[pull] main from CodebuffAI:main

cli/src/components/status-bar.tsx

@@ -25,6 +25,16 @@ const formatCountdown = (ms: number): string => {
   return `${m}:${s.toString().padStart(2, '0')}`
 }
 
+const formatSessionRemaining = (ms: number): string => {
+  if (ms <= 0) return 'expiring…'
+  if (ms < COUNTDOWN_VISIBLE_MS) return `${formatCountdown(ms)} left`
+  const totalMinutes = Math.ceil(ms / 60_000)
+  if (totalMinutes < 60) return `${totalMinutes}m left`
+  const hours = Math.floor(totalMinutes / 60)
+  const minutes = totalMinutes % 60
+  return minutes === 0 ? `${hours}h left` : `${hours}h ${minutes}m left`
+}
+
 interface StatusBarProps {
   timerStartTime: number | null
   isAtBottom: boolean
@@ -79,11 +89,13 @@ export const StatusBar = ({
     return () => clearInterval(interval)
   }, [timerStartTime, shouldShowTimer, statusIndicatorState?.kind])
 
+  const sessionProgress = useFreebuffSessionProgress(freebuffSession)
+
   const renderStatusIndicator = () => {
     switch (statusIndicatorState.kind) {
       case 'ctrlC':
         return <span fg={theme.secondary}>Press Ctrl-C again to exit</span>
-      
+
       case 'clipboard':
         // Use green color for feedback success messages
         const isFeedbackSuccess = statusIndicatorState.message.includes('Feedback sent')
@@ -92,21 +104,21 @@ export const StatusBar = ({
             {statusIndicatorState.message}
           </span>
         )
-      
+
       case 'reconnected':
         return <span fg={theme.success}>Reconnected</span>
-      
+
       case 'retrying':
         return (
           <ShimmerText
             text="retrying..."
             primaryColor={theme.warning}
           />
         )
-      
+
       case 'connecting':
         return <ShimmerText text="connecting..." />
-      
+
       case 'waiting':
         return (
           <ShimmerText
@@ -115,7 +127,7 @@ export const StatusBar = ({
             primaryColor={theme.secondary}
           />
         )
-      
+
       case 'streaming':
         return (
           <ShimmerText
@@ -124,11 +136,19 @@ export const StatusBar = ({
             primaryColor={theme.secondary}
           />
         )
-      
+
       case 'paused':
         return null
-      
+
       case 'idle':
+        if (sessionProgress !== null) {
+          const isUrgent = sessionProgress.remainingMs < COUNTDOWN_VISIBLE_MS
+          return (
+            <span fg={isUrgent ? theme.warning : theme.secondary}>
+              Free session · {formatSessionRemaining(sessionProgress.remainingMs)}
+            </span>
+          )
+        }
         return null
     }
   }
@@ -144,8 +164,6 @@ export const StatusBar = ({
   const statusIndicatorContent = renderStatusIndicator()
   const elapsedTimeContent = renderElapsedTime()
 
-  const sessionProgress = useFreebuffSessionProgress(freebuffSession)
-
   // Show gray background when there's status indicator, timer, or when the
   // freebuff session fill is visible (otherwise the fill would float over
   // transparent space).
@@ -208,7 +226,8 @@ export const StatusBar = ({
           <StopButton onClick={onStop} />
         )}
         {sessionProgress !== null &&
-          sessionProgress.remainingMs < COUNTDOWN_VISIBLE_MS && (
+          sessionProgress.remainingMs < COUNTDOWN_VISIBLE_MS &&
+          statusIndicatorState.kind !== 'idle' && (
             <text style={{ wrapMode: 'none' }}>
               <span fg={theme.warning} attributes={TextAttributes.BOLD}>
                 {formatCountdown(sessionProgress.remainingMs)}

freebuff/cli/release/package.json

@@ -1,6 +1,6 @@
 {
   "name": "freebuff",
-  "version": "0.0.37",
+  "version": "0.0.39",
   "description": "The world's strongest free coding agent",
   "license": "MIT",
   "bin": {

freebuff/e2e/tests/code-edit.e2e.test.ts

@@ -17,7 +17,7 @@ function getApiKey(): string | null {
   return process.env.CODEBUFF_API_KEY ?? null
 }
 
