@cartanova/qgrid-cli 2.3.1 → 2.3.3

package/bundle/src/utils/providers/common/model-cost.test.ts

@@ -1,6 +1,6 @@
 import { describe, expect, it } from "vitest";
 
-import { calculateCostUsd } from "./model-cost";
+import { calculateCostUsd, getModelCosts } from "./model-cost";
 
 describe("calculateCostUsd", () => {
   it("Anthropic cache read/write 를 전체 입력에서 분리해 각각 단가를 적용한다", () => {
@@ -36,4 +36,22 @@ describe("calculateCostUsd", () => {
     expect(cost).toBeGreaterThan(0);
     expect(cost).toBeCloseTo(0.0027222, 10);
   });
+
+  it("[1m] suffix 는 cost lookup 에서 strip 하고 long-context 할증은 붙이지 않는다", () => {
+    const base = getModelCosts("claude-sonnet-4-6");
+    const suffixed = getModelCosts("claude-sonnet-4-6[1m]");
+    expect(suffixed).toBe(base);
+    expect(suffixed.longContext).toBeUndefined();
+
+    const usage = {
+      inputTokens: 250_000,
+      outputTokens: 1_000,
+      cachedInputTokens: 200_000,
+      cacheCreationInputTokens: 0,
+    };
+    expect(calculateCostUsd("claude-sonnet-4-6[1m]", usage)).toBeCloseTo(
+      calculateCostUsd("claude-sonnet-4-6", usage),
+      10,
+    );
+  });
 });

package/bundle/src/utils/providers/common/model-cost.ts

@@ -164,7 +164,8 @@ const ANTHROPIC_COSTS: Record<string, ModelCosts> = {
 const DEFAULT_COSTS: ModelCosts = { inputTokens: 3, outputTokens: 15, cachedInputTokens: 0.3 };
 
 export function getModelCosts(model: string): ModelCosts {
-  return OPENAI_COSTS[model] ?? ANTHROPIC_COSTS[model] ?? DEFAULT_COSTS;
+  const normalizedModel = model.replace(/\[1m\]$/i, "");
+  return OPENAI_COSTS[normalizedModel] ?? ANTHROPIC_COSTS[normalizedModel] ?? DEFAULT_COSTS;
 }
 
 export function calculateCostUsd(

package/bundle/src/utils/providers/common/provider-dispatcher.ts

@@ -57,16 +57,19 @@ export interface GenerateResult {
   threadCoord: ReuseThreadCoord;
 }
 
-// 상위(qgrid.dispatcher)로 가는 스트림 콜백. onComplete 는 non-stream 의 GenerateResult 와
-// 동일 shape(tokenName/threadCoord 포함)를 받아, 두 경로가 같은 일급 타입을 공유한다.
-export interface GenerateStreamCallbacks {
+// 스트림 콜백 컨테이너. 계층별로 onComplete payload 만 다르고, 스트림 이벤트 shape 는 동일하다.
+export interface StreamCallbacks<TComplete> {
   onDelta: (text: string) => void;
-  onComplete: (result: GenerateResult) => void;
+  onComplete: (result: TComplete) => void;
   onError: (error: Error) => void;
   onThreadId?: (threadId: string) => void;
   onTurnId?: (turnId: string) => void;
 }
 
+// 상위(qgrid.dispatcher)로 가는 스트림 콜백. onComplete 는 non-stream 의 GenerateResult 와
+// 동일 shape(tokenName/threadCoord 포함)를 받아, 두 경로가 같은 일급 타입을 공유한다.
+export type GenerateStreamCallbacks = StreamCallbacks<GenerateResult>;
+
 export interface ProviderDispatcher {
   generate(req: GenerateRequest): Promise<GenerateResult>;
   start(): Promise<void>;

package/bundle/src/utils/providers/openai/codex-worker.ts

@@ -9,6 +9,7 @@ import { type ThreadStartResponse } from "../../../codex-protocol/v2/ThreadStart
 import { type TokenUsageBreakdown } from "../../../codex-protocol/v2/TokenUsageBreakdown";
 import { type TurnStartParams } from "../../../codex-protocol/v2/TurnStartParams";
 import { type TurnStartResponse } from "../../../codex-protocol/v2/TurnStartResponse";
+import { type StreamCallbacks as CommonStreamCallbacks } from "../common/provider-dispatcher";
 import { CodexRpcClient } from "./codex-rpc";
 
 const logger = getLogger(["qgrid", "codex-worker"]);
@@ -46,13 +47,7 @@ export interface TurnResult {
   model: string;
 }
 
