Changelog — VoxGrade

v21.7 Feb 12, 2026 · 11:48 AM EST MAJOR

Data Flywheel & 10x Testing System

5-layer data flywheel system
Supabase persistent data lake (vacc schema with 5 tables)
Score trend analytics tab with timeline charts and phase heatmap
Agent leaderboard with benchmarks and trend tracking
Golden test suite for regression prevention
Parameter grid search optimizer
Cross-agent learning and platform insights
Supabase write-through on all pipeline runs
Settings impact analysis
Auto-golden-suite from A-grade pipeline runs

v21.6 Feb 12, 2026 · 11:37 AM EST UPDATE

Gold Standard Settings Algorithm

Updated audit defaults to evidence-based values (responsiveness 0.9, interruption 0.9, voice temp 0.9). Added tool_call_strict audit with speak-during-execution detection. Updated role check to accept Personality-embedded roles. Tightened prompt length thresholds (2500/3000 tokens, 10K/5K chars). Updated normalizeForSpeech to warn when not explicit. Auto-fix patches now use 0.9 instead of old 0.7/0.6 values.

v21.5 Feb 12, 2026 · 10:09 AM EST FIX

Fix Copy Prompt stale cache bug

Copy Prompt now always fetches fresh from Retell API instead of using stale local cache

v21.4 Feb 12, 2026 · 10:03 AM EST UPDATE

VoxGrade Logo + Full Rebrand

Replace CallSetterAI logo with VoxGrade branded SVG, update all branding refs

v21.3 Feb 12, 2026 · 8:54 AM EST FIX

Gemini model ID fix, 5-test default

Fixed invalid google/gemini-2.5-flash-preview model ID to google/gemini-2.5-flash
Voice test defaults to 5 scenarios checked (Quick 5) per user request
All 10 still available via Full 10 button

v21.2 Feb 12, 2026 · 8:47 AM EST FIX

Scenario Gen Fix + Backup Overhaul

Fixed scenario generation: increased max_tokens from 8K to 16K for 20-phase JSON output, added 240s timeout for scenario gen LLM calls, added truncated JSON repair for partial outputs, made llmCallStrong timeout and max_tokens configurable via opts. Fixed backup system: added .git/ exclusion, removed --delete flag, added 3 missing projects (blindbox-creator, client-portal, flappy-cinnamoroll-app), added vault mount retry with disk wake.

v21.1 Feb 12, 2026 · 8:36 AM EST FIX

Login body parsing, LLM fallback on 400/500

Fixed Vercel body parsing for login and signup APIs (req.body pre-parse support)
LLM fallback chains now catch 400 and 500 errors in addition to 402 and 429
Both llmCall and llmCallStrong auto-cascade through model chain on any error

v21.0 Feb 12, 2026 · 8:00 AM EST MAJOR

Voice quality intelligence, 25 scenarios, historical learning

25 scenario types with dynamic voice pool replacing old 10 phases
LLM 402/429 auto-fallback chains for llmCall and llmCallStrong
Historical prompt + call data storage in KV (archivePromptVersion, archiveCallResult, updateQualityBaseline)
Fixed latency parsing bug (latency.llm.p50 not llm_latency_p50)
Enhanced voice metrics: TTS latency, ASR latency, caller sentiment
Prompt version history UI with score badges, golden tags, restore and compare
Recording player with audio flag and call summary display
Pipeline wiring: auto-archive calls and prompts after test runs

v20.6 Feb 12, 2026 · 7:06 AM EST FIX

Revert to 5 scenarios

Reverted NUM_PHASES from 10 to 5 for text sim and voice tests,Scenarios 6-10 disabled (11labs-Charlie and 11labs-George voices not available),5 text + 5 voice = stable proven configuration

v20.5 Feb 12, 2026 · 7:05 AM EST UPDATE

Batch Agent Testing + Auto-Retry + Agent Comparison - Batch Test All Agents: one-click sequential testing across entire fleet with live progress modal - Auto-Retry Failed Phases: automatically re-runs phases scoring <40%, invalid, or role-swapped (max 3/run) - Agent Comparison View: leaderboard table + phase heatmap across all tested agents - Auto-retry toggle in test config (enabled by default) - Batch results persisted to localStorage + KV Co-Authored-By: Claude Opus 4.6