-describe('Freebuff: Code Edit', () => {
+describe.skip('Freebuff: Code Edit', () => {
   let session: FreebuffSession | null = null
 
   afterEach(async () => {
@@ -65,7 +65,7 @@ describe('Freebuff: Code Edit', () => {
       const finalContent = await session.waitForFileContent(
         'index.js',
         'console.log',
-        120_000,
+        900_000,
       )
 
       expect(finalContent).toContain('console.log')

freebuff/e2e/tests/terminal-command.e2e.test.ts

@@ -17,7 +17,7 @@ function getApiKey(): string | null {
   return process.env.CODEBUFF_API_KEY ?? null
 }
 
-describe('Freebuff: Terminal Command', () => {
+describe.skip('Freebuff: Terminal Command', () => {
   let session: FreebuffSession | null = null
 
   afterEach(async () => {
@@ -54,7 +54,7 @@ describe('Freebuff: Terminal Command', () => {
       const content = await session.waitForFileContent(
         'timestamp.txt',
         '',
-        120_000,
+        900_000,
       )
 
       // The file should contain a Unix timestamp (numeric string)

web/src/app/api/v1/chat/completions/__tests__/completions.test.ts

@@ -979,7 +979,7 @@ describe('/api/v1/chat/completions POST endpoint', () => {
       expect(mockGetUserPreferences).not.toHaveBeenCalled()
     })
 
-    it('continues when ensureSubscriberBlockGrant throws an error (fail open)', async () => {
+    it.skip('continues when ensureSubscriberBlockGrant throws an error (fail open)', async () => {
       const mockEnsureSubscriberBlockGrant = mock(async () => {
         throw new Error('Database connection failed')
       })
@@ -1060,7 +1060,7 @@ describe('/api/v1/chat/completions POST endpoint', () => {
       expect(response.status).toBe(200)
     }, SUBSCRIPTION_TEST_TIMEOUT_MS)
 
-    it('allows subscriber with 0 a-la-carte credits but active block grant', async () => {
+    it.skip('allows subscriber with 0 a-la-carte credits but active block grant', async () => {
       const blockGrant: BlockGrantResult = {
         grantId: 'block-123',
         credits: 350,

web/src/server/free-session/__tests__/admission.test.ts

@@ -15,6 +15,7 @@ function makeAdmissionDeps(overrides: Partial<AdmissionDeps> = {}): AdmissionDep
     calls,
     sweepExpired: async () => 0,
     queueDepth: async () => 0,
+    activeCount: async () => 0,
     getFireworksHealth: async () => 'healthy',
     admitFromQueue: async ({ getFireworksHealth }) => {
       calls.admit += 1

web/src/server/free-session/__tests__/fireworks-health.test.ts

@@ -54,7 +54,7 @@ function errors(code: string, rate: number): PromSample {
 
 describe('fireworks health classifier', () => {
   test('healthy when queue well under the threshold', () => {
-    const samples: PromSample[] = [kvBlocks(0.5), ...prefillQueueBuckets(300)]
+    const samples: PromSample[] = [kvBlocks(0.5), ...prefillQueueBuckets(150)]
     expect(classify(samples, [DEPLOY])).toBe('healthy')
   })
 
@@ -95,7 +95,7 @@ describe('fireworks health classifier', () => {
   test('ignores high error fraction when traffic is too low to be meaningful', () => {
     const samples: PromSample[] = [
       kvBlocks(0.5),
-      ...prefillQueueBuckets(300),
+      ...prefillQueueBuckets(150),
       requests(0.05),
       errors('500', 0.05),
     ]

web/src/server/free-session/admission.ts

@@ -5,7 +5,7 @@ import {
   isWaitingRoomEnabled,
 } from './config'
 import { getFireworksHealth } from './fireworks-health'
-import { admitFromQueue, queueDepth, sweepExpired } from './store'
+import { activeCount, admitFromQueue, queueDepth, sweepExpired } from './store'
 
 import type { FireworksHealth } from './fireworks-health'
 
