promote: dev → main (Phase 1 reasoning, telemetry rebrand, qwen-logger nuke) — bumps to v0.28.0

.github/workflows/release.yml

@@ -1,10 +1,13 @@
 name: 'Release'
 
-# Fires when the dev→main promotion PR is merged.
+# Fires when a release-bearing PR is merged into main.
 # Builds the bundle, publishes @protolabsai/proto to npm, tags, and creates a GitHub Release.
 #
-# Flow: feature PRs → dev → prepare-release.yml bumps version on dev
-#       → dev→main promotion PR merges → this workflow tags and releases.
+# Two flows trigger this:
+#   1. Auto-release: auto-release.yml opens an `auto-release/v*` PR with the
+#      version bump, which merges to main and triggers this workflow.
+#   2. Manual dev→main promotion: dev branch PR'd into main (legacy path,
+#      still supported).
 
 on:
   pull_request:
@@ -24,7 +27,10 @@ jobs:
         github.event_name == 'workflow_dispatch' ||
         (
           github.event.pull_request.merged == true &&
-          github.event.pull_request.head.ref == 'dev'
+          (
+            github.event.pull_request.head.ref == 'dev' ||
+            startsWith(github.event.pull_request.head.ref, 'auto-release/')
+          )
         )
       )
     timeout-minutes: 30

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@protolabsai/proto",
-  "version": "0.27.0",
+  "version": "0.28.0",
   "publishConfig": {
     "access": "public"
   },
@@ -20,7 +20,7 @@
     "url": "https://github.com/protoLabsAI/protoCLI/issues"
   },
   "config": {
-    "sandboxImageUri": "ghcr.io/qwenlm/qwen-code:0.27.0"
+    "sandboxImageUri": "ghcr.io/qwenlm/qwen-code:0.28.0"
   },
   "scripts": {
     "start": "cross-env node scripts/start.js",

packages/cli/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@protolabs/proto",
-  "version": "0.27.0",
+  "version": "0.28.0",
   "description": "proto",
   "repository": {
     "type": "git",
@@ -37,7 +37,7 @@
     "dist"
   ],
   "config": {
-    "sandboxImageUri": "ghcr.io/qwenlm/qwen-code:0.27.0"
+    "sandboxImageUri": "ghcr.io/qwenlm/qwen-code:0.28.0"
   },
   "dependencies": {
     "@agentclientprotocol/sdk": "^0.14.1",

packages/cli/src/acp-integration/acpAgent.ts

@@ -159,7 +159,7 @@ class QwenAgent implements Agent {
     return {
       protocolVersion: PROTOCOL_VERSION,
       agentInfo: {
-        name: 'qwen-code',
+        name: 'proto-cli',
         title: 'proto',
         version,
       },

packages/cli/src/ui/components/messages/ConversationMessages.test.tsx

@@ -0,0 +1,73 @@
+/**
+ * @license
+ * Copyright 2025 Qwen Team
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+import { describe, expect, it } from 'vitest';
+import { render } from 'ink-testing-library';
+import { ThinkMessage, ThinkMessageContent } from './ConversationMessages.js';
+
+describe('ThinkMessage', () => {
+  it('renders the streaming text expanded while pending', () => {
+    const { lastFrame } = render(
+      <ThinkMessage
+        text="Let me consider this carefully and weigh the options."
+        isPending={true}
+        contentWidth={80}
+      />,
+    );
+    const output = lastFrame() ?? '';
+    expect(output).toContain('Let me consider this');
+    // Streaming render uses the existing ⟡ glyph, not the ▸ summary marker.
+    expect(output).toContain('⟡');
+    expect(output).not.toContain('thinking (');
+  });
+
+  it('renders compact "thinking (N chars)" summary once stream finalizes', () => {
+    const text = 'reasoning '.repeat(10).trim(); // 99 chars
+    const { lastFrame } = render(
+      <ThinkMessage text={text} isPending={false} contentWidth={80} />,
+    );
+    const output = lastFrame() ?? '';
+    expect(output).toContain('▸');
+    expect(output).toContain(`thinking (${text.length} chars)`);
+    // Underlying reasoning text is not rendered inline post-stream.
+    expect(output).not.toContain('reasoning reasoning');
+  });
+
+  it('formats large char counts with thousands separator', () => {
+    const text = 'x'.repeat(12_345);
+    const { lastFrame } = render(
+      <ThinkMessage text={text} isPending={false} contentWidth={80} />,
+    );
+    const output = lastFrame() ?? '';
+    expect(output).toContain('thinking (12,345 chars)');
+  });
+});
+
+describe('ThinkMessageContent', () => {
+  it('renders the continuation text while pending', () => {
+    const { lastFrame } = render(
+      <ThinkMessageContent
+        text="continued reasoning text"
+        isPending={true}
+        contentWidth={80}
+      />,
+    );
+    const output = lastFrame() ?? '';
+    expect(output).toContain('continued reasoning text');
+  });
+
+  it('renders nothing once stream finalizes (the parent ThinkMessage owns the summary)', () => {
+    const { lastFrame } = render(
+      <ThinkMessageContent
+        text="continued reasoning text"
+        isPending={false}
+        contentWidth={80}
+      />,
+    );
+    const output = lastFrame() ?? '';
+    expect(output.trim()).toBe('');
+  });
+});

packages/cli/src/ui/components/messages/ConversationMessages.tsx

@@ -227,35 +227,70 @@ export const AssistantMessageContent: React.FC<
   />
 );
 
