@cartanova/qgrid-ai-sdk 2.12.0 → 2.14.0

package/README.ko.md

@@ -2,7 +2,7 @@
 
 [English](./README.md) · **한국어**
 
-AI SDK v6 custom `LanguageModelV3` / `ImageModelV3` provider for [qgrid](https://github.com/cartanova-ai/Qgrid).
+[qgrid](https://github.com/cartanova-ai/Qgrid)용 AI SDK 7 custom provider입니다. AI SDK 7이 받는 `LanguageModelV3` / `ImageModelV3` 명세를 구현합니다.
 
 **기존 AI SDK 코드 변경 없이, `model` 한 줄만 바꾸면 구독 토큰 풀링(토큰 N개 × concurrent permit) + request log 대시보드를 사용할 수 있습니다.**
 
@@ -24,7 +24,7 @@ AI SDK v6 custom `LanguageModelV3` / `ImageModelV3` provider for [qgrid](https:/
  const { text } = await generateText({
    model: google("gemini-3-flash"),
    prompt: "복잡한 질문",
-+  experimental_telemetry: createQgridLogger({ serverUrl: "http://localhost:44900" }),
++  telemetry: createQgridLogger({ serverUrl: "http://localhost:44900" }),
  });
 ```
 
@@ -34,7 +34,7 @@ AI SDK v6 custom `LanguageModelV3` / `ImageModelV3` provider for [qgrid](https:/
 pnpm add @cartanova/qgrid-ai-sdk
 ```
 
-Peer dependencies: `ai@^6.0.0`, `@ai-sdk/provider@^3.0.0`
+Peer dependencies: `ai@^7.0.0`, `@ai-sdk/provider@^4.0.0`
 
 ## 빠른 시작
 
@@ -64,7 +64,7 @@ import { qgrid } from "@cartanova/qgrid-ai-sdk";
 
 const { text } = await generateText({
   model: qgrid("openai/gpt-6-luna"),
-  system: "당신은 학술 논문 요약가입니다.",
+  instructions: "당신은 학술 논문 요약가입니다.",
   prompt: paperText,
 });
 ```
@@ -80,7 +80,7 @@ import { z } from "zod";
 
 const { output } = await generateText({
   model: qgrid("openai/gpt-5.6-terra"),
-  system: "논문 메타데이터를 추출해주세요.",
+  instructions: "논문 메타데이터를 추출해주세요.",
   prompt: paperText,
   output: Output.object({
     schema: z.object({
@@ -118,7 +118,7 @@ for await (const chunk of textStream) {
 ### Tool Calling
 
 ```typescript
-import { generateText, stepCountIs, tool } from "ai";
+import { generateText, isStepCount, tool } from "ai";
 import { qgrid } from "@cartanova/qgrid-ai-sdk";
 import { z } from "zod";
 
@@ -134,7 +134,7 @@ const { text } = await generateText({
       },
     }),
   },
-  stopWhen: stepCountIs(3),
+  stopWhen: isStepCount(3),
 });
 ```
 
@@ -146,7 +146,7 @@ tool-call 턴 뒤에도 계속 진행해 모델의 최종 응답을 받습니다
 `tools`와 `Output.object`를 함께 사용할 수도 있습니다.
 
 ```typescript
-import { generateText, Output, stepCountIs, tool } from "ai";
+import { generateText, Output, isStepCount, tool } from "ai";
 import { qgrid } from "@cartanova/qgrid-ai-sdk";
 import { z } from "zod";
 
@@ -160,7 +160,7 @@ const { output } = await generateText({
       execute: async ({ city }) => ({ city, temperature: 22 }),
     }),
   },
-  stopWhen: stepCountIs(3),
+  stopWhen: isStepCount(3),
   output: Output.object({
     schema: z.object({
       city: z.string(),
@@ -345,10 +345,10 @@ qgrid는 `grok/*` 모델을 qgrid 대시보드에 등록한 xAI 구독 토큰(
 | Factory | 모델 ID | AI SDK 함수 |
 | --- | --- | --- |
 | `qgrid(id)` | `grok/grok-4.7`(카탈로그 기본 모델), `grok/grok-4.7-build-fast`, `grok/grok-4.6`, `grok/grok-4.5`, 그 밖의 `grok/<id>` | `generateText`, `streamText` |
-| `qgrid.image(id)` | `grok/grok-imagine-image`, `grok/grok-imagine-image-2.0`, `grok/grok-imagine-image-quality` | `generateImage` |
+| `qgrid.image(id)` | `grok/grok-imagine-image`, `grok/grok-imagine-image-2.0` | `generateImage` |
 | `qgrid.video(id)` | `grok/grok-imagine-video`, `grok/grok-imagine-video-1.5`, `grok/grok-imagine-video-1.5-lite` | `experimental_generateVideo` |
-| `qgrid.transcription(id)` | `grok/grok-voice-transcribe-2.0` | `experimental_transcribe` |
-| `qgrid.speech(id)` | `grok/grok-tts`(ID는 request log 라벨로만 쓰임) | `experimental_generateSpeech` |
+| `qgrid.transcription(id)` | `grok/grok-voice-transcribe-2.0` | `transcribe` |
+| `qgrid.speech(id)` | `grok/grok-tts`(ID는 request log 라벨로만 쓰임) | `generateSpeech` |
 
 qgrid는 카탈로그에 없는 language model ID를 거부하지 않고 그대로 전달합니다(OpenAI와 같음). 판단은
 subscription backend가 하며 다른 모델로 응답할 수 있으므로, 실제로 실행된 모델은 request log의 응답 모델(또는
@@ -403,16 +403,19 @@ Grok language model은 user 이미지 part(https URL 또는 inline bytes)를 버
 
 ### Grok 이미지
 
+생성과 편집에는 `grok-imagine-image-2.0`을 사용합니다. `quality`는 `low`, `medium`, `auto`이며,
+생략하면 `auto`입니다(현재 생성은 low, 편집은 medium).
+
 ```typescript
 import { generateImage } from "ai";
 import { qgrid } from "@cartanova/qgrid-ai-sdk";
 
 const { images, providerMetadata } = await generateImage({
-  model: qgrid.image("grok/grok-imagine-image"),
+  model: qgrid.image("grok/grok-imagine-image-2.0"),
   prompt: "A lighthouse at dusk, oil painting",
   n: 2,
   aspectRatio: "16:9",
-  providerOptions: { xai: { resolution: "2k", quality: "high", output_format: "jpeg" } },
+  providerOptions: { xai: { resolution: "2k", quality: "low", output_format: "jpeg" } },
 });
 images[0].mediaType; // "image/jpeg"
 providerMetadata.xai?.images[0]; // revisedPrompt와 해당 호출의 costInUsdTicks
@@ -459,8 +462,8 @@ try {
 ```typescript
 import { readFile } from "node:fs/promises";
 import {
-  experimental_generateSpeech as generateSpeech,
-  experimental_transcribe as transcribe,
+  generateSpeech,
+  transcribe,
 } from "ai";
 import { qgrid } from "@cartanova/qgrid-ai-sdk";
 
