[pull] main from CodebuffAI:main

agents/__tests__/editor.test.ts

@@ -67,6 +67,11 @@ describe('editor agent', () => {
       expect(glmEditor.model).toBe('z-ai/glm-5.1')
     })
 
+    test('creates kimi editor', () => {
+      const kimiEditor = createCodeEditor({ model: 'kimi' })
+      expect(kimiEditor.model).toBe('moonshotai/kimi-k2.6')
+    })
+
     test('creates minimax editor', () => {
       const minimaxEditor = createCodeEditor({ model: 'minimax' })
       expect(minimaxEditor.model).toBe('minimax/minimax-m2.7')
@@ -84,6 +89,12 @@ describe('editor agent', () => {
       expect(glmEditor.instructionsPrompt).not.toContain('</think>')
     })
 
+    test('kimi editor does not include think tags in instructions', () => {
+      const kimiEditor = createCodeEditor({ model: 'kimi' })
+      expect(kimiEditor.instructionsPrompt).not.toContain('<think>')
+      expect(kimiEditor.instructionsPrompt).not.toContain('</think>')
+    })
+
     test('minimax editor does not include think tags in instructions', () => {
       const minimaxEditor = createCodeEditor({ model: 'minimax' })
       expect(minimaxEditor.instructionsPrompt).not.toContain('<think>')

agents/base2/base2.ts

@@ -25,7 +25,7 @@ export function createBase2(
   const isFree = mode === 'free' || mode === 'lite'
 
   const isSonnet = false
-  const model = isFree ? 'z-ai/glm-5.1' : 'anthropic/claude-opus-4.7'
+  const model = isFree ? 'moonshotai/kimi-k2.6' : 'anthropic/claude-opus-4.7'
 
   return {
     publisher,

agents/editor/editor-lite.ts

@@ -3,7 +3,7 @@ import { createCodeEditor } from './editor'
 import type { AgentDefinition } from '../types/agent-definition'
 
 const definition: AgentDefinition = {
-  ...createCodeEditor({ model: 'glm' }),
+  ...createCodeEditor({ model: 'kimi' }),
   id: 'editor-lite',
 }
 export default definition

agents/editor/editor.ts

@@ -4,7 +4,7 @@ import { publisher } from '../constants'
 import type { AgentDefinition } from '../types/agent-definition'
 
 export const createCodeEditor = (options: {
-  model: 'gpt-5' | 'opus' | 'glm' | 'minimax'
+  model: 'gpt-5' | 'opus' | 'glm' | 'kimi' | 'minimax'
 }): Omit<AgentDefinition, 'id'> => {
   const { model } = options
   return {
@@ -14,6 +14,8 @@ export const createCodeEditor = (options: {
         ? 'openai/gpt-5.1'
         : options.model === 'minimax'
           ? 'minimax/minimax-m2.7'
+        : options.model === 'kimi'
+          ? 'moonshotai/kimi-k2.6'
         : options.model === 'glm'
           ? 'z-ai/glm-5.1'
           : 'anthropic/claude-opus-4.7',
@@ -67,7 +69,7 @@ OR for new files or major rewrites:
 }
 </codebuff_tool_call>
 
-${model === 'gpt-5' || model === 'glm' || model === 'minimax'
+${model === 'gpt-5' || model === 'glm' || model === 'kimi' || model === 'minimax'
         ? ''
         : `Before you start writing your implementation, you should use <think> tags to think about the best way to implement the changes.
 

agents/reviewer/code-reviewer-lite.ts

@@ -5,7 +5,7 @@ import { createReviewer } from './code-reviewer'
 const definition: SecretAgentDefinition = {
   id: 'code-reviewer-lite',
   publisher,
-  ...createReviewer('z-ai/glm-5.1'),
+  ...createReviewer('moonshotai/kimi-k2.6'),
 }
 
 export default definition

agents/types/agent-definition.ts

@@ -423,6 +423,7 @@ export type ModelName =
   // Other open source models
   | 'moonshotai/kimi-k2'
   | 'moonshotai/kimi-k2:nitro'
+  | 'moonshotai/kimi-k2.6'
   | 'z-ai/glm-5'
   | 'z-ai/glm-5.1'
   | 'z-ai/glm-4.6'

cli/src/chat.tsx

@@ -174,7 +174,7 @@ export const Chat = ({
   })
   const hasSubscription = subscriptionData?.hasSubscription ?? false
 
-  const { adData, recordImpression } = useGravityAd({
+  const { ads, recordImpression } = useGravityAd({
     enabled: IS_FREEBUFF || !hasSubscription,
     provider: 'gravity',
     fallbackProvider: 'carbon',
@@ -1463,11 +1463,8 @@ export const Chat = ({
           />
         )}
 
-        {adData && (IS_FREEBUFF || getAdsEnabled()) && (
-          <ChoiceAdBanner
-            ads={adData.variant === 'choice' ? adData.ads : [adData.ad]}
-            onImpression={recordImpression}
-          />
+        {ads && (IS_FREEBUFF || getAdsEnabled()) && (
+          <ChoiceAdBanner ads={ads} onImpression={recordImpression} />
         )}
 
