replicas-engine 0.1.987 → 0.1.989

package/dist/src/{asp-host-LOTCGOVD.js → asp-host-O3JY7A6A.js}

@@ -5,12 +5,12 @@ import {
   getCodexAspHost,
   restartCodexAspHost,
   restartCodexAspHostIfRunning
-} from "./chunk-D5B72SUV.js";
-import "./chunk-TYH3IEXZ.js";
-import "./chunk-KAJYMTYA.js";
-import "./chunk-2PMHC6IL.js";
+} from "./chunk-VUUSZ5JS.js";
+import "./chunk-LM4J2GLY.js";
+import "./chunk-BAJGQWIO.js";
+import "./chunk-TFUEJW43.js";
 import "./chunk-UZSNFLDQ.js";
-import "./chunk-QU4FXL4X.js";
+import "./chunk-RZBMLP2H.js";
 import "./chunk-VEQXQN22.js";
 export {
   getCodexAspHost,

package/dist/src/{chunk-KAJYMTYA.js → chunk-BAJGQWIO.js}

@@ -3,10 +3,10 @@ import { createRequire as __createRequire } from 'node:module';
 const require = __createRequire(import.meta.url);
 import {
   execFileAsync
-} from "./chunk-2PMHC6IL.js";
+} from "./chunk-TFUEJW43.js";
 import {
   isVersionBelow
-} from "./chunk-QU4FXL4X.js";
+} from "./chunk-RZBMLP2H.js";
 
 // src/utils/codex-agent-env.ts
 function buildCodexAgentEnv(source = process.env) {
@@ -240,7 +240,7 @@ function getCodexModelProvider(method) {
   if (method === "foundry") return "azure";
   return method === "ollama" || method === "openai-compatible" ? "replicas_openai_compatible" : "openai";
 }
-var ENGINE_PACKAGE_VERSION = "0.1.987";
+var ENGINE_PACKAGE_VERSION = "0.1.989";
 var INITIALIZE_METHOD = "initialize";
 var INITIALIZED_NOTIFICATION = "initialized";
 var ACCOUNT_LOGIN_START_METHOD = "account/login/start";

package/dist/src/{chunk-G6QBYUZR.js → chunk-JBMI42EJ.js}

@@ -3,18 +3,18 @@ import { createRequire as __createRequire } from 'node:module';
 const require = __createRequire(import.meta.url);
 import {
   monolithRequest
-} from "./chunk-TYH3IEXZ.js";
+} from "./chunk-LM4J2GLY.js";
 import {
   SUBPROCESS_MAX_BUFFER,
   execFileAsync
-} from "./chunk-2PMHC6IL.js";
+} from "./chunk-TFUEJW43.js";
 import {
   extractCommandProtectionCommandText,
   findCommandProtectionGitLookups,
   findGitCommitSignals,
   findPrMergeSignals,
   parseCommandProtectionRequest
-} from "./chunk-QU4FXL4X.js";
+} from "./chunk-RZBMLP2H.js";
 
 // src/services/command-protection-service.ts
 import { readFile, stat } from "fs/promises";

package/dist/src/{chunk-TYH3IEXZ.js → chunk-LM4J2GLY.js}

@@ -14,7 +14,7 @@ import {
   isValidAgentProvider,
   parsePosixEnvFile,
   readReplicasRuntimeEnv
-} from "./chunk-QU4FXL4X.js";
+} from "./chunk-RZBMLP2H.js";
 
 // src/runtime-env-loader.ts
 import { readFileSync } from "fs";

package/dist/src/{chunk-QU4FXL4X.js → chunk-RZBMLP2H.js}

@@ -3274,7 +3274,7 @@ replicas computer record stop
 replicas media upload /tmp/walkthrough.mp4 --name "<Project> walkthrough"
 \`\`\`
 
-Keep the recording short (one to three minutes) and focused on the flow that proves the environment works. \`replicas media upload\` prints a chat embed line and a \`View in Replicas\` link ending in \`?media=<media-id>\`; keep both. The media id goes into the proposal, and the embed line goes into your final message. Follow \`MEDIA.md\` for anything shared outside the dashboard.
+Keep the recording short (one to three minutes) and focused on the flow that proves the environment works. \`replicas media upload\` prints a chat embed line and a \`View in Replicas\` link ending in \`?mode=media&media=<media-id>\`; keep both. The media id goes into the proposal, and the embed line goes into your final message. Follow \`MEDIA.md\` for anything shared outside the dashboard.
 
 Add screenshots of important states (a passing test run, the running app) with \`replicas computer screenshot\` and upload them the same way when they help the user trust the result.
 
