});
-								textStarted = true;
+								textStarted = false;
+							}
+							const done = event.data;
+							if (matchedClientRun) clientRuns.delete(matchedClientRun.key);
+							if (done.finishReason === "tool-calls") rememberClientRun(done.runContext, (done.content ?? []).filter((c) => c.type === "tool-call").map((c) => c.toolCallId));
+							if (cacheAffinityKey && done.runContext?.threadCoord) setThreadCoord(cacheAffinityKey, done.runContext.threadCoord);
+							if (done.content) {
+								for (const item of done.content) if (item.type === "text" && !deltaTextEmitted) {
+									const tid = `${TEXT_STREAM_ID_PREFIX}_${Math.random().toString(36).slice(2, 10)}`;
+									controller.enqueue({
+										type: "text-start",
+										id: tid
+									});
+									controller.enqueue({
+										type: "text-delta",
+										id: tid,
+										delta: item.text
+									});
+									controller.enqueue({
+										type: "text-end",
+										id: tid
+									});
+								} else if (item.type === "tool-call") {
+									controller.enqueue({
+										type: "tool-input-start",
+										id: item.toolCallId,
+										toolName: item.toolName
+									});
+									controller.enqueue({
+										type: "tool-input-delta",
+										id: item.toolCallId,
+										delta: item.input
+									});
+									controller.enqueue({
+										type: "tool-input-end",
+										id: item.toolCallId
+									});
+									controller.enqueue({
+										type: "tool-call",
+										toolCallId: item.toolCallId,
+										toolName: item.toolName,
+										input: item.input
+									});
+								}
 							}
 							controller.enqueue({
-								type: "text-delta",
-								id: textId,
-								delta: text
+								type: "response-metadata",
+								modelId: done.model
 							});
-							deltaTextEmitted = true;
-						}
-					} else if (event.type === "done") {
-						if (textStarted) {
 							controller.enqueue({
-								type: "text-end",
-								id: textId
+								type: "finish",
+								finishReason: done.finishReason === "tool-calls" ? {
+									unified: "tool-calls",
+									raw: "tool_call"
+								} : {
+									unified: "stop",
+									raw: "stop"
+								},
+								usage: toAiSdkUsage(done.usage),
+								providerMetadata: qgridProviderMetadata(done)
 							});
-							textStarted = false;
+							streamCompleted = true;
+							controller.close();
+							return;
+						} else if (event.type === "error") {
+							streamCompleted = true;
+							controller.error(new Error(event.data.message));
+							return;
 						}
-						const done = event.data;
-						if (matchedClientRun) clientRuns.delete(matchedClientRun.key);
-						if (done.finishReason === "tool-calls") rememberClientRun(done.runContext, (done.content ?? []).filter((c) => c.type === "tool-call").map((c) => c.toolCallId));
-						if (cacheAffinityKey && done.runContext?.threadCoord) setThreadCoord(cacheAffinityKey, done.runContext.threadCoord);
-						if (done.content) {
-							for (const item of done.content) if (item.type === "text" && !deltaTextEmitted) {
-								const tid = `${TEXT_STREAM_ID_PREFIX}_${Math.random().toString(36).slice(2, 10)}`;
-								controller.enqueue({
-									type: "text-start",
-									id: tid
-								});
-								controller.enqueue({
-									type: "text-delta",
-									id: tid,
-									delta: item.text
-								});
-								controller.enqueue({
-									type: "text-end",
-									id: tid
-								});
-							} else if (item.type === "tool-call") {
-								controller.enqueue({
-									type: "tool-input-start",
-									id: item.toolCallId,
-									toolName: item.toolName
-								});
-								controller.enqueue({
-									type: "tool-input-delta",
-									id: item.toolCallId,
-									delta: item.input
-								});
-								controller.enqueue({
-									type: "tool-input-end",
-									id: item.toolCallId
-								});
-								controller.enqueue({
-									type: "tool-call",
-									toolCallId: item.toolCallId,
-									toolName: item.toolName,
-									input: item.input
-								});
-							}
-						}
-						controller.enqueue({
-							type: "response-metadata",
-							modelId: done.model
-						});
-						controller.enqueue({
-							type: "finish",
-							finishReason: done.finishReason === "tool-calls" ? {
-								unified: "tool-calls",
-								raw: "tool_call"
-							} : {
-								unified: "stop",
-								raw: "stop"
-							},
-							usage: toAiSdkUsage(done.usage),
-							providerMetadata: qgridProviderMetadata(done)
-						});
-						streamCompleted = true;
-						controller.close();
-						return;
-					} else if (event.type === "error") {
-						streamCompleted = true;
-						controller.error(new Error(event.data.message));
-						return;
 					}
-					if (!streamCompleted) controller.error(/* @__PURE__ */ new Error("qgrid stream ended unexpectedly"));
+					if (!streamCompleted && !cancelled) {
+						signal.throwIfAborted();
+						controller.error(/* @__PURE__ */ new Error("qgrid stream ended unexpectedly"));
+					}
 				} catch (e) {
-					controller.error(e);
+					if (!cancelled) controller.error(e);
 				}
-			} }) };
+			}
+			return { stream };
 		}
 	};
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@cartanova/qgrid-ai-sdk",
-  "version": "2.9.10",
+  "version": "2.10.1",
   "description": "AI SDK LanguageModelV3 provider for qgrid",
   "repository": {
     "type": "git",