@@ -482,14 +485,14 @@ const speech = await generateSpeech({
 speech.audio.uint8Array;
 ```
 
-- `experimental_transcribe`는 bytes로 오디오 형식을 감지하고 실패하면 `audio/wav`를 씁니다. 서버는 `audio/*`가 아닌 값을 거부합니다. transcript에는 segment가 없습니다(`segments: []`).
+- `transcribe`는 bytes로 오디오 형식을 감지하고 실패하면 `audio/wav`를 씁니다. 서버는 `audio/*`가 아닌 값을 거부합니다. transcript에는 segment가 없습니다(`segments: []`).
 - 음성 합성 텍스트는 최대 15,000자입니다. `voice`, `language`, `outputFormat`을 지정하지 않으면 서버 기본값 `ara`, `en`, `mp3`가 쓰이며, live 검증된 조합은 이것뿐입니다. 다른 voice, language(`@ai-sdk/xai`의 `auto` 포함), codec은 그대로 전달되지만 미검증입니다. 서버가 돌려준 오디오 MIME 타입은 `providerMetadata.qgrid.mediaType`에 있고, AI SDK의 `speech.audio.mediaType`은 bytes에서 추정한 값입니다.
 
 ## 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.
+서버와 AI SDK 2.10.0부터 qgrid provider의 `generateText` 결과에서는 `result.finalStep.providerMetadata?.qgrid?.requestLogId`로 요청 로그 ID를 확인할 수 있습니다. `streamText`는 스트림 소비 후 `(await result.finalStep).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.
+qgrid provider가 아닌 모델(google, openai 직접 호출)에서도 같은 request log 대시보드를 사용하려면 `createQgridLogger`를 `telemetry`에 넣으면 됩니다.
 
 ```typescript
 import { generateText } from "ai";
@@ -499,7 +502,7 @@ import { createQgridLogger } from "@cartanova/qgrid-ai-sdk";
 const { text } = await generateText({
   model: google("gemini-3-flash"),
   prompt: "안녕하세요",
-  experimental_telemetry: createQgridLogger({
+  telemetry: createQgridLogger({
     serverUrl: "http://localhost:44900",
   }),
 });
@@ -537,7 +540,7 @@ const { text } = await generateText({
   providerOptions: {
     qgrid: { logger: false } satisfies QgridProviderOptions,
   },
-  experimental_telemetry: createQgridLogger({ serverUrl: "http://localhost:44900" }),
+  telemetry: createQgridLogger({ serverUrl: "http://localhost:44900" }),
 });
 ```
 
@@ -698,6 +701,12 @@ QGRID_REAL_PROVIDER_ACCEPTANCE=1 QGRID_URL=http://localhost:44900 QGRID_MODEL=op
 
 ## 요구사항
 
-- Node.js >= 20
-- AI SDK v6 (`ai@^6.0.0`)
+- Node.js >= 22
+- AI SDK 7 (`ai@^7.0.0`)
 - 실행 중인 qgrid 서버
+
+### AI SDK 7로 업그레이드
+
+- logger는 `experimental_telemetry` 대신 `telemetry: createQgridLogger(...)`로 넘깁니다.
+- `metadata.qgridRunId`는 더 이상 필요 없습니다. logger가 AI SDK의 호출별 `callId`로 run을 구분하므로 logger 하나를 동시 호출에 함께 써도 됩니다.
+- 요청 로그 ID는 `result.finalStep.providerMetadata`에서 읽습니다. AI SDK 7에서 최상위 `providerMetadata`는 deprecated alias로만 남아 있습니다.

package/README.md

@@ -2,7 +2,7 @@
 
 **English** · [한국어](./README.ko.md)
 
-AI SDK v6 custom `LanguageModelV3` / `ImageModelV3` provider for [qgrid](https://github.com/cartanova-ai/Qgrid).
+AI SDK 7 custom provider for [qgrid](https://github.com/cartanova-ai/Qgrid), implementing the `LanguageModelV3` / `ImageModelV3` specification that AI SDK 7 accepts.
 
 **Without changing your existing AI SDK code, swap a single `model` line to get subscription-token pooling (N tokens × concurrent permits) + the request log dashboard.**
 
@@ -24,7 +24,7 @@ If you already use another provider (google, openai, ...) directly, add **a sing
  const { text } = await generateText({
    model: google("gemini-3-flash"),
    prompt: "A complex question",
-+  experimental_telemetry: createQgridLogger({ serverUrl: "http://localhost:44900" }),
++  telemetry: createQgridLogger({ serverUrl: "http://localhost:44900" }),
  });
 ```
 
@@ -34,7 +34,7 @@ If you already use another provider (google, openai, ...) directly, add **a sing
 pnpm add @cartanova/qgrid-ai-sdk
 ```
 
-Peer dependencies: `ai@^6.0.0`, `@ai-sdk/provider@^3.0.0`
+Peer dependencies: `ai@^7.0.0`, `@ai-sdk/provider@^4.0.0`
 
 ## Quick start
 
@@ -64,7 +64,7 @@ import { qgrid } from "@cartanova/qgrid-ai-sdk";
 
 const { text } = await generateText({
   model: qgrid("openai/gpt-6-luna"),
-  system: "You are an academic paper summarizer.",
+  instructions: "You are an academic paper summarizer.",
   prompt: paperText,
 });
 ```
@@ -80,7 +80,7 @@ import { z } from "zod";
 
 const { output } = await generateText({
   model: qgrid("openai/gpt-5.6-terra"),
-  system: "Extract the paper metadata.",
+  instructions: "Extract the paper metadata.",
   prompt: paperText,
   output: Output.object({
     schema: z.object({
@@ -126,7 +126,7 @@ for await (const chunk of textStream) {
 ### Tool Calling
 
 ```typescript
-import { generateText, stepCountIs, tool } from "ai";
+import { generateText, isStepCount, tool } from "ai";
 import { qgrid } from "@cartanova/qgrid-ai-sdk";
 import { z } from "zod";
 
@@ -142,7 +142,7 @@ const { text } = await generateText({
       },
     }),
   },
-  stopWhen: stepCountIs(3),
+  stopWhen: isStepCount(3),
 });
 ```
 