@@ -3798,7 +3798,7 @@ For each media file, the CLI prints an embed line and a dashboard link:
 
 1. Images always print \`![filename](<public-url>)\`. This public, stable link works in Replicas chat, GitHub, and anywhere else Markdown images render. No flag is needed.
 2. Video and audio print \`Replicas chat embed: ![filename](<api-url>)\` by default; this authenticated URL renders inline in Replicas chat **only** \u2014 see below. \`--access public\` prints a public embed instead.
-3. \`View in Replicas: <deep-link>\` is a per-file dashboard URL of the form \`https://app.replicas.dev/workspaces/<workspace-id>?media=<media-id>\` that opens directly to that specific file.
+3. \`View in Replicas: <deep-link>\` is a per-file dashboard URL of the form \`https://app.replicas.dev/workspaces/<workspace-id>?mode=media&media=<media-id>\` that opens directly to that specific file.
 
 Match each "View in Replicas" line to the embed line directly above it \u2014 that's the deep link for that file.
 
@@ -3829,7 +3829,7 @@ Permanently delete an upload with \`replicas media delete <media-id>\`. The back
 | \`https://api.replicas.dev/v1/media/<id>\` (inside \`![\u2026](\u2026)\`) | Authenticated API redirect printed for private video, audio, and HTML \u2014 requires the Replicas chat session | **Only** in the Replicas chat reply. Nowhere else. |
 | \`https://replicas.dev/organizations/<organization-id>/media/<media-id>\` | Organization-only page printed by \`--access organization\` | Pull requests and other places where every viewer belongs to the Replicas organization. |
 | \`https://api.replicas.dev/v1/media/public/<id>/<token>\` | Public bearer URL printed for every image, and for video or audio with \`--access public\` | The Replicas chat reply and GitHub embeds. |
-| \`https://app.replicas.dev/workspaces/<workspace-id>?media=<media-id>\` | Authenticated dashboard deep link that opens the media file in the workspace | Everywhere external: Slack, GitHub, GitLab, Linear, PR bodies, comments, emails. |
+| \`https://app.replicas.dev/workspaces/<workspace-id>?mode=media&media=<media-id>\` | Authenticated dashboard deep link that opens the media file in the workspace | Everywhere external: Slack, GitHub, GitLab, Linear, PR bodies, comments, emails. |
 
 ### NEVER link the API redirect URL outside the Replicas chat reply
 
@@ -3848,10 +3848,10 @@ This rule has no exceptions:
 For an organization-only HTML page, use the \`Forge link\` printed by \`--access organization\`. For other authenticated media references, use the dashboard form:
 
 \`\`\`
-https://app.replicas.dev/workspaces/<workspace-id>?media=<media-id>
+https://app.replicas.dev/workspaces/<workspace-id>?mode=media&media=<media-id>
 \`\`\`
 
-The \`media=<media-id>\` query param opens that specific file in the workspace.
+The \`mode=media&media=<media-id>\` query params open that specific file in the workspace.
 
 When you have a specific file in mind, never link the bare workspace URL without the \`media=<id>\` param.
 
@@ -3861,6 +3861,8 @@ Always render as a markdown / mrkdwn hyperlink labeled **View in Replicas** \u20
 
 Include each markdown embed line **verbatim** where you want that file to render inline. The chat substitutes each one with an embedded image, video, or audio player. For HTML, use the printed \`HTML page\` link. Multiple embeds can appear in a single reply.
 
+Never reply with only dashboard links (for example a "Links:" list of \`?mode=media&media=\` URLs) in place of the embeds: a dashboard link renders as plain text, so the user sees nothing inline. This applies even when the task came from Slack, Linear, GitHub, or GitLab; the chat reply still gets the embeds.
+
 After (or alongside) the embeds, include a \`[View in Canvas](<deep-link>)\` hyperlink for each file using the per-file dashboard URL printed by the CLI as **View in Replicas**. Keep that URL unchanged so it opens the specific media item in the workspace's Canvas. External messages still label this dashboard link **View in Replicas**.
 
 ## On external platforms: embed the file AND link the dashboard
@@ -3868,7 +3870,7 @@ After (or alongside) the embeds, include a \`[View in Canvas](<deep-link>)\` hyp
 After \`replicas media upload\` succeeds, external delivery has two required follow-up steps for every file:
 
 1. **Use the destination's inline representation:** upload the raw bytes to Slack, Linear, or GitLab; embed the image's printed public link on GitHub.
-2. **Link the Replicas dashboard deep link** (\`?media=<id>\` form) so the user can jump straight to the file in the workspace.
+2. **Link the Replicas dashboard deep link** (\`?mode=media&media=<id>\` form) so the user can jump straight to the file in the workspace.
 