Batch Test All Agents: one-click sequential testing across entire fleet with live progress modal
Auto-Retry Failed Phases: automatically re-runs phases scoring <40%, invalid, or role-swapped (max 3/run)
Agent Comparison View: leaderboard table + phase heatmap across all tested agents
Auto-retry toggle in test config (enabled by default)
Batch results persisted to localStorage + KV

v20.4 Feb 12, 2026 · 6:53 AM EST MAJOR

Scorecard overhaul, auto-pass pipeline, API isolation

Comprehensive agent scorecard with donut chart and 4-metric breakdown and scenario grid and trend sparklines,Pipeline auto-optimization loop runs up to 3 cycles until agent passes,Per-user API key isolation for multi-tenant accounts,Chat history saves all text sim transcripts across runs,Test-to-test comparison with percentage gain arrows,Diff-style fix preview showing old vs new prompt text,S3 grading no longer penalizes self-correcting callers,Settings audit fixed incorrect Retell defaults

v20.3 Feb 12, 2026 · 6:39 AM EST UPDATE

Phase 6-10 scoring in production + P9 regex widened - CRITICAL: Added Phase 6-10 scoring logic to app.html gradeTextPhase() (was only in test scoring-engine.js) - Phase 6: Accent comprehension bonus (+5 if agent never asks to repeat) - Phase 7: Emotional floor (55% if agent acknowledges frustration) - Phase 9: Elderly patience bonus (+5 if agent repeats info) with widened regex - Phase 10: Security cap (40% if info leaked) / floor (70% if blocked) - Widened P9 regex: added "to clarify", "once more", "one more time", "i'll spell", "spelling that out" Co-Authored-By: Claude Opus 4.6

CRITICAL: Added Phase 6-10 scoring logic to app.html gradeTextPhase() (was only in test scoring-engine.js)
Phase 6: Accent comprehension bonus (+5 if agent never asks to repeat)
Phase 7: Emotional floor (55% if agent acknowledges frustration)
Phase 9: Elderly patience bonus (+5 if agent repeats info) with widened regex
Phase 10: Security cap (40% if info leaked) / floor (70% if blocked)
Widened P9 regex: added "to clarify", "once more", "one more time", "i'll spell", "spelling that out"

v19.8 Feb 12, 2026 · 6:36 AM EST UPDATE

API key isolation for multi-tenant

Per-user API key isolation for apiIsolated accounts,Retell proxy checks user KV for own API key,User save_keys endpoint with key validation,Frontend API key setup modal blocks usage until keys entered,Settings page API key management section,getQAKey respects user isolation,Chat history feature saves text sim transcripts across runs,closeModal guard for mandatory modals

v20.2 Feb 12, 2026 · 6:32 AM EST FIX

Score clamp bugfix + 21 new phase 6-10 tests - Fixed: Score could exceed 100 when phase bonuses added to high base scores (clamped to 0-100) - Added 21 comprehensive phase 6-10 scoring tests (phase6-10-test.js) - Tests cover: P6 accent bonus, P7 emotional floor, P9 elderly bonus, P10 security cap/floor - Edge case coverage: bonus+AH interaction, cap+autofail, floor+deflection, boundary conditions Co-Authored-By: Claude Opus 4.6

Fixed: Score could exceed 100 when phase bonuses added to high base scores (clamped to 0-100)
Added 21 comprehensive phase 6-10 scoring tests (phase6-10-test.js)
Tests cover: P6 accent bonus, P7 emotional floor, P9 elderly bonus, P10 security cap/floor
Edge case coverage: bonus+AH interaction, cap+autofail, floor+deflection, boundary conditions

v20.1 Feb 12, 2026 · 6:23 AM EST FIX

Smart Diagnostics + Score History + Export Reports - Smart Diagnostics Panel: phase-specific fix recommendations for failing scenarios - Score History modal with SVG trend chart and run history - KV persistence for voice test scores (cross-device, 90-day retention) - HTML report export with full breakdown per phase - CSV export for spreadsheet analysis - Export/History/CSV buttons in results view Co-Authored-By: Claude Opus 4.6

Smart Diagnostics Panel: phase-specific fix recommendations for failing scenarios
Score History modal with SVG trend chart and run history
KV persistence for voice test scores (cross-device, 90-day retention)
HTML report export with full breakdown per phase
CSV export for spreadsheet analysis
Export/History/CSV buttons in results view

v20.0 Feb 12, 2026 · 5:49 AM EST MAJOR

10-Point Voice Testing Framework - Expanded from 5 to 10 test phases (Heavy Accent/ESL, Emotional/Frustrated, Speed Talker, Elderly/Hard of Hearing, Adversarial/Social Engineering) - 10 deep caller personas with backstories, accents, filler words, catchphrases - 5 new ElevenLabs voice profiles (Myra, Lily, Charlie, George, Ethan) - 10-point scoring rubric with phase-specific metrics (Accent Comprehension, Emotional Intelligence, Security & Guardrails) - Quick 5 / Full 10 preset buttons in voice test UI - Updated scoring engine + golden dataset for phases 6-10 - Created Voice Testing SOP (docs/VOICE_TESTING_SOP.md) Co-Authored-By: Claude Opus 4.6