         {reviewMode ? (

cli/src/components/freebuff-model-selector.tsx

@@ -5,7 +5,7 @@ import React, { useCallback, useEffect, useMemo, useState } from 'react'
 import { Button } from './button'
 import {
   FALLBACK_FREEBUFF_MODEL_ID,
-  FREEBUFF_GLM_MODEL_ID,
+  FREEBUFF_KIMI_MODEL_ID,
   FREEBUFF_MODELS,
   getFreebuffDeploymentAvailabilityLabel,
   isFreebuffModelAvailable,
@@ -25,8 +25,8 @@ import {
 import type { KeyEvent } from '@opentui/core'
 
 const FREEBUFF_MODEL_SELECTOR_MODELS = [
-  ...FREEBUFF_MODELS.filter((model) => model.id === FREEBUFF_GLM_MODEL_ID),
-  ...FREEBUFF_MODELS.filter((model) => model.id !== FREEBUFF_GLM_MODEL_ID),
+  ...FREEBUFF_MODELS.filter((model) => model.id === FREEBUFF_KIMI_MODEL_ID),
+  ...FREEBUFF_MODELS.filter((model) => model.id !== FREEBUFF_KIMI_MODEL_ID),
 ]
 
 /**
@@ -72,7 +72,7 @@ export const FreebuffModelSelector: React.FC = () => {
     // unavailable (e.g. deployment hours close while the picker is open),
     // swap to the always-available fallback so Enter doesn't POST a model
     // the server will immediately reject. In-memory only — the user's saved
-    // preference (e.g. GLM) is preserved for the next launch.
+    // preference (e.g. Kimi) is preserved for the next launch.
     if (
       (session?.status === 'none' || !session) &&
       !isFreebuffModelAvailable(selectedModel, new Date(now))

cli/src/components/waiting-room-screen.tsx

@@ -115,7 +115,7 @@ export const WaitingRoomScreen: React.FC<WaitingRoomScreenProps> = ({
   // forceStart bypasses the "wait for first user message" gate inside the hook,
   // which would otherwise block ads here since no conversation exists yet.
   // Try Gravity first, then fall back to Carbon when Gravity doesn't fill.
-  const { adData, recordImpression } = useGravityAd({
+  const { ads, recordImpression } = useGravityAd({
     enabled: true,
     forceStart: true,
     provider: 'gravity',
@@ -260,7 +260,7 @@ export const WaitingRoomScreen: React.FC<WaitingRoomScreenProps> = ({
                   <span>Elapsed </span>
                   {formatElapsed(elapsedMs)}
                 </text>
-                {/* Per-model session quota (e.g. GLM 5.1 caps at 5/12h). Only
+                {/* Per-model session quota (e.g. Kimi K2.6 caps at 5/12h). Only
                     rendered for rate-limited models so the Minimax queue stays
                     clutter-free. */}
                 {session.rateLimit && (
@@ -343,7 +343,7 @@ export const WaitingRoomScreen: React.FC<WaitingRoomScreenProps> = ({
             </>
           )}
 
-          {/* Per-model session quota exhausted (e.g. 5+ GLM sessions in the
+          {/* Per-model session quota exhausted (e.g. 5+ Kimi sessions in the
               last 12h). Terminal for this run — the user can exit and come
               back once the oldest session in the window rolls off. */}
           {session?.status === 'rate_limited' && (
@@ -369,17 +369,14 @@ export const WaitingRoomScreen: React.FC<WaitingRoomScreenProps> = ({
       </box>
 
       {/* Ad banner pinned to the bottom, same look-and-feel as in chat. */}
-      {adData && (
+      {ads && (
         <box style={{ flexShrink: 0 }}>
-          <ChoiceAdBanner
-            ads={adData.variant === 'choice' ? adData.ads : [adData.ad]}
-            onImpression={recordImpression}
-          />
+          <ChoiceAdBanner ads={ads} onImpression={recordImpression} />
         </box>
       )}
 
       {/* Horizontal separator (mirrors chat input divider style) */}
-      {!adData && (
+      {!ads && (
         <text style={{ fg: theme.muted, flexShrink: 0 }}>
           {'─'.repeat(terminalWidth)}
         </text>

cli/src/hooks/use-freebuff-session.ts

@@ -104,7 +104,7 @@ async function callSession(
       return body
     }
   }
-  // 429 from POST is the per-model session-quota reject (e.g. too many GLM
+  // 429 from POST is the per-model session-quota reject (e.g. too many Kimi
   // sessions in the last 12h). Terminal for the current poll — the CLI shows
   // a screen explaining the limit and when the user can try again. The 429
   // status (rather than 200) keeps older CLIs in their error path so they
@@ -442,9 +442,9 @@ export function useFreebuffSession(): UseFreebuffSessionResult {
         }
         if (next.status === 'model_unavailable') {
           // Server says the requested model isn't available right now (e.g.
-          // GLM outside deployment hours). Flip to the always-available
+          // Kimi outside deployment hours). Flip to the always-available
           // fallback for this run. In-memory only — `setSelectedModel`
-          // doesn't persist, so the user's saved preference (e.g. GLM)
+          // doesn't persist, so the user's saved preference (e.g. Kimi)
           // is preserved for their next launch during deployment hours.
           useFreebuffModelStore
             .getState()