 A dashboard link alone is not enough. An embed without the dashboard link is not enough. Do both.
 
@@ -3897,7 +3899,7 @@ await replicas.slack.call('files.completeUploadExternal', {
   files: [{ id: reservation.file_id, title: 'Screenshot' }],
   channel_id: 'CHANNEL_ID',
   thread_ts: 'THREAD_TS',
-  initial_comment: '<https://app.replicas.dev/workspaces/<workspace-id>?media=<media-id>|View in Replicas>',
+  initial_comment: '<https://app.replicas.dev/workspaces/<workspace-id>?mode=media&media=<media-id>|View in Replicas>',
 });
 \`\`\`
 
@@ -3920,7 +3922,7 @@ Use \`jq\` to build the JSON bodies \u2014 the GraphQL variables (\`size\`, \`fi
 FILE=/abs/path/to/screenshot.png
 FILENAME=$(basename "$FILE")
 SIZE=$(stat -c%s "$FILE" 2>/dev/null || stat -f%z "$FILE")
-CONTENT_TYPE=image/png
+CONTENT_TYPE=image/png   # the file's real MIME type, e.g. video/mp4 for a recording
 ISSUE_UUID=...           # the Linear issue's UUID
 MEDIA_ID=...             # the media ID printed by \`replicas media upload\`
 
@@ -3957,7 +3959,7 @@ curl -s -X PUT "$UPLOAD_URL" \\
   --data-binary @"$FILE"
 
 # 3. Reference assetUrl inline in the comment body, alongside the dashboard link.
-BODY=$(printf '![screenshot](%s)\\n\\n[View in Replicas](https://app.replicas.dev/workspaces/%s?media=%s)' \\
+BODY=$(printf '![screenshot](%s)\\n\\n[View in Replicas](https://app.replicas.dev/workspaces/%s?mode=media&media=%s)' \\
   "$ASSET_URL" "$REPLICAS_WORKSPACE_ID" "$MEDIA_ID")
 
 jq -n \\
@@ -3969,6 +3971,8 @@ jq -n \\
     }' | linear_graphql
 \`\`\`
 
+For a recording, upload the MP4 with its own MIME type, then upload a poster frame (\`ffmpeg -i /tmp/demo.mp4 -frames:v 1 /tmp/demo-poster.png\`) the same way as \`image/png\` and link it: \`[![Demo recording](<poster-asset-url>)](<recording-asset-url>)\`.
+
 The same flow works for issue bodies (use \`issueCreate\` / \`issueUpdate\` with the same markdown body).
 
 Never paste the \`api.replicas.dev/v1/media/<id>\` URL into a Linear body or comment \u2014 it will render as a broken image to anyone outside your Replicas chat.
@@ -4008,7 +4012,7 @@ Do **not** commit screenshots to the repo, use placeholder URLs, expose the auth
 
 GitLab supports native project Markdown uploads. For a GitLab destination, always upload the media to Replicas first, then POST the same original file to the repository's \`/projects/:id/uploads\` API using the workspace GitLab credential. Insert the returned \`markdown\` value into the merge request body or comment and include the Replicas dashboard link. Image uploads render inline; MP4, MOV, and WebM uploads render as inline video players.
 
-Do not use a Replicas public link for GitLab when the native project upload succeeds; native uploads preserve project access controls.
+Do not use a Replicas public link for GitLab when the native project upload succeeds; native uploads preserve project access controls. \`replicas.gitlab.request\` only sends JSON and cannot POST multipart file bytes, so without another supported upload path use the GitHub flow above instead: the image's printed public embed, and a poster linked to the recording's **View in Replicas** URL.
 
 ## Recording defaults
 
@@ -4056,7 +4060,7 @@ replicas media upload chart-a.svg chart-b.svg --name "Latency chart" --name "Err
 var MEDIA_ABILITY = {
   label: "Media",
   description: "Deliver media to Replicas and destination workflows with inline uploads.",
-  bullet: "- Producing or showing the user any media \u2014 screenshots, screen recordings, generated images or diagrams, audio clips, or HTML recap pages \u2014 including in your Replicas chat reply, PR descriptions/comments, and other external platforms",
+  bullet: "- Producing or showing the user any media \u2014 screenshots, screen recordings, generated images or diagrams, audio clips, or HTML recap pages \u2014 including in your Replicas chat reply, PR descriptions/comments, and other external platforms. Paste the printed `![\u2026](\u2026)` embeds in your chat reply, never only dashboard links, and upload the file itself to Slack, Linear, or GitLab",
   section: SECTION7,
   referenceFile: { name: "MEDIA.md", content: REFERENCE7 }
 };
@@ -7180,10 +7184,6 @@ function detectLanguageByPath(filePath) {
 }
 