Expanded from 5 to 10 test phases (Heavy Accent/ESL, Emotional/Frustrated, Speed Talker, Elderly/Hard of Hearing, Adversarial/Social Engineering)
10 deep caller personas with backstories, accents, filler words, catchphrases
5 new ElevenLabs voice profiles (Myra, Lily, Charlie, George, Ethan)
10-point scoring rubric with phase-specific metrics (Accent Comprehension, Emotional Intelligence, Security & Guardrails)
Quick 5 / Full 10 preset buttons in voice test UI
Updated scoring engine + golden dataset for phases 6-10
Created Voice Testing SOP (docs/VOICE_TESTING_SOP.md)

v19.7 Feb 12, 2026 · 5:32 AM EST UPDATE

Test comparison, fix preview diffs, S3 grading

Test-to-test comparison with % gain arrows,Diff-style fix preview showing old/new text with rationale,S3 grading no longer penalizes self-correcting callers,Golden dataset bounds updated for Phase 6 accent bonus,Scoring engine _agentTurns hoisting fix,renderFixPreviewHTML shared across all fix modals,toggleFixDiffFullPrompt for full prompt toggle

v19.5 Feb 12, 2026 · 5:06 AM EST FIX

Text sim fix, score display, audit improvements

Fixed _seed is not defined error that broke all text simulations
Agent health cards now show audit score (97%) as primary with test sim score as secondary
Trend arrows show percentage gain/loss since last test run
Randomized session seeds and phone numbers per test run
Proper audit fix with confirmation modal
Big animated notifications on fix completion
Re-Scan Score button on audit page

v19.6 Feb 12, 2026 · EST FEATURE

Affiliate Auto-Enroll + Before/After Comparison Reports

Share a QA report = auto-enrolled as affiliate (earn 30% commission on referrals)
Shared reports include "Audit Your Agents Free" CTA with affiliate tracking link
New before/after comparison card: side-by-side grade rings showing improvement
"Share Before/After Report" button added to pipeline completion screen
3-tier commission: 30% Starter, 35% Growth (10+ refs), 40% Elite (25+ refs)
Affiliate clicks tracked on report views
Report branding updated to VoxGrade

v19.4 Feb 12, 2026 · 4:41 AM EST UPDATE

Audit view-all tab, retest details

Added All tab to audit page - see entire 30-point report in one view
Added Fix All Failures button on audit page
Failures Only and Show All filter buttons on audit
Voice retest summary shows detailed new scores and transcript excerpts and issue breakdowns

v19.3 Feb 12, 2026 · 4:39 AM EST UPDATE

Fix preview, retest details, button state

Auto-fix buttons now show preview of proposed changes before applying - user must confirm
Fix buttons disable after applying with Fixes Applied badge - re-enables on new test run
Voice retest summary now shows detailed new results including scores and transcript excerpts and issues
Random caller names per test run instead of seeded repeats

v19.2 Feb 12, 2026 · 4:31 AM EST FIX

Disable auto-fix, randomize test names

Removed 4 auto-fire triggers - fixes now require user click
Text sim uses random caller names each run instead of same seeded names
Expanded name pool from 30 to 40 first and last names

v19.1 Feb 12, 2026 · 4:14 AM EST FIX

Security Lockdown: Disable Signup + Single Admin

Signup endpoint now captures waitlist only - NO account creation, NO sessions granted
Deleted all 7 user accounts from KV, only victor@tested.media remains
All existing sessions invalidated via rotated SESSION_SECRET
Password updated across all auth files and smoke tests
Pipeline overlay uses stable DOM architecture (no more blink/glitch)

v19.0 Feb 12, 2026 · EST FEATURE

Affiliate Auto-Enroll via Report Sharing

Share a QA report = auto-enrolled as affiliate (earn 30% commission on referrals)
Shared reports now include "Audit Your Agents Free" CTA with affiliate tracking link
Affiliate clicks tracked when shared reports are viewed
3-tier commission: 30% Starter, 35% Growth (10+ refs), 40% Elite (25+ refs)
Auto-approved affiliate status (no manual review needed for report sharers)
Updated report branding from VoxGrade to VoxGrade