@@ -154,7 +154,7 @@ tool-call step and asks the model for the final response.
 Tools can be combined with `Output.object`:
 
 ```typescript
-import { generateText, Output, stepCountIs, tool } from "ai";
+import { generateText, Output, isStepCount, tool } from "ai";
 import { qgrid } from "@cartanova/qgrid-ai-sdk";
 import { z } from "zod";
 
@@ -168,7 +168,7 @@ const { output } = await generateText({
       execute: async ({ city }) => ({ city, temperature: 22 }),
     }),
   },
-  stopWhen: stepCountIs(3),
+  stopWhen: isStepCount(3),
   output: Output.object({
     schema: z.object({
       city: z.string(),
@@ -355,10 +355,10 @@ Grok model accepts the same factory config as the other models: `serverUrl` (def
 | Factory | Model ids | AI SDK function |
 | --- | --- | --- |
 | `qgrid(id)` | `grok/grok-4.7` (default catalog model), `grok/grok-4.7-build-fast`, `grok/grok-4.6`, `grok/grok-4.5`, any other `grok/<id>` | `generateText`, `streamText` |
-| `qgrid.image(id)` | `grok/grok-imagine-image`, `grok/grok-imagine-image-2.0`, `grok/grok-imagine-image-quality` | `generateImage` |
+| `qgrid.image(id)` | `grok/grok-imagine-image`, `grok/grok-imagine-image-2.0` | `generateImage` |
 | `qgrid.video(id)` | `grok/grok-imagine-video`, `grok/grok-imagine-video-1.5`, `grok/grok-imagine-video-1.5-lite` | `experimental_generateVideo` |
-| `qgrid.transcription(id)` | `grok/grok-voice-transcribe-2.0` | `experimental_transcribe` |
-| `qgrid.speech(id)` | `grok/grok-tts` (the id only labels the request log) | `experimental_generateSpeech` |
+| `qgrid.transcription(id)` | `grok/grok-voice-transcribe-2.0` | `transcribe` |
+| `qgrid.speech(id)` | `grok/grok-tts` (the id only labels the request log) | `generateSpeech` |
 
 qgrid forwards a language-model id that is not in its catalog unchanged instead of rejecting it,
 as it does for OpenAI; the subscription backend decides. It may serve a different model, so check
@@ -414,16 +414,19 @@ error. Compress the input, or pass an https image URL where the option accepts o
 
 ### Grok images
 
+Use `grok-imagine-image-2.0` for generation and editing. Its `quality` values are `low`, `medium`,
+and `auto`; omission uses `auto` (currently low for generation, medium for editing).
+
 ```typescript
 import { generateImage } from "ai";
 import { qgrid } from "@cartanova/qgrid-ai-sdk";
 
 const { images, providerMetadata } = await generateImage({
-  model: qgrid.image("grok/grok-imagine-image"),
+  model: qgrid.image("grok/grok-imagine-image-2.0"),
   prompt: "A lighthouse at dusk, oil painting",
   n: 2,
   aspectRatio: "16:9",
-  providerOptions: { xai: { resolution: "2k", quality: "high", output_format: "jpeg" } },
+  providerOptions: { xai: { resolution: "2k", quality: "low", output_format: "jpeg" } },
 });
 images[0].mediaType; // "image/jpeg"
 providerMetadata.xai?.images[0]; // revisedPrompt and the call's costInUsdTicks
@@ -471,8 +474,8 @@ try {
 ```typescript
 import { readFile } from "node:fs/promises";
 import {
-  experimental_generateSpeech as generateSpeech,
-  experimental_transcribe as transcribe,
+  generateSpeech,
+  transcribe,
 } from "ai";
 import { qgrid } from "@cartanova/qgrid-ai-sdk";
 
@@ -494,14 +497,14 @@ const speech = await generateSpeech({
 speech.audio.uint8Array;
 ```
 
-- `experimental_transcribe` detects the audio type from its bytes and falls back to `audio/wav`; the server rejects anything that is not `audio/*`. Transcripts carry no segments (`segments: []`).
+- `transcribe` detects the audio type from its bytes and falls back to `audio/wav`; the server rejects anything that is not `audio/*`. Transcripts carry no segments (`segments: []`).
 - Speech text is limited to 15,000 characters. Unset `voice`, `language` and `outputFormat` use the server defaults `ara`, `en` and `mp3`, the only combination verified live. Other voices, languages (including `@ai-sdk/xai`'s `auto`) and codecs are passed through but unverified. The server's audio MIME type is in `providerMetadata.qgrid.mediaType`; AI SDK's `speech.audio.mediaType` is sniffed from the bytes.
 
 ## 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.
+With qgrid server and SDK 2.10.0+, read the parent request-log ID from `result.finalStep.providerMetadata?.qgrid?.requestLogId` after `generateText`. For `streamText`, consume the stream first, then read `(await result.finalStep).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`.
+To use the same request log dashboard with models that don't go through the qgrid provider (direct google/openai calls), pass `createQgridLogger` to `telemetry`.
 
 ```typescript
 import { generateText } from "ai";
@@ -511,7 +514,7 @@ import { createQgridLogger } from "@cartanova/qgrid-ai-sdk";
 const { text } = await generateText({
   model: google("gemini-3-flash"),
   prompt: "Hello",
-  experimental_telemetry: createQgridLogger({
+  telemetry: createQgridLogger({
     serverUrl: "http://localhost:44900",
   }),
 });
@@ -550,7 +553,7 @@ const { text } = await generateText({
   providerOptions: {
     qgrid: { logger: false } satisfies QgridProviderOptions,
   },
-  experimental_telemetry: createQgridLogger({ serverUrl: "http://localhost:44900" }),
+  telemetry: createQgridLogger({ serverUrl: "http://localhost:44900" }),
 });
 ```
 
@@ -714,6 +717,12 @@ Keep new tests focused on observable contracts or concrete regressions. Preserve
 
 ## Requirements
 
-- Node.js >= 20
-- AI SDK v6 (`ai@^6.0.0`)
+- Node.js >= 22
+- AI SDK 7 (`ai@^7.0.0`)
 - A running qgrid server
+
+### Upgrading to AI SDK 7
+
+- Pass the logger as `telemetry: createQgridLogger(...)` instead of `experimental_telemetry`.
+- Drop `metadata.qgridRunId`. The logger keys each run by AI SDK's per-call `callId`, so one logger can serve concurrent calls.
+- Read the request-log ID from `result.finalStep.providerMetadata`. AI SDK 7 keeps the top-level `providerMetadata` only as a deprecated alias.