 // ../shared/src/context-usage.ts
-var CODEX_CATEGORY_COLORS = {
-  input: "#3eeba3",
-  output: "#56b6c2"
-};
 function isRecord2(value) {
   return typeof value === "object" && value !== null;
 }
@@ -7214,9 +7214,7 @@ function normalizeAutoCompactThreshold(threshold, maxTokens) {
   return clampPercentage(value);
 }
 function clampTokensToWindow(rawTokens, maxTokens) {
-  if (!maxTokens || maxTokens <= 0) return { totalTokens: rawTokens };
-  if (rawTokens <= maxTokens) return { totalTokens: rawTokens };
-  return { totalTokens: maxTokens, totalProcessedTokens: rawTokens };
+  return maxTokens && maxTokens > 0 ? Math.min(rawTokens, maxTokens) : rawTokens;
 }
 function compactCategories(categories) {
   return categories.filter((category) => Boolean(category && category.tokens > 0)).map((category) => ({
@@ -7228,16 +7226,12 @@ function buildCodexTokenUsageContextUsagePayload(usage) {
   const maxTokens = usage.modelContextWindow;
   const inputTokens = usage.last.inputTokens;
   const outputTokens = usage.last.outputTokens;
-  const lastTurnTokens = usage.last.totalTokens;
-  const { totalTokens } = clampTokensToWindow(lastTurnTokens, maxTokens);
-  const cumulativeTotalTokens = usage.total?.totalTokens ?? null;
-  const totalProcessedTokens = cumulativeTotalTokens !== null && cumulativeTotalTokens > totalTokens ? cumulativeTotalTokens : clampTokensToWindow(lastTurnTokens, maxTokens).totalProcessedTokens;
+  const totalTokens = clampTokensToWindow(usage.last.totalTokens, maxTokens);
   return {
     provider: "codex",
     source: "codex_token_count",
     ...usage.model !== void 0 ? { model: usage.model } : {},
     totalTokens,
-    ...totalProcessedTokens !== void 0 ? { totalProcessedTokens } : {},
     maxTokens,
     rawMaxTokens: maxTokens,
     percentage: percentage(totalTokens, maxTokens),
@@ -7246,23 +7240,14 @@ function buildCodexTokenUsageContextUsagePayload(usage) {
       {
         name: "Input context",
         tokens: inputTokens,
-        percentage: percentage(inputTokens, maxTokens),
-        color: CODEX_CATEGORY_COLORS.input
+        percentage: percentage(inputTokens, maxTokens)
       },
       {
         name: "Generated output",
         tokens: outputTokens,
-        percentage: percentage(outputTokens, maxTokens),
-        color: CODEX_CATEGORY_COLORS.output
+        percentage: percentage(outputTokens, maxTokens)
       }
     ]),
-    apiUsage: {
-      inputTokens,
-      outputTokens,
-      ...usage.last.cachedInputTokens !== null && usage.last.cachedInputTokens !== void 0 ? { cachedInputTokens: usage.last.cachedInputTokens } : {},
-      ...usage.last.reasoningOutputTokens !== null && usage.last.reasoningOutputTokens !== void 0 ? { reasoningOutputTokens: usage.last.reasoningOutputTokens } : {},
-      ...cumulativeTotalTokens !== null ? { totalTokens: cumulativeTotalTokens } : {}
-    },
     updatedAt: usage.updatedAt
   };
 }
@@ -7288,10 +7273,7 @@ function coerceContextUsagePayload(payload, timestamp) {
   const reportedMaxTokens = numberOrNull(payload.maxTokens);
   const { maxTokens, usedKnownWindow } = resolveClaudeContextWindow(provider, model, reportedMaxTokens);
   const rawMaxTokens = usedKnownWindow ? maxTokens : numberOrNull(payload.rawMaxTokens);
-  const clamped = clampTokensToWindow(rawTotalTokens, maxTokens);
-  const totalTokens = clamped.totalTokens;
-  const payloadTotalProcessedTokens = numberOrNull(payload.totalProcessedTokens);
-  const totalProcessedTokens = payloadTotalProcessedTokens !== null ? payloadTotalProcessedTokens : clamped.totalProcessedTokens ?? null;
+  const totalTokens = clampTokensToWindow(rawTotalTokens, maxTokens);
   const source = payload.source === "codex_token_count" || payload.source === "claude_usage" || payload.source === "provider_usage" ? payload.source : "claude_context";
   const categories = Array.isArray(payload.categories) ? compactCategories(payload.categories.map((item) => {