+// Post-stream summary line ("▸ thinking (N chars)"). Phase 1 of the reasoning
+// rendering work (see #162): full text remains live in Langfuse, ACP
+// `agent_thought_chunk` notifications, and ChatRecord for back-compat. An
+// in-TUI expand affordance is a follow-up. Note: when a long thought was
+// split mid-stream into a gemini_thought + gemini_thought_content pair, this
+// counts only the first chunk — the continuation renders nothing (see below).
+// True total requires post-finalize coalescing in useGeminiStream and is
+// deferred since splits are rare and the count is a hint, not a contract.
+const ThinkSummary: React.FC<{ text: string }> = ({ text }) => {
+  const charCount = text.length;
+  return (
+    <PrefixedTextMessage
+      text={`thinking (${charCount.toLocaleString()} chars)`}
+      prefix="▸"
+      prefixColor={theme.text.secondary}
+      textColor={theme.text.secondary}
+    />
+  );
+};
+
 export const ThinkMessage: React.FC<ThinkMessageProps> = ({
   text,
   isPending,
   availableTerminalHeight,
   contentWidth,
-}) => (
-  <PrefixedMarkdownMessage
-    text={text}
-    prefix="⟡"
-    prefixColor={theme.text.secondary}
-    isPending={isPending}
-    availableTerminalHeight={availableTerminalHeight}
-    contentWidth={contentWidth}
-    textColor={theme.text.secondary}
-  />
-);
+}) => {
+  if (!isPending) {
+    return <ThinkSummary text={text} />;
+  }
+  return (
+    <PrefixedMarkdownMessage
+      text={text}
+      prefix="⟡"
+      prefixColor={theme.text.secondary}
+      isPending={isPending}
+      availableTerminalHeight={availableTerminalHeight}
+      contentWidth={contentWidth}
+      textColor={theme.text.secondary}
+    />
+  );
+};
 
 export const ThinkMessageContent: React.FC<ThinkMessageContentProps> = ({
   text,
   isPending,
   availableTerminalHeight,
   contentWidth,
-}) => (
-  <ContinuationMarkdownMessage
-    text={text}
-    isPending={isPending}
-    availableTerminalHeight={availableTerminalHeight}
-    contentWidth={contentWidth}
-    basePrefix="⟡"
-    textColor={theme.text.secondary}
-  />
-);
+}) => {
+  // When the stream has finalized, suppress the continuation block. The
+  // adjacent ThinkMessage already renders the summary line; rendering this
+  // continuation as another summary would double-count and drop chars across
+  // the split boundary. Streaming-time renders unchanged so live thoughts
+  // still appear.
+  if (!isPending) {
+    return null;
+  }
+  return (
+    <ContinuationMarkdownMessage
+      text={text}
+      isPending={isPending}
+      availableTerminalHeight={availableTerminalHeight}
+      contentWidth={contentWidth}
+      basePrefix="⟡"
+      textColor={theme.text.secondary}
+    />
+  );
+};

packages/core/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@qwen-code/qwen-code-core",
-  "version": "0.27.0",
+  "version": "0.28.0",
   "description": "proto core",
   "repository": {
     "type": "git",

packages/core/src/config/config.test.ts

@@ -13,7 +13,6 @@ import { setGeminiMdFilename as mockSetGeminiMdFilename } from '../tools/memoryT
 import {
   DEFAULT_TELEMETRY_TARGET,
   DEFAULT_OTLP_ENDPOINT,
-  QwenLogger,
 } from '../telemetry/index.js';
 import type {
   ContentGenerator,
@@ -235,9 +234,6 @@ describe('Server Config (config.ts)', () => {
   beforeEach(() => {
     // Reset mocks if necessary
     vi.clearAllMocks();
-    vi.spyOn(QwenLogger.prototype, 'logStartSessionEvent').mockImplementation(
-      async () => undefined,
-    );
 
     // Setup default mock for resolveContentGeneratorConfigWithSources
     vi.mocked(resolveContentGeneratorConfigWithSources).mockImplementation(
@@ -635,16 +631,6 @@ describe('Server Config (config.ts)', () => {
         expect(config.getUsageStatisticsEnabled()).toBe(enabled);
       },
     );
-
-    it('logs the session start event', async () => {
-      const config = new Config({
-        ...baseParams,
-        usageStatisticsEnabled: true,
-      });
-      await config.initialize();
-
-      expect(QwenLogger.prototype.logStartSessionEvent).toHaveBeenCalledOnce();
-    });
   });
 
   describe('Telemetry Settings', () => {

packages/core/src/core/client.test.ts

@@ -146,7 +146,6 @@ vi.mock('../telemetry/index.js', async (importOriginal) => {
     ...actual,
     uiTelemetryService: mockUiTelemetryService,
     // We keep the real implementations of logChatCompression, etc.
-    // but we can spy on QwenLogger if needed
   };
 });
 vi.mock('../ide/ideContext.js');