-export interface StreamCallbacks {
-  onDelta: (text: string) => void;
-  onComplete: (result: TurnResult) => void;
-  onError: (error: Error) => void;
-  onThreadId?: (threadId: string) => void;
-  onTurnId?: (turnId: string) => void;
-}
+export type StreamCallbacks = CommonStreamCallbacks<TurnResult>;
 
 type RefreshableWorkerCredentials = Pick<WorkerConfig, "accessToken" | "accountId" | "planType">;
 type ActiveTurnAbort = (error: Error) => void;
@@ -191,7 +186,6 @@ export class CodexAppServerWorker {
     await this.waitForLoginCompleted();
     this.ready = true;
     this.restartAttempts = 0;
-    logger.info(`worker ${this.config.tokenName} ready`);
   }
 
   async startBrowserLogin(): Promise<string> {

package/bundle/src/utils/providers/openai/openai-dispatcher.ts

@@ -89,12 +89,16 @@ export class OpenAIDispatcher implements ProviderDispatcher {
 
   async onTokenRemoved(id: number): Promise<void> {
     const workers = this.workerPool.get(id) ?? [];
+    const tokenName = workers[0]?.tokenName;
     this.workerPool.delete(id);
     await Promise.allSettled(workers.map((w) => w.kill()));
     if (this.getAllReadyActiveWorkers().length === 0) {
       this.rejectAllQueued("NO_OPENAI_WORKERS");
     }
-    if (workers.length > 0) logger.info(`workers removed for token ${id}: ${workers.length}`);
+    if (workers.length > 0) {
+      const label = tokenName ?? `token ${id}`;
+      logger.info(`workers removed: ${label} (${workers.length})`);
+    }
   }
 
   async onTokenUpdated(id: number, name: string, credentials: OpenAICredentials): Promise<void> {
@@ -106,7 +110,7 @@ export class OpenAIDispatcher implements ProviderDispatcher {
 
     if (existing.every((w) => w.canReuseForToken(name, credentials))) {
       existing.forEach((w) => w.updateTokenState(name, credentials));
-      logger.info(`workers updated in-place for token ${id} (${name})`);
+      logger.info(`workers updated in-place: ${name}`);
       this.drainQueue();
       return;
     }
@@ -117,21 +121,27 @@ export class OpenAIDispatcher implements ProviderDispatcher {
   }
 
   onTokenDeactivated(id: number): void {
-    (this.workerPool.get(id) ?? []).forEach((w) => {
+    const workers = this.workerPool.get(id) ?? [];
+    const tokenName = workers[0]?.tokenName;
+    workers.forEach((w) => {
       w.active = false;
     });
     if (this.getAllReadyActiveWorkers().length === 0) {
       this.rejectAllQueued("NO_ACTIVE_WORKERS");
     }
-    logger.info(`workers deactivated: token ${id}`);
+    const label = tokenName ?? `token ${id}`;
+    logger.info(`workers deactivated: ${label}`);
   }
 
   onTokenActivated(id: number): void {
-    (this.workerPool.get(id) ?? []).forEach((w) => {
+    const workers = this.workerPool.get(id) ?? [];
+    const tokenName = workers[0]?.tokenName;
+    workers.forEach((w) => {
       w.active = true;
     });
     this.drainQueue();
-    logger.info(`workers activated: token ${id}`);
+    const label = tokenName ?? `token ${id}`;
+    logger.info(`workers activated: ${label}`);
   }
 
   // ── Generate ────────────────────────────────────────────────────
@@ -403,10 +413,7 @@ export class OpenAIDispatcher implements ProviderDispatcher {
       const worker = await this.spawnSingleWorker(tokenId, tokenName, credentials, i);
       if (worker) workers.push(worker);
     }
-    if (workers.length > 0) {
-      this.workerPool.set(tokenId, workers);
-      logger.info(`${workers.length}/${WORKERS_PER_TOKEN} workers spawned for ${tokenName}`);
-    }
+    if (workers.length > 0) this.workerPool.set(tokenId, workers);
   }
 
   async spawnSingleWorker(
@@ -435,7 +442,7 @@ export class OpenAIDispatcher implements ProviderDispatcher {
 
     try {
       await worker.initialize();
-      logger.info(`worker spawned: ${tokenName}[${workerIndex}] (id=${tokenId + 1})`);
+      logger.info(`worker spawned: ${tokenName}[${workerIndex}]`);
       return worker;
     } catch (e) {
       logger.warn(`worker spawn failed: ${tokenName}[${workerIndex}]: ${(e as Error).message}`);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@cartanova/qgrid-cli",
-  "version": "2.3.1",
+  "version": "2.3.3",
   "repository": {
     "type": "git",
     "url": "https://github.com/cartanova-ai/qgrid"