     if (!isRecord2(item)) return null;
@@ -7299,23 +7281,13 @@ function coerceContextUsagePayload(payload, timestamp) {
     const tokens = numberOrNull(item.tokens);
     const categoryPercentage = numberOrNull(item.percentage);
     if (!name || tokens === null || categoryPercentage === null) return null;
+    if (item.isDeferred === true || name === "Free space" || name === "Autocompact buffer") return null;
     return {
       name,
       tokens,
-      percentage: usedKnownWindow ? percentage(tokens, maxTokens) : categoryPercentage,
-      ...typeof item.color === "string" ? { color: item.color } : {},
-      ...typeof item.isDeferred === "boolean" ? { isDeferred: item.isDeferred } : {}
+      percentage: usedKnownWindow ? percentage(tokens, maxTokens) : categoryPercentage
     };
   })) : [];
-  const apiUsage = isRecord2(payload.apiUsage) ? {
-    ...numberOrNull(payload.apiUsage.inputTokens) !== null ? { inputTokens: numberOrNull(payload.apiUsage.inputTokens) } : {},
-    ...numberOrNull(payload.apiUsage.outputTokens) !== null ? { outputTokens: numberOrNull(payload.apiUsage.outputTokens) } : {},
-    ...numberOrNull(payload.apiUsage.cacheCreationInputTokens) !== null ? { cacheCreationInputTokens: numberOrNull(payload.apiUsage.cacheCreationInputTokens) } : {},
-    ...numberOrNull(payload.apiUsage.cacheReadInputTokens) !== null ? { cacheReadInputTokens: numberOrNull(payload.apiUsage.cacheReadInputTokens) } : {},
-    ...numberOrNull(payload.apiUsage.cachedInputTokens) !== null ? { cachedInputTokens: numberOrNull(payload.apiUsage.cachedInputTokens) } : {},
-    ...numberOrNull(payload.apiUsage.reasoningOutputTokens) !== null ? { reasoningOutputTokens: numberOrNull(payload.apiUsage.reasoningOutputTokens) } : {},
-    ...numberOrNull(payload.apiUsage.totalTokens) !== null ? { totalTokens: numberOrNull(payload.apiUsage.totalTokens) } : {}
-  } : null;
   const compactsAutomatically = typeof payload.compactsAutomatically === "boolean" ? payload.compactsAutomatically : void 0;
   const autoCompactThreshold = normalizeAutoCompactThreshold(payload.autoCompactThreshold, maxTokens);
   return {
@@ -7323,14 +7295,12 @@ function coerceContextUsagePayload(payload, timestamp) {
     source,
     ...model ? { model } : {},
     totalTokens,
-    ...totalProcessedTokens !== null ? { totalProcessedTokens } : {},
     maxTokens,
     rawMaxTokens,
     percentage: usedKnownWindow ? percentage(totalTokens, maxTokens) : clampPercentage(percentageValue),
     ...compactsAutomatically !== void 0 ? { compactsAutomatically } : {},
     ...autoCompactThreshold !== null ? { autoCompactThreshold } : {},
     categories,
-    apiUsage,
     updatedAt: typeof payload.updatedAt === "string" ? payload.updatedAt : timestamp
   };
 }
@@ -7341,20 +7311,15 @@ function codexTokenCountToContextUsage(event) {
   if (!info) return null;
   const lastUsage = isRecord2(info.last_token_usage) ? info.last_token_usage : null;
   if (!lastUsage) return null;
-  const totalUsage = isRecord2(info.total_token_usage) ? info.total_token_usage : null;
   const maxTokens = numberOrNull(info.model_context_window);
   const inputTokens = numberOrNull(lastUsage.input_tokens) ?? 0;
   const outputTokens = numberOrNull(lastUsage.output_tokens) ?? 0;
-  const cumulativeTotalTokens = numberOrNull(totalUsage?.total_tokens);
   return buildCodexTokenUsageContextUsagePayload({
     last: {
       inputTokens,
       outputTokens,
-      totalTokens: numberOrNull(lastUsage.total_tokens) ?? inputTokens + outputTokens,
-      cachedInputTokens: numberOrNull(lastUsage.cached_input_tokens),
-      reasoningOutputTokens: numberOrNull(lastUsage.reasoning_output_tokens)
+      totalTokens: numberOrNull(lastUsage.total_tokens) ?? inputTokens + outputTokens
     },
-    total: cumulativeTotalTokens !== null ? { totalTokens: cumulativeTotalTokens } : null,
     modelContextWindow: maxTokens,
     updatedAt: event.timestamp
   });