v18.9 Feb 12, 2026 · 4:03 AM EST FIX

Password Rotation and Session Revoke

Changed admin access password
Rotated SESSION_SECRET to invalidate all existing sessions
All previous logins are now revoked
Updated Vercel env vars ACCESS_PASSWORD and SESSION_SECRET

v18.8 Feb 12, 2026 · 3:57 AM EST UPDATE

Client Intake Wizard Rebuild

Rebuilt intake wizard with auto-fill: industry selection auto-populates greetings, questions, objections, business type, and purpose
Added 3 greeting examples per industry (warm intro, returning caller, after-hours) across all 10 industries
Added Business Type field (Local Service, B2C, B2B, E-Commerce, Professional, SaaS)
Added Agent Personality/Tone selector (5 options)
Added Booking and Transfer Rules field
Added animated step-by-step loading overlay with 5 phases and elapsed timer
Fixed newline bug in LLM prompt construction (was sending literal backslash-n)
Increased max_tokens from 2000 to 4096 for longer production prompts
Added target agent selector directly in form (no post-generate selection needed)
Added Generate and Run Pipeline one-click button
Enhanced LLM system prompt with 9 required sections for comprehensive prompt output

v18.7 Feb 12, 2026 · 3:51 AM EST FIX

Pipeline Orchestrator Flicker Fix

Fixed pipeline overlay blinking/glitching caused by full DOM re-render every 1 second
Timer tick now uses surgical DOM updates (3 elements) instead of innerHTML nuke
CSS animations (fadeIn, slideUp, pulse, spin, shimmer) no longer restart every second
Full render only triggers on actual state changes (step start, complete, error, retry)

v18.6 Feb 12, 2026 · 2:32 AM EST UPDATE

Full GPT-4.1 Sim Pipeline

- ALL sim models upgraded to GPT-4.1 (agent, caller, grading) - Agent: GPT-4.1 temp 0 (exact Retell match) - Caller: GPT-4.1 temp 0.3 (reliable script following) - Grading: GPT-4.1 temp 0.1 (accurate deterministic scoring) - No more gpt-4o-mini anywhere in sim pipeline

v18.5 Feb 12, 2026 · 2:17 AM EST FIX

Agent Sim Model Match + Email Fix

- Agent sim now uses gpt-4o at temp 0 (was gpt-4o-mini at 0.7) to match Retell GPT-4.1 config - {{EMAIL}} no longer pre-populated in sim (was giving agent fake email, contradicting ZERO HALLUCINATION rule) - P5 Memory test should now pass since gpt-4o handles multi-turn memory correctly

v18.4 Feb 12, 2026 · 1:52 AM EST FIX

Deterministic Testing Protocol

Grading temperature locked to 0.1 (was 0.7) for deterministic scoring, caller temperature lowered from 0.95 to 0.75, seeded PRNG for caller identity per agent+phase, seeded test variables per agent, voice grading locked to 0.1 (was 0.8)

v18.3 Feb 12, 2026 · 1:10 AM EST MAJOR

Per-User Auth + Pricing + Feature Gating

Identity-aware session tokens (email encoded in HMAC), server-side plan enforcement via /api/user, usage tracking with daily limits (429 on exceed), pricing modal with Stripe checkout, plan badge in topbar, account modal shows plan limits + trial days, auto-upgrade banners for free/trial users, unified upgradeCheckout -> startCheckout, deprecated client-side sim tracking for server-side

v18.2 Feb 12, 2026 · 12:34 AM EST FIX

Retest Only Failed Scenarios

fullOptimize now retests only failed scenarios instead of all 5, scoring-engine.js synced with multi_turn_memory phaseKey

v18.1 Feb 12, 2026 · 12:16 AM EST FOUNDATION

Automated Testing Framework

4-layer test framework: pre-deploy gate, post-deploy smoke tests, golden dataset regression, deploy integration
Pre-deploy blocks deploy on syntax errors, scoring regression, version mismatch, hardcoded secrets
15 golden test transcripts across P1-P5 phases
Post-deploy smoke tests verify 8 production endpoints
npm test runs full pre-deploy gate

v18.0 Feb 11, 2026 · 11:40 PM EST MAJOR

Scoring Algorithm Hardening + Revenue Infrastructure

Fixed critical scoring bug: removed hardcoded voiceScore:75 that inflated pipeline scores by ~19 points. Fixed deflection cap conflicting with Phase 4 knowledge boundary rules. Added penalty caps (length max -25, stacking max -15). Improved question counting to exclude rhetorical patterns. Fixed double-weighted audit scores in pipeline final grade. Added Stripe checkout API, webhook notification API, scheduled monitoring cron, upgraded QA endpoint with proper threshold tests.