package/dist/index.d.ts

@@ -1,6 +1,6 @@
 import { z } from "zod";
 import { Experimental_VideoModelV3, Experimental_VideoModelV3VideoData, ImageModelV3, LanguageModelV3, SharedV3ProviderMetadata, SpeechModelV3, TranscriptionModelV3 } from "@ai-sdk/provider";
-import { TelemetrySettings } from "ai";
+import { TelemetryOptions } from "ai";
 //#region src/grok-options.d.ts
 /** Effort values qgrid forwards to the Grok subscription backend (`reasoning.effort`). */
 type QgridGrokEffort = "low" | "medium" | "high" | "xhigh";
@@ -9,7 +9,7 @@ type XaiOptionsIssue = {
   message: string;
 };
 /** Thrown when `providerOptions.xai` has an invalid shape or value type. */
-declare class QgridXaiOptionsError extends Error {
+export declare class QgridXaiOptionsError extends Error {
   readonly issues: XaiOptionsIssue[];
   constructor(issues: XaiOptionsIssue[]);
 }
@@ -36,7 +36,7 @@ type QgridCommonProviderConfig = {
   projectName?: string;
 };
 /** Codex Images request ID, not a guaranteed serving version; does not select a chat model. */
-type QgridImageModel = "openai/gpt-image-2" | "grok/grok-imagine-image" | "grok/grok-imagine-image-2.0" | "grok/grok-imagine-image-quality";
+type QgridImageModel = "openai/gpt-image-2" | "grok/grok-imagine-image" | "grok/grok-imagine-image-2.0";
 /** Grok video generation model ids (xAI subscription; not yet wired to a provider factory). */
 type QgridVideoModel = "grok/grok-imagine-video" | "grok/grok-imagine-video-1.5" | "grok/grok-imagine-video-1.5-lite";
 /** Grok speech-to-text model ids. */
@@ -138,10 +138,10 @@ type QgridGrokModel = "grok/grok-4.7" | "grok/grok-4.7-build-fast" | "grok/grok-
  */
 declare const qgridXaiProviderOptionsSchema: z.ZodObject<{
   reasoningEffort: z.ZodOptional<z.ZodEnum<{
-    none: "none";
     low: "low";
     medium: "medium";
     high: "high";
+    none: "none";
     xhigh: "xhigh";
   }>>;
   maxTurns: z.ZodOptional<z.ZodNumber>;
@@ -200,29 +200,27 @@ type QgridLoggerConfig = {
    */
   tokenName?: string;
   /**
-   * Fallback timeout for runs that receive onStart but never receive onFinish.
-   * AI SDK TelemetryIntegration does not expose an error hook, so provider
-   * failures before a final step can otherwise leave request logs running.
+   * Safety-net timeout for runs that start but never report an end, error, or
+   * abort, for example a streamText result whose stream is never consumed.
    *
    * Set to 0 to disable. Defaults to 30 minutes, or the AI SDK total timeout
    * plus a short grace period when one is provided.
    */
   staleRunTimeoutMs?: number;
   /**
-   * Receives qgrid logging failures. When reusing one logger integration across
-   * overlapping AI SDK calls, pass a unique `metadata.qgridRunId` per call so
-   * lifecycle events can be attributed to the correct run.
+   * Receives qgrid logging failures. One logger can serve concurrent AI SDK
+   * calls; each call is logged as its own run.
    */
   onLogError?: (error: Error) => void;
 };
 //#endregion
 //#region src/image-model.d.ts
 /** Server-side Fastify body limit (`sonamu.config.ts` `bodyLimit`). Larger bodies get a 413. */
-declare const QGRID_MAX_REQUEST_BODY_BYTES: number;
+export declare const QGRID_MAX_REQUEST_BODY_BYTES: number;
 /** `config` for `qgrid.image/video/transcription/speech`; same fields as the language model. */
 type QgridMediaProviderConfig = QgridImageProviderConfig;
 /** Non-2xx response from a qgrid media endpoint. */
-declare class QgridMediaHttpError extends Error {
+export declare class QgridMediaHttpError extends Error {
   readonly status: number;
   constructor(message: string, status: number);
 }
@@ -285,12 +283,12 @@ type QgridXaiVideoProviderOptions = {
  * Raised when the poll budget runs out. The job keeps running on xAI; pass `jobHandle` to
  * `resumeQgridVideo` (with the same `projectName`) to keep waiting.
  */
-declare class QgridVideoTimeoutError extends Error {
+export declare class QgridVideoTimeoutError extends Error {
   readonly jobHandle: string;
   constructor(jobHandle: string, timeoutMs: number);
 }
 /** The upstream job finished as failed (including content-policy blocks). */
-declare class QgridVideoGenerationError extends Error {
+export declare class QgridVideoGenerationError extends Error {
   readonly jobHandle: string;
   constructor(jobHandle: string, detail: string | undefined);
 }
@@ -315,28 +313,28 @@ type QgridVideoResult = {
  * AI SDK v3 video models expose only `doGenerate` (no `doStart` / `doStatus`), so resuming is
  * a qgrid helper rather than an `experimental_generateVideo` option.
  */
-declare function resumeQgridVideo(jobHandle: string, options?: QgridVideoPollOptions & QgridMediaProviderConfig & {
+export declare function resumeQgridVideo(jobHandle: string, options?: QgridVideoPollOptions & QgridMediaProviderConfig & {
   modelId?: QgridVideoModel;
 }): Promise<QgridVideoResult>;
 declare function createQgridVideo(modelId: QgridVideoModel, config?: QgridMediaProviderConfig): Experimental_VideoModelV3;
 //#endregion
 //#region src/logger.d.ts
-declare function createQgridLogger(config?: QgridLoggerConfig): TelemetrySettings;
+export declare function createQgridLogger(config?: QgridLoggerConfig): TelemetryOptions;
 //#endregion
 //#region src/index.d.ts
 /** `qgrid()` config for Grok models; `defaultEffort` uses the Grok catalog effort vocabulary. */
-type QgridGrokProviderConfig = Omit<QgridOpenAIProviderConfig, "defaultEffort"> & {
+export type QgridGrokProviderConfig = Omit<QgridOpenAIProviderConfig, "defaultEffort"> & {
   /** Omitted: the server applies the model's catalog default effort. */
   defaultEffort?: QgridGrokEffort;
 };
-declare function qgrid(modelId: QgridOpenAIModel, config?: QgridOpenAIProviderConfig): LanguageModelV3;
-declare function qgrid(modelId: QgridAnthropicModel, config?: QgridAnthropicProviderConfig): LanguageModelV3;
-declare function qgrid(modelId: QgridGrokModel, config?: QgridGrokProviderConfig): LanguageModelV3;
-declare namespace qgrid {
+export declare function qgrid(modelId: QgridOpenAIModel, config?: QgridOpenAIProviderConfig): LanguageModelV3;
+export declare function qgrid(modelId: QgridAnthropicModel, config?: QgridAnthropicProviderConfig): LanguageModelV3;
+export declare function qgrid(modelId: QgridGrokModel, config?: QgridGrokProviderConfig): LanguageModelV3;
+export declare namespace qgrid {
   var image: typeof createQgridImage;
   var video: typeof createQgridVideo;
   var transcription: typeof createQgridTranscription;
   var speech: typeof createQgridSpeech;
 }
 //#endregion
-export { QGRID_MAX_REQUEST_BODY_BYTES, type QgridAnthropicEffort, type QgridAnthropicModel, type QgridAnthropicProviderConfig, type QgridAnthropicProviderOptions, type QgridGrokEffort, type QgridGrokMediaProviderOptions, type QgridGrokModel, QgridGrokProviderConfig, type QgridImageGenerationMetadata, type QgridImageModel, type QgridImageProviderConfig, type QgridImageProviderOptions, type QgridLoggerConfig, QgridMediaHttpError, type QgridMediaProviderConfig, type QgridOpenAIEffort, type QgridOpenAIModel, type QgridOpenAIProviderConfig, type QgridOpenAIProviderOptions, type QgridProviderConfig, type QgridProviderOptions, type QgridSpeechModel, type QgridSupportedModel, type QgridTranscriptionModel, QgridVideoGenerationError, type QgridVideoModel, type QgridVideoPollOptions, type QgridVideoResult, QgridVideoTimeoutError, type QgridXaiImageProviderOptions, QgridXaiOptionsError, type QgridXaiProviderOptions, type QgridXaiTranscriptionProviderOptions, type QgridXaiVideoProviderOptions, createQgridLogger, qgrid as default, qgrid, resumeQgridVideo };
+export { type QgridAnthropicEffort, type QgridAnthropicModel, type QgridAnthropicProviderConfig, type QgridAnthropicProviderOptions, type QgridGrokEffort, type QgridGrokMediaProviderOptions, type QgridGrokModel, type QgridImageGenerationMetadata, type QgridImageModel, type QgridImageProviderConfig, type QgridImageProviderOptions, type QgridLoggerConfig, type QgridMediaProviderConfig, type QgridOpenAIEffort, type QgridOpenAIModel, type QgridOpenAIProviderConfig, type QgridOpenAIProviderOptions, type QgridProviderConfig, type QgridProviderOptions, type QgridSpeechModel, type QgridSupportedModel, type QgridTranscriptionModel, type QgridVideoModel, type QgridVideoPollOptions, type QgridVideoResult, type QgridXaiImageProviderOptions, type QgridXaiProviderOptions, type QgridXaiTranscriptionProviderOptions, type QgridXaiVideoProviderOptions, qgrid as default };