package/dist/src/{chunk-2PMHC6IL.js → chunk-TFUEJW43.js}

@@ -3,7 +3,7 @@ import { createRequire as __createRequire } from 'node:module';
 const require = __createRequire(import.meta.url);
 import {
   HOOK_EXEC_MAX_BUFFER_BYTES
-} from "./chunk-QU4FXL4X.js";
+} from "./chunk-RZBMLP2H.js";
 
 // src/utils/exec.ts
 import { exec, execFile } from "child_process";

package/dist/src/{chunk-D5B72SUV.js → chunk-VUUSZ5JS.js}

@@ -5,15 +5,15 @@ import {
   ENGINE_ENV,
   monolithRequest,
   setAgentCredentialSnapshot
-} from "./chunk-TYH3IEXZ.js";
+} from "./chunk-LM4J2GLY.js";
 import {
   AppServerProcess,
   buildCodexAgentEnv
-} from "./chunk-KAJYMTYA.js";
+} from "./chunk-BAJGQWIO.js";
 import {
   SUBPROCESS_MAX_BUFFER,
   execFileAsync
-} from "./chunk-2PMHC6IL.js";
+} from "./chunk-TFUEJW43.js";
 import {
   isRecord
 } from "./chunk-UZSNFLDQ.js";
@@ -27,7 +27,7 @@ import {
   listOpenAICompatibleModels,
   readOllamaAuthEnv,
   resolveOpenAICompatibleModelsUrl
-} from "./chunk-QU4FXL4X.js";
+} from "./chunk-RZBMLP2H.js";
 
 // src/managers/codex-token-manager.ts
 import { promises as fs } from "fs";
@@ -260,7 +260,7 @@ var CodexTokenManager = class extends BaseRefreshManager {
     await this.applyCredentialUpdate(++this.credentialGeneration, async () => {
       await this.applyCredentialsResponse(data);
       setAgentCredentialSnapshot("codex", data);
-      const { restartCodexAspHostIfRunning: restartCodexAspHostIfRunning2 } = await import("./asp-host-LOTCGOVD.js");
+      const { restartCodexAspHostIfRunning: restartCodexAspHostIfRunning2 } = await import("./asp-host-O3JY7A6A.js");
       await restartCodexAspHostIfRunning2();
     });
     await this.start(true);
@@ -280,7 +280,7 @@ var CodexTokenManager = class extends BaseRefreshManager {
     await this.applyCredentialUpdate(generation, async () => {
       await this.applyCredentialsResponse(data);
       if (restartOnChange && CODEX_AUTH_ENV_KEYS.some((key, index) => process.env[key] !== previousEnv[index])) {
-        const { restartCodexAspHostIfRunning: restartCodexAspHostIfRunning2 } = await import("./asp-host-LOTCGOVD.js");
+        const { restartCodexAspHostIfRunning: restartCodexAspHostIfRunning2 } = await import("./asp-host-O3JY7A6A.js");
         await restartCodexAspHostIfRunning2();
       }
       setAgentCredentialSnapshot("codex", data);

package/dist/src/{chunk-5EHMJ4OT.js → chunk-WYERDOJD.js}

@@ -5,7 +5,7 @@ import {
   extractCommandProtectionCommandText,
   isRecord,
   parseCodeHostPrUrl
-} from "./chunk-QU4FXL4X.js";
+} from "./chunk-RZBMLP2H.js";
 
 // src/services/published-pull-request-urls.ts
 function publishedPullRequestUrls(result) {

package/dist/src/command-protection-hook.js

@@ -3,13 +3,13 @@ import { createRequire as __createRequire } from 'node:module';
 const require = __createRequire(import.meta.url);
 import {
   evaluateCommandProtection
-} from "./chunk-G6QBYUZR.js";
-import "./chunk-TYH3IEXZ.js";
-import "./chunk-2PMHC6IL.js";
+} from "./chunk-JBMI42EJ.js";
+import "./chunk-LM4J2GLY.js";
+import "./chunk-TFUEJW43.js";
 import {
   isRecord
 } from "./chunk-UZSNFLDQ.js";
-import "./chunk-QU4FXL4X.js";
+import "./chunk-RZBMLP2H.js";
 import "./chunk-VEQXQN22.js";
 
 // src/command-protection-hook.ts

package/dist/src/headless-agent.js

@@ -7,12 +7,12 @@ import {
 import {
   AppServerProcess,
   buildCodexAgentEnv
-} from "./chunk-KAJYMTYA.js";
-import "./chunk-2PMHC6IL.js";
+} from "./chunk-BAJGQWIO.js";
+import "./chunk-TFUEJW43.js";
 import {
   AGENT,
   headlessAgentRequestSchema
-} from "./chunk-QU4FXL4X.js";
+} from "./chunk-RZBMLP2H.js";
 import "./chunk-VEQXQN22.js";
 