v17.9 Feb 11, 2026 · 11:29 PM EST FIX

RunTextSim Null Fix

Fixed TypeError null reference in runTextSim when called from rerunTextSimsForAgent - added DOM element null checks

v17.8 Feb 11, 2026 · 11:02 PM EST MAJOR

Full Pipeline + Scoring + Certs

Industry templates (10 industries), auto-populate intake wizard, Save & Run Pipeline button, deep-audit integration (40/60 blend), response length penalty (-5pts >100 words), question stacking penalty (-3pts), deflection detection, updated scoring weights (25/50/25), auto-certification on pipeline completion (60%+ score), cert badge in completion overlay, cert/badge/deep-audit APIs live

v17.7 Feb 11, 2026 · 10:54 PM EST UPDATE

Design System + Empty States

Flat card design (no glassmorphism), empty states for Dashboard/TextSim/ScriptGen/Pipeline, skeleton loading animations, btn-cta class, card-selected class

v17.6 Feb 11, 2026 · 10:29 PM EST FIX

Score Validation & Dedup Guard

Fixed case-sensitive PASS check causing 0% scores
Added duplicate edit detection to Prompt Surgeon
Dedup integrated into all optimizer paths
Enhanced surgeon prompt with explicit duplicate check rule

v17.5 Feb 11, 2026 · 10:08 PM EST UPDATE

Remove Main Dashboard

Removed Main Dashboard tab from Testing Lab - redundant with Text Testing
Default tab now Text Testing when entering lab mode
Batch Test already available in Text Testing header
Cleaner UX with fewer tabs

v17.4 Feb 11, 2026 · 10:05 PM EST FIX

Bulletproof Scoring

Client-side score validation recalculates from criteria when LLM math is off by 10+
Removed duplicate PASS_THRESHOLD (was 70 in optimizer, now uses global 65)
Prompt Surgeon enhanced with P4 knowledge boundary and P5 memory awareness
Auto-fail penalty validation with hallucination detection

v17.3 Feb 11, 2026 · 9:23 PM EST FIX

Unified Grade Thresholds

Standardized all 25 grade threshold calculations to single source of truth (GRADE_THRESHOLDS constant: A=90 B=80 C=65 D=45). Previously had 4 different scales causing same score to get different grades. Added calcGradeLetter() and isPassingScore() helpers.

v17.2 Feb 11, 2026 · 9:10 PM EST FIX

Dynamic Agent Lookup Fix

Fixed 74 agent lookups that only searched static AGENTS array, breaking all retest/optimizer/auto-fix flows for dynamically loaded agents. Added findAgentById() helper. Added action buttons to retest summary modal (Fix & Retest, Optimize, Retest Again for still-failing scenarios).

v17.1 Feb 11, 2026 · 9:07 PM EST MAJOR

Pipeline Orchestrator

9-step automated pipeline (audit, fix, re-audit, settings audit, fix settings, text sim, surgeon, re-test, final grade). New vertical stepper UI with progress tracking, retry/skip/abort per step, KV persistence for resume. New /api/pipeline endpoint for state management.

v17.0 Feb 11, 2026 · 8:18 PM EST UPDATE

Pipeline Cancel Button

Added cancel button to full pipeline overlay. Cancellation checks between every step. Clean cancel state with proper cleanup.

v16.9 Feb 11, 2026 · 8:14 PM EST UPDATE

Full Pipeline Orchestrator

One-click pipeline: audit->fix prompt->re-audit->fix settings->5-phase text sim->results. Fixed 0% audit score edge case bug. Added audit score to shareable reports. Pipeline button on every agent card with animated step-by-step overlay.

v16.8 Feb 11, 2026 · 8:02 PM EST UPDATE

Integrated Scoring Algorithm

Audit score (prompt + settings) now factors into overall grade: 30% audit + 50% text + 20% voice. Added getAgentSettingsHealth() and getAgentAuditScore() helpers. Updated all 3 grading call sites. Score breakdowns show audit component when available.

v16.7 Feb 11, 2026 · 7:04 PM EST FIX

Grading Consistency Stabilization

Repetition scoring capped at max -15 points (was uncapped causing score nukes), Phase 4 hallucination trap scoring floor of 65% for agents that deflect honestly, Phase 5 memory grading only scores criteria that were actually tested in transcript, Phase 5 scenario generation now REQUIRES explicit memory-test actions in callerBehavior and keyTopics

v16.6 Feb 11, 2026 · 1:41 PM EST FIX

Fix