package/dist/index.js

@@ -1,5 +1,5 @@
 import { z } from "zod";
-import { Agent } from "undici";
+import { Agent, fetch as fetch$1 } from "undici";
 //#region src/envelope-stream-parser.ts
 const ESCAPE_MAP = /* @__PURE__ */ new Map([
 	["\"", "\""],
@@ -520,19 +520,21 @@ function mapXaiProviderOptions(options, modelId) {
 		}
 	}
 	const effort = values.reasoningEffort;
-	if (effort !== void 0) if (effort === "none") {
-		result.warnings.push({
-			key: "reasoningEffort",
-			reason: "reasoningEffort \"none\" is not supported; the qgrid Grok route always reasons"
-		});
-		result.rejected.push("reasoningEffort");
-	} else if (isGrokEffort(effort) && grokEffortsForModel(modelId).includes(effort)) result.request.effort = effort;
-	else {
-		result.warnings.push({
-			key: "reasoningEffort",
-			reason: `reasoningEffort "${effort}" is not supported by ${modelId}`
-		});
-		result.rejected.push("reasoningEffort");
+	if (effort !== void 0) {
+		if (effort === "none") {
+			result.warnings.push({
+				key: "reasoningEffort",
+				reason: "reasoningEffort \"none\" is not supported; the qgrid Grok route always reasons"
+			});
+			result.rejected.push("reasoningEffort");
+		} else if (isGrokEffort(effort) && grokEffortsForModel(modelId).includes(effort)) result.request.effort = effort;
+		else {
+			result.warnings.push({
+				key: "reasoningEffort",
+				reason: `reasoningEffort "${effort}" is not supported by ${modelId}`
+			});
+			result.rejected.push("reasoningEffort");
+		}
 	}
 	if (values.promptCacheKey !== void 0) result.request.promptCacheKey = values.promptCacheKey;
 	return result;
@@ -567,9 +569,19 @@ async function closeQgridRequestTransport(transport) {
 	if (!transport) return;
 	await transport.dispatcher.close().catch(() => void 0);
 }
-async function qgridFetch(input, init, context) {
+async function qgridFetch(target, init, context) {
 	try {
-		return await fetch(input, init);
+		const url = new URL(target);
+		if (url.protocol !== "http:" && url.protocol !== "https:") throw new TypeError(`qgrid server URL must use http or https, not ${url.protocol}`);
+		const href = url.href;
+		if (init.dispatcher) return await fetch$1(href, init);
+		const { method, body, signal, headers } = init;
+		return await fetch(href, {
+			method,
+			body,
+			signal,
+			headers
+		});
 	} catch (error) {
 		throw toQgridTransportError(error, context);
 	}
@@ -630,7 +642,7 @@ function serverOrigin(serverUrl) {
 //#endregion
 //#region src/qgrid.constant.ts
 const QGRID_PROVIDER_NAME = "qgrid";
-const THREAD_COORD_TTL_MS = 10 * 6e4;
+const THREAD_COORD_TTL_MS = 6e5;
 const TEXT_STREAM_ID_PREFIX = "text";
 const TOP_LEVEL_SCHEMA_WARNING = "qgrid server-enforced structured output requires \"object\". Falling back to client-side parsing.";
 const PENDING_TOOL_RESULTS_WARNING = "[qgrid] tool results did not match any pending qgrid run; sending request without runContext";
@@ -709,6 +721,13 @@ function extractToolResultsFromHistory(messages) {
 	});
 	return out;
 }
+function parseSseData(raw) {
+	try {
+		return JSON.parse(raw);
+	} catch {
+		return;
+	}
+}
 async function* parseSSE(body, signal) {
 	const reader = body.getReader();
 	const decoder = new TextDecoder();
@@ -728,13 +747,11 @@ async function* parseSSE(body, signal) {
 			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 {}
+				const data = parseSseData(line.slice(6));
+				if (data !== void 0) yield {
+					type: eventType || "message",
+					data
+				};
 				eventType = "";
 			}
 		}
@@ -925,11 +942,9 @@ function getErrorMessage(cause) {
 		return "unknown error";
 	}
 }
-function extractUserPrompt(prompt, messages) {
-	if (prompt !== void 0 && !Array.isArray(prompt)) return prompt;
-	const messageList = messages ?? prompt ?? [];
-	for (let i = messageList.length - 1; i >= 0; i--) {
-		const msg = messageList[i];
+function extractUserPrompt(messages) {
+	for (let i = messages.length - 1; i >= 0; i--) {
+		const msg = messages[i];
 		if (msg?.role === "user") return extractTextContent(msg.content);
 	}
 	return "";
@@ -974,7 +989,7 @@ function imageInputUrl(file) {
 	return `data:${file.mediaType};base64,${data}`;
 }
 /** Server-side Fastify body limit (`sonamu.config.ts` `bodyLimit`). Larger bodies get a 413. */
-const QGRID_MAX_REQUEST_BODY_BYTES = 10 * 1024 * 1024;
+const QGRID_MAX_REQUEST_BODY_BYTES = 10485760;
 function resolveMediaConfig(config) {
 	return {
 		serverUrl: config?.serverUrl ?? process.env.QGRID_URL ?? "http://localhost:44900",
@@ -993,7 +1008,7 @@ function formatMiB(bytes) {
 * inline images of one request together. Derived from the 10 MiB body limit: base64 grows bytes by
 * 4/3, so 7 MiB decoded is about 9.3 MiB encoded, leaving room for the rest of the JSON body.
 */
-const QGRID_MAX_INLINE_MEDIA_BYTES = 7 * 1024 * 1024;
+const QGRID_MAX_INLINE_MEDIA_BYTES = 7340032;
 function base64DecodedBytes(data) {
 	const padding = data.endsWith("==") ? 2 : data.endsWith("=") ? 1 : 0;
 	return Math.floor(data.length * 3 / 4) - padding;
@@ -1493,14 +1508,16 @@ function createQgridSpeech(modelId, config) {
 			const xai = parseXaiMediaOptions(options.providerOptions?.xai, xaiSpeechOptionsSchema, /* @__PURE__ */ new Set(), LABEL$2);
 			const warnings = [];
 			let format;
-			if (options.outputFormat !== void 0) if (isSpeechFormat(options.outputFormat)) format = options.outputFormat;
-			else {
-				format = "mp3";
-				warnings.push({
-					type: "unsupported",
-					feature: "outputFormat",
-					details: `Unsupported output format: ${options.outputFormat}. Using mp3 instead.`
-				});
+			if (options.outputFormat !== void 0) {
+				if (isSpeechFormat(options.outputFormat)) format = options.outputFormat;
+				else {
+					format = "mp3";
+					warnings.push({
+						type: "unsupported",
+						feature: "outputFormat",