 // src/headless-agent.ts

package/dist/src/index.js

@@ -32,7 +32,7 @@ import {
   reportCommandProtectionBlock,
   resolveDiffBase,
   runGitCommand
-} from "./chunk-G6QBYUZR.js";
+} from "./chunk-JBMI42EJ.js";
 import {
   ACCOUNT_RATE_LIMITS_UPDATED_METHOD,
   AGENT_MESSAGE_DELTA_METHOD,
@@ -57,7 +57,7 @@ import {
 } from "./chunk-P73LGPQO.js";
 import {
   publishedPullRequestUrls
-} from "./chunk-5EHMJ4OT.js";
+} from "./chunk-WYERDOJD.js";
 import {
   BaseRefreshManager,
   CodexAspAuthMethodChangedError,
@@ -72,7 +72,7 @@ import {
   refreshDelayMs,
   restartCodexAspHost,
   routineRefreshExclusions
-} from "./chunk-D5B72SUV.js";
+} from "./chunk-VUUSZ5JS.js";
 import {
   ENGINE_ENV,
   monolithRequest,
@@ -80,16 +80,16 @@ import {
   readRuntimeEnvFile,
   reloadEngineEnv,
   setAgentCredentialSnapshot
-} from "./chunk-TYH3IEXZ.js";
+} from "./chunk-LM4J2GLY.js";
 import {
   AspClient,
   getCodexModelProvider
-} from "./chunk-KAJYMTYA.js";
+} from "./chunk-BAJGQWIO.js";
 import {
   SUBPROCESS_MAX_BUFFER,
   execAsync,
   execFileAsync
-} from "./chunk-2PMHC6IL.js";
+} from "./chunk-TFUEJW43.js";
 import {
   isRecord as isRecord2
 } from "./chunk-UZSNFLDQ.js";
@@ -315,7 +315,7 @@ import {
   workspaceCredentialProviderSchema,
   workspaceTokenRequestSchema,
   workspaceTokenResponseSchema
-} from "./chunk-QU4FXL4X.js";
+} from "./chunk-RZBMLP2H.js";
 import {
   cloneMcpConfig,
   ensureCompatibilityImports,
@@ -14392,32 +14392,23 @@ var ClaudeManager = class _ClaudeManager extends CodingAgentManager {
     const reportedMaxTokens = Number.isFinite(usage.maxTokens) ? usage.maxTokens : null;
     const knownMaxTokens = getClaudeModelContextWindow(usage.model);
     const maxTokens = knownMaxTokens ?? reportedMaxTokens;
-    const clamped = clampTokensToWindow(usage.totalTokens, maxTokens);
+    const totalTokens = clampTokensToWindow(usage.totalTokens, maxTokens);
     const autoCompactThreshold = normalizeAutoCompactThreshold(usage.autoCompactThreshold, maxTokens);
     return {
       provider: "claude",
       source: "claude_context",
       model: usage.model,
-      totalTokens: clamped.totalTokens,
-      ...clamped.totalProcessedTokens !== void 0 ? { totalProcessedTokens: clamped.totalProcessedTokens } : {},
+      totalTokens,
       maxTokens,
       rawMaxTokens: knownMaxTokens ?? (Number.isFinite(usage.rawMaxTokens) ? usage.rawMaxTokens : maxTokens),
-      percentage: percentage(clamped.totalTokens, maxTokens),
+      percentage: percentage(totalTokens, maxTokens),
       compactsAutomatically: usage.isAutoCompactEnabled === true,
       ...autoCompactThresholdField(autoCompactThreshold),
-      categories: usage.categories.map((category) => ({
+      categories: usage.categories.filter((category) => category.kind === "used").map((category) => ({
         name: category.name,
         tokens: category.tokens,
-        percentage: percentage(category.tokens, maxTokens),
-        color: category.color,
-        ...category.isDeferred !== void 0 ? { isDeferred: category.isDeferred } : {}
+        percentage: percentage(category.tokens, maxTokens)
       })),
-      apiUsage: usage.apiUsage ? {
-        inputTokens: usage.apiUsage.input_tokens,
-        outputTokens: usage.apiUsage.output_tokens,
-        cacheCreationInputTokens: usage.apiUsage.cache_creation_input_tokens,
-        cacheReadInputTokens: usage.apiUsage.cache_read_input_tokens
-      } : null,
       updatedAt: (/* @__PURE__ */ new Date()).toISOString()
     };
   }
@@ -16404,12 +16395,7 @@ ${text.slice(-24e3)}`
       last: {
         inputTokens: tokenUsage.last.inputTokens,
         outputTokens: tokenUsage.last.outputTokens,
-        totalTokens: tokenUsage.last.totalTokens,
-        cachedInputTokens: tokenUsage.last.cachedInputTokens,