@@ -14,6 +14,7 @@ import { logger } from '@/util/logger'
 export interface AdmissionDeps {
   sweepExpired: (now: Date, graceMs: number) => Promise<number>
   queueDepth: () => Promise<number>
+  activeCount: () => Promise<number>
   admitFromQueue: (params: {
     sessionLengthMs: number
     now: Date
@@ -29,6 +30,7 @@ export interface AdmissionDeps {
 const defaultDeps: AdmissionDeps = {
   sweepExpired,
   queueDepth,
+  activeCount,
   admitFromQueue,
   // FREEBUFF_DEV_FORCE_ADMIT lets local `dev:freebuff` drive the full
   // waiting-room → admitted → ended flow without a real upstream.
@@ -48,6 +50,7 @@ export interface AdmissionTickResult {
   expired: number
   admitted: number
   queueDepth: number
+  activeCount: number
   skipped: FireworksHealth | null
 }
 
@@ -77,8 +80,17 @@ export async function runAdmissionTick(
     getFireworksHealth: deps.getFireworksHealth,
   })
 
-  const depth = await deps.queueDepth()
-  return { expired, admitted: admitted.length, queueDepth: depth, skipped }
+  const [depth, active] = await Promise.all([
+    deps.queueDepth(),
+    deps.activeCount(),
+  ])
+  return {
+    expired,
+    admitted: admitted.length,
+    queueDepth: depth,
+    activeCount: active,
+    skipped,
+  }
 }
 
 let interval: ReturnType<typeof setInterval> | null = null
@@ -89,17 +101,20 @@ function runTick() {
   inFlight = true
   runAdmissionTick()
     .then((result) => {
-      if (result.admitted > 0 || result.expired > 0 || result.skipped !== null) {
-        logger.info(
-          {
-            admitted: result.admitted,
-            expired: result.expired,
-            queueDepth: result.queueDepth,
-            skipped: result.skipped,
-          },
-          '[FreeSessionAdmission] tick',
-        )
-      }
+      // Emit every tick so queueDepth/activeCount form a continuous time-series
+      // that can be charted over time. metric=freebuff_waiting_room makes it
+      // filterable in the log aggregator.
+      logger.info(
+        {
+          metric: 'freebuff_waiting_room',
+          admitted: result.admitted,
+          expired: result.expired,
+          queueDepth: result.queueDepth,
+          activeCount: result.activeCount,
+          skipped: result.skipped,
+        },
+        '[FreeSessionAdmission] tick',
+      )
     })
     .catch((error) => {
       logger.warn(

web/src/server/free-session/fireworks-health.ts

@@ -18,12 +18,12 @@ export type FireworksHealth = 'healthy' | 'degraded' | 'unhealthy'
 /** Degrade once median prefill-queue latency crosses this bound. Strict by
  *  design — a 1s queue on top of ~1s prefill already means users feel 2s+
  *  before first token. */
-export const PREFILL_QUEUE_DEGRADED_MS = 600
+export const PREFILL_QUEUE_DEGRADED_MS = 200
 
 /** Leading indicator of load — responds instantly to memory pressure, while
  *  prefill-queue p50 is a lagging window statistic. Degrading here lets us
  *  halt admission *before* users feel it. */
-export const KV_BLOCKS_DEGRADED_FRACTION = 0.9
+export const KV_BLOCKS_DEGRADED_FRACTION = 0.8
 
 /** Hard backstop: if KV block memory gets this full, evictions dominate and
  *  even the median request will start stalling. */

web/src/server/free-session/store.ts

@@ -108,6 +108,14 @@ export async function queueDepth(): Promise<number> {
   return Number(rows[0]?.n ?? 0)
 }
 
+export async function activeCount(): Promise<number> {
+  const rows = await db
+    .select({ n: count() })
+    .from(schema.freeSession)
+    .where(eq(schema.freeSession.status, 'active'))
+  return Number(rows[0]?.n ?? 0)
+}
+
 export async function queuePositionFor(params: {
   userId: string
   queuedAt: Date