+						details: `Unsupported output format: ${options.outputFormat}. Using mp3 instead.`
+					});
+				}
 			}
 			if (options.instructions !== void 0) warnings.push({
 				type: "unsupported",
@@ -1896,6 +1913,18 @@ function createQgridVideo(modelId, config) {
 				type: "unsupported",
 				feature: "seed"
 			});
+			if (options.frameImages !== void 0) warnings.push({
+				type: "unsupported",
+				feature: "frameImages"
+			});
+			if (options.inputReferences !== void 0) warnings.push({
+				type: "unsupported",
+				feature: "inputReferences"
+			});
+			if (options.generateAudio !== void 0) warnings.push({
+				type: "unsupported",
+				feature: "generateAudio"
+			});
 			let resolution = xaiOptions?.resolution ?? void 0;
 			if (resolution === void 0 && options.resolution !== void 0) {
 				resolution = RESOLUTION_MAP.get(options.resolution);
@@ -1906,12 +1935,14 @@ function createQgridVideo(modelId, config) {
 				});
 			}
 			let lastFrame;
-			if (xaiOptions?.lastFrame !== void 0) if (modelId === LAST_FRAME_MODEL) lastFrame = toImageRefUrl(xaiOptions.lastFrame, LABEL);
-			else warnings.push({
-				type: "unsupported",
-				feature: "providerOptions.xai.lastFrame",
-				details: `xAI only supports last_frame with "${LAST_FRAME_MODEL}". The last frame was ignored.`
-			});
+			if (xaiOptions?.lastFrame !== void 0) {
+				if (modelId === LAST_FRAME_MODEL) lastFrame = toImageRefUrl(xaiOptions.lastFrame, LABEL);
+				else warnings.push({
+					type: "unsupported",
+					feature: "providerOptions.xai.lastFrame",
+					details: `xAI only supports last_frame with "${LAST_FRAME_MODEL}". The last frame was ignored.`
+				});
+			}
 			const image = options.image === void 0 ? void 0 : toImageRef(options.image, LABEL);
 			assertInlineMediaFits([
 				image,
@@ -1957,11 +1988,21 @@ function createQgridVideo(modelId, config) {
 }
 //#endregion
 //#region src/logger.ts
-const DEFAULT_RUN_KEY = "__qgrid_default_run__";
-const DEFAULT_STALE_RUN_TIMEOUT_MS = 1800 * 1e3;
+const DEFAULT_STALE_RUN_TIMEOUT_MS = 18e5;
 const STALE_RUN_GRACE_MS = 5e3;
-const OVERLAPPING_RUN_ERROR = "createQgridLogger received overlapping runs for the same telemetry key. Pass a unique metadata.qgridRunId per AI SDK call or create a fresh logger integration per call.";
-const qgridRunIdSchema = z.string().min(1);
+const STALE_RUN_ERROR = "AI SDK generation reported no end, error, or abort before the stale-run timeout";
+const TEXT_OPERATIONS = /* @__PURE__ */ new Set(["ai.generateText", "ai.streamText"]);
+const errorEventSchema = z.object({
+	callId: z.string(),
+	error: z.unknown()
+});
+const abortErrorSchema = z.object({ name: z.enum(["AbortError", "TimeoutError"]) });
+function isTextStart(event) {
+	return TEXT_OPERATIONS.has(event.operationId);
+}
+function isTextEnd(event) {
+	return "finalStep" in event;
+}
 function isParseableJson(value) {
 	try {
 		JSON.parse(value ?? "");
@@ -1975,46 +2016,25 @@ function fullModelName(provider, modelId) {
 	const prefix = `${provider.split(".", 1)[0] ?? provider}/`;
 	return modelId.startsWith(prefix) ? modelId : `${prefix}${modelId}`;
 }
-function timedKeySet() {
-	const keys = /* @__PURE__ */ new Set();
-	const timers = /* @__PURE__ */ new Map();
-	return {
-		has: (k) => keys.has(k),
-		add(k, ttlMs) {
-			keys.add(k);
-			const existing = timers.get(k);
-			if (existing) clearTimeout(existing);
-			const t = setTimeout(() => {
-				keys.delete(k);
-				timers.delete(k);
-			}, ttlMs);
-			t.unref?.();
-			timers.set(k, t);
-		},
-		remove(k) {
-			keys.delete(k);
-			const t = timers.get(k);
-			if (t) clearTimeout(t);
-			timers.delete(k);
-		}
-	};
+function staleRunTimeout(config, timeout) {
+	if (config.staleRunTimeoutMs !== void 0) return config.staleRunTimeoutMs;
+	if (timeout instanceof Object) {
+		const totalMs = timeout.totalMs;
+		return totalMs !== void 0 && totalMs > 0 ? totalMs + STALE_RUN_GRACE_MS : DEFAULT_STALE_RUN_TIMEOUT_MS;
+	}
+	return timeout !== void 0 && timeout > 0 ? timeout + STALE_RUN_GRACE_MS : DEFAULT_STALE_RUN_TIMEOUT_MS;
 }
 function createQgridLogger(config = {}) {
 	const serverUrl = config.serverUrl ?? process.env.QGRID_URL ?? "http://localhost:44900";
 	const projectName = config.projectName ?? process.env.QGRID_PROJECT_NAME;
 	const onLogError = config.onLogError ?? ((e) => console.warn(`[qgrid-logger] ${e.message}`));
 	const runs = /* @__PURE__ */ new Map();
-	const starting = /* @__PURE__ */ new Map();
-	const keyTtl = config.staleRunTimeoutMs !== void 0 && config.staleRunTimeoutMs > 0 ? config.staleRunTimeoutMs : DEFAULT_STALE_RUN_TIMEOUT_MS;
-	const suppressedRuns = timedKeySet();
-	const quarantined = timedKeySet();
-	const finalizeRun = async (runKey, result) => {
-		const run = runs.get(runKey);
+	const finalizeRun = async (callId, result) => {
+		const run = runs.get(callId);
 		if (!run || run.finishing) return;
 		run.finishing = true;
-		runs.delete(runKey);
+		runs.delete(callId);
 		if (run.watchdog) clearTimeout(run.watchdog);
-		run.cleanupAbortListener?.();
 		for (const pending of run.pendingToolCalls) run.pendingSteps.push(appendStep(serverUrl, {
 			requestLogId: run.requestLogId,
 			stepIndex: pending.stepIndex,
@@ -2046,147 +2066,63 @@ function createQgridLogger(config = {}) {
 		if (result.errorMessage) finishInput.errorMessage = result.errorMessage;
 		await finishRun(serverUrl, finishInput).catch((e) => onLogError(e instanceof Error ? e : new Error(String(e))));
 	};
-	let autoRunIdCounter = 0;
-	const resolveRunKey = (qgridRunId, functionId) => {
-		if (qgridRunId !== void 0) return `qgridRunId:${qgridRunId}`;
-		if (functionId) return `functionId:${functionId}`;
-		return DEFAULT_RUN_KEY;
-	};
 	return {
 		isEnabled: true,
 		integrations: [{
 			async onStart(event) {
-				if (!event.metadata?.qgridRunId && !event.functionId && event.metadata) event.metadata.qgridRunId = `auto-${++autoRunIdCounter}`;
-				const runKey = resolveRunKey(qgridRunIdSchema.safeParse(event.metadata?.qgridRunId).data, event.functionId);
-				if (event.model.provider === "qgrid") return;
-				if (quarantined.has(runKey)) {
-					onLogError(/* @__PURE__ */ new Error("createQgridLogger: telemetry key is quarantined after overlap"));
-					return;
-				}
-				if (runs.has(runKey) || starting.has(runKey) || suppressedRuns.has(runKey)) {
-					const reservation = starting.get(runKey);
-					if (reservation) {
-						reservation.errorMessage = OVERLAPPING_RUN_ERROR;
-						starting.delete(runKey);
-					}
-					suppressedRuns.remove(runKey);
-					quarantined.add(runKey, keyTtl);
-					const finalizing = finalizeRun(runKey, {
-						status: "error",
-						errorMessage: OVERLAPPING_RUN_ERROR
-					});
-					onLogError(/* @__PURE__ */ new Error(OVERLAPPING_RUN_ERROR));
-					await finalizing;
-					return;
-				}
-				if (event.providerOptions?.qgrid?.logger === false) {
-					suppressedRuns.add(runKey, keyTtl);
-					return;
-				}
-				const requestedModelName = fullModelName(event.model.provider, event.model.modelId) ?? event.model.modelId;
-				const reservation = {
-					requestedModelName,
-					startTime: Date.now()
-				};
-				starting.set(runKey, reservation);
+				if (!isTextStart(event)) return;
+				if (event.provider === "qgrid") return;
+				if (event.providerOptions?.qgrid?.logger === false) return;
+				const requestedModelName = fullModelName(event.provider, event.modelId) ?? event.modelId;
+				const startTime = Date.now();
 				try {
-					const messages = event.messages ?? (Array.isArray(event.prompt) ? event.prompt : void 0);
-					const history = serializeHistory(messages);
 					const responseFormat = event.output ? await event.output.responseFormat : void 0;
 					const isStructured = responseFormat?.type === "json";
 					const jsonSchema = isStructured && responseFormat.schema !== void 0 ? JSON.stringify(responseFormat.schema) ?? null : null;
 					const result = await createRun(serverUrl, {
-						userPrompt: extractUserPrompt(event.prompt, messages),
-						systemPrompt: extractSystemPrompt(event.system),
+						userPrompt: extractUserPrompt(event.messages),
+						systemPrompt: extractSystemPrompt(event.instructions),
 						modelName: requestedModelName,
 						projectName,
-						history,
+						history: serializeHistory(event.messages),
 						isStructured,
 						jsonSchema
 					});
-					if (reservation.errorMessage || quarantined.has(runKey) || starting.get(runKey) !== reservation) {
-						if (starting.get(runKey) === reservation) starting.delete(runKey);
-						await finishRun(serverUrl, {
-							requestLogId: result.requestLogId,
-							status: "error",
-							tokenName: config.tokenName ?? "external",
-							requestedModelName: reservation.requestedModelName,
-							totalInputTokens: 0,
-							totalOutputTokens: 0,
-							totalCacheReadTokens: 0,
-							totalCacheCreationTokens: 0,
-							totalDurationMs: Date.now() - reservation.startTime,
-							errorMessage: reservation.errorMessage ?? OVERLAPPING_RUN_ERROR
-						}).catch((e) => onLogError(e instanceof Error ? e : new Error(String(e))));
-						return;
-					}
-					let watchdogTimeout = DEFAULT_STALE_RUN_TIMEOUT_MS;
-					const timeout = event.timeout;
-					if (config.staleRunTimeoutMs !== void 0) watchdogTimeout = config.staleRunTimeoutMs;
-					else if (timeout instanceof Object) {
-						const totalMs = timeout.totalMs;
-						if (totalMs !== void 0 && totalMs > 0) watchdogTimeout = totalMs + STALE_RUN_GRACE_MS;
-					} else if (timeout !== void 0 && timeout > 0) watchdogTimeout = timeout + STALE_RUN_GRACE_MS;
+					const watchdogTimeout = staleRunTimeout(config, event.timeout);
 					let watchdog;
 					if (watchdogTimeout > 0) {
 						watchdog = setTimeout(() => {
-							finalizeRun(runKey, {
+							finalizeRun(event.callId, {
 								status: "error",
-								errorMessage: "AI SDK generation ended before onFinish was emitted"
+								errorMessage: STALE_RUN_ERROR
 							});
 						}, watchdogTimeout);
 						watchdog.unref?.();
 					}
-					let cleanupAbortListener;
-					const signal = event.abortSignal;
-					if (signal) {
-						const onAbort = () => {
-							finalizeRun(runKey, {
-								status: "aborted",
-								errorMessage: getErrorMessage(signal.reason)
-							});
-						};
-						if (signal.aborted) queueMicrotask(onAbort);
-						else {
-							signal.addEventListener("abort", onAbort, { once: true });
-							cleanupAbortListener = () => signal.removeEventListener("abort", onAbort);
-						}
-					}
-					starting.delete(runKey);
-					runs.set(runKey, {
+					runs.set(event.callId, {
 						requestLogId: result.requestLogId,
 						requestedModelName,
 						isStructured,
 						pendingSteps: [],
 						pendingToolCalls: [],
-						startTime: Date.now(),
+						startTime,
 						toolDurations: /* @__PURE__ */ new Map(),
 						watchdog,
-						cleanupAbortListener,
 						finishing: false
 					});
 				} catch (e) {
-					if (starting.get(runKey) === reservation) starting.delete(runKey);
 					onLogError(e instanceof Error ? e : new Error(String(e)));
 				}
 			},
-			onToolCallFinish(event) {
-				if (event.model?.provider === "qgrid") return;
-				const runKey = resolveRunKey(qgridRunIdSchema.safeParse(event.metadata?.qgridRunId).data, event.functionId);
-				if (suppressedRuns.has(runKey) || quarantined.has(runKey)) return;
-				const run = runs.get(runKey);
-				if (!run) return;
-				run.toolDurations.set(event.toolCall.toolCallId, Math.round(event.durationMs));
+			onToolExecutionEnd(event) {
+				runs.get(event.callId)?.toolDurations.set(event.toolCall.toolCallId, Math.round(event.toolExecutionMs));
 			},
-			onStepFinish(event) {
-				if (event.model?.provider === "qgrid") return;
-				const runKey = resolveRunKey(qgridRunIdSchema.safeParse(event.metadata?.qgridRunId).data, event.functionId);
-				if (suppressedRuns.has(runKey) || quarantined.has(runKey)) return;
-				const run = runs.get(runKey);
+			onStepEnd(event) {
+				const run = runs.get(event.callId);
 				if (!run) return;
 				const { content, usage, reasoningText, finishReason, stepNumber } = event;
-				const requestedModelName = fullModelName(event.model?.provider, event.model?.modelId) ?? run.requestedModelName;
-				const modelName = fullModelName(event.model?.provider, event.response?.modelId);
+				const requestedModelName = fullModelName(event.model.provider, event.model.modelId) ?? run.requestedModelName;
+				const modelName = fullModelName(event.model.provider, event.response.modelId);
 				const remainingPending = [];
 				for (const pending of run.pendingToolCalls) {
 					const tr = content.find((p) => p.type === "tool-result" && p.toolCallId === pending.toolCallId);
@@ -2204,7 +2140,7 @@ function createQgridLogger(config = {}) {
 							requestedModelName: pending.requestedModelName,
 							toolResult: tr && "output" in tr ? loggedPayloadText.parse(tr.output) : void 0,
 							toolDurationMs: run.toolDurations.get(pending.toolCallId),
-							error: te && "error" in te ? loggedPayloadText.parse(te.error) : void 0
+							error: te && "error" in te ? getErrorMessage(te.error) : void 0
 						}).catch((e) => onLogError(e instanceof Error ? e : new Error(String(e)))));
 						run.toolDurations.delete(pending.toolCallId);
 					} else remainingPending.push(pending);
@@ -2241,7 +2177,7 @@ function createQgridLogger(config = {}) {
 							requestedModelName,
 							toolResult: tr && "output" in tr ? loggedPayloadText.parse(tr.output) : void 0,
 							toolDurationMs: run.toolDurations.get(tc.toolCallId),
-							error: te && "error" in te ? loggedPayloadText.parse(te.error) : void 0
+							error: te && "error" in te ? getErrorMessage(te.error) : void 0
 						}).catch((e) => onLogError(e instanceof Error ? e : new Error(String(e)))));
 						run.toolDurations.delete(tc.toolCallId);
 					} else run.pendingToolCalls.push({
@@ -2255,28 +2191,31 @@ function createQgridLogger(config = {}) {
 					});
 				}
 			},
-			async onFinish(event) {
-				if (event.model?.provider === "qgrid") return;
-				const runKey = resolveRunKey(qgridRunIdSchema.safeParse(event.metadata?.qgridRunId).data, event.functionId);
-				if (suppressedRuns.has(runKey)) {
-					suppressedRuns.remove(runKey);
-					return;
-				}
-				if (quarantined.has(runKey)) return;
-				const run = runs.get(runKey);
-				if (!run) return;
-				const status = event.finishReason === "error" ? "error" : "succeeded";
-				const requestedModelName = fullModelName(event.model?.provider, event.model?.modelId) ?? run.requestedModelName;
-				const modelName = fullModelName(event.model?.provider, event.response?.modelId);
-				const result = {
-					status,
+			async onEnd(event) {
+				const run = runs.get(event.callId);
+				if (!run || !isTextEnd(event)) return;
+				await finalizeRun(event.callId, {
+					status: event.finishReason === "error" ? "error" : "succeeded",
 					response: event.text,
-					totalUsage: event.totalUsage,
-					modelName,
-					requestedModelName
-				};
-				if (status === "error" && "error" in event) result.errorMessage = getErrorMessage(event.error);
-				await finalizeRun(runKey, result);
+					totalUsage: event.usage,
+					modelName: fullModelName(event.model.provider, event.finalStep.response.modelId),
+					requestedModelName: fullModelName(event.model.provider, event.model.modelId) ?? run.requestedModelName
+				});
+			},
+			async onError(event) {
+				const parsed = errorEventSchema.safeParse(event);
+				if (!parsed.success) return;
+				const { callId, error } = parsed.data;
+				await finalizeRun(callId, {
+					status: abortErrorSchema.safeParse(error).success ? "aborted" : "error",
+					errorMessage: getErrorMessage(error)
+				});
+			},
+			async onAbort(event) {
+				await finalizeRun(event.callId, {
+					status: "aborted",
+					errorMessage: getErrorMessage(event.reason)
+				});
 			}
 		}]
 	};
@@ -2665,13 +2604,16 @@ function qgrid(modelId, config) {
 			if (tokenName) args.tokenName = tokenName;
 			if (runContext) args.runContext = runContext;
 			if (toolResultsPayload) args.toolResults = toolResultsPayload;
-			const init = {
+			const request = {
 				method: "POST",
 				headers: { "Content-Type": "application/json" },
 				body: JSON.stringify({ args }),
 				signal: options.abortSignal
 			};
-			if (transport) init.dispatcher = transport.dispatcher;
+			const init = transport ? {
+				...request,
+				dispatcher: transport.dispatcher
+			} : request;
 			const data = await qgridFetch(`${serverUrl}/api/qgrid/query`, init, {
 				operation: "query",
 				serverUrl,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@cartanova/qgrid-ai-sdk",
-  "version": "2.12.0",
+  "version": "2.14.0",
   "description": "AI SDK LanguageModelV3 and ImageModelV3 provider for qgrid",
   "repository": {
     "type": "git",
@@ -20,21 +20,24 @@
     }
   },
   "dependencies": {
-    "undici": "^6.28.0"
+    "undici": "^7.30.1"
   },
   "devDependencies": {
-    "@ai-sdk/provider": "^3.0.0",
-    "ai": "^6.0.190",
-    "tsdown": "^0.22.3",
-    "typescript": "^5.9.3",
-    "vitest": "^4.1.10",
-    "zod": "4.4.3"
+    "@ai-sdk/provider": "^4.0.0",
+    "ai": "^7.0.0",
+    "tsdown": "^0.23.1",
+    "typescript": "npm:@typescript/typescript6@6.0.2",
+    "vitest": "^5.0.5",
+    "zod": "^4.6.7"
   },
   "peerDependencies": {
-    "@ai-sdk/provider": "^3.0.0",
-    "ai": "^6.0.0",
+    "@ai-sdk/provider": "^4.0.0",
+    "ai": "^7.0.0",
     "zod": "^3.25.76 || ^4.1.8"
   },
+  "engines": {
+    "node": ">=22"
+  },
   "scripts": {
     "build": "tsdown",
     "test": "vitest run",