-        reasoningOutputTokens: tokenUsage.last.reasoningOutputTokens
-      },
-      total: {
-        totalTokens: tokenUsage.total.totalTokens
+        totalTokens: tokenUsage.last.totalTokens
       },
       updatedAt
     });
@@ -16514,12 +16500,6 @@ var CURSOR_CANCEL_TIMEOUT_MS = 1e4;
 var CURSOR_RECONNECT_NOTICE_DELAY_MS = 15e3;
 var TURN_ABORTED = /* @__PURE__ */ Symbol("cursor-turn-aborted");
 var CURSOR_COMPOSER_CONTEXT_WINDOW = 2e5;
-var CURSOR_CATEGORY_COLORS = {
-  input: "#3eeba3",
-  cacheRead: "#61afef",
-  cacheWrite: "#e5c07b",
-  output: "#56b6c2"
-};
 var CURSOR_TRANSPORT_ERROR_PATTERNS = [
   /connection failed/,
   /connection stalled/,
@@ -16879,31 +16859,27 @@ ${instructions}
     if (rawTotalTokens <= 0) return;
     this.onUsageUpdate(update, (/* @__PURE__ */ new Date()).toISOString());
     const maxTokens = CURSOR_COMPOSER_CONTEXT_WINDOW;
-    const { totalTokens, totalProcessedTokens } = clampTokensToWindow(rawTotalTokens, maxTokens);
+    const totalTokens = clampTokensToWindow(rawTotalTokens, maxTokens);
     const categories = [
       {
         name: "Input",
         tokens: inputTokens,
-        percentage: percentage(inputTokens, maxTokens),
-        color: CURSOR_CATEGORY_COLORS.input
+        percentage: percentage(inputTokens, maxTokens)
       },
       {
         name: "Cache read",
         tokens: cacheReadTokens,
-        percentage: percentage(cacheReadTokens, maxTokens),
-        color: CURSOR_CATEGORY_COLORS.cacheRead
+        percentage: percentage(cacheReadTokens, maxTokens)
       },
       {
         name: "Cache write",
         tokens: cacheWriteTokens,
-        percentage: percentage(cacheWriteTokens, maxTokens),
-        color: CURSOR_CATEGORY_COLORS.cacheWrite
+        percentage: percentage(cacheWriteTokens, maxTokens)
       },
       {
         name: "Output",
         tokens: outputTokens,
-        percentage: percentage(outputTokens, maxTokens),
-        color: CURSOR_CATEGORY_COLORS.output
+        percentage: percentage(outputTokens, maxTokens)
       }
     ].filter((category) => category.tokens > 0);
     const payload = {
@@ -16911,18 +16887,11 @@ ${instructions}
       source: "provider_usage",
       model: this.activeModel,
       totalTokens,
-      ...totalProcessedTokens !== void 0 ? { totalProcessedTokens } : {},
       maxTokens,
       rawMaxTokens: maxTokens,
       percentage: percentage(totalTokens, maxTokens),
       compactsAutomatically: true,
       categories,
-      apiUsage: {
-        inputTokens,
-        outputTokens,
-        cacheReadInputTokens: cacheReadTokens,
-        cacheCreationInputTokens: cacheWriteTokens
-      },
       updatedAt: (/* @__PURE__ */ new Date()).toISOString()
     };
     this.historyFile.append(this.emitContextUsage(payload));

package/dist/src/post-tool-pr-hook.js

@@ -3,13 +3,13 @@ import { createRequire as __createRequire } from 'node:module';
 const require = __createRequire(import.meta.url);
 import {
   publishedPullRequestUrls
-} from "./chunk-5EHMJ4OT.js";
+} from "./chunk-WYERDOJD.js";
 import {
   isRecord
 } from "./chunk-UZSNFLDQ.js";
 import {
   shouldRefreshPullRequests
-} from "./chunk-QU4FXL4X.js";
+} from "./chunk-RZBMLP2H.js";
 import "./chunk-VEQXQN22.js";
 
 // src/services/post-tool-pr-notifier.ts

package/dist/src/subagent-mcp.js

@@ -9,7 +9,7 @@ import {
   spawnSubagentRequestSchema,
   subagentResponseSchema,
   waitSubagentRequestSchema
-} from "./chunk-QU4FXL4X.js";
+} from "./chunk-RZBMLP2H.js";
 import "./chunk-VEQXQN22.js";
 
 // src/subagent-mcp.ts

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "replicas-engine",
-  "version": "0.1.987",
+  "version": "0.1.989",
   "description": "Lightweight API server for Replicas workspaces",
   "type": "module",
   "main": "dist/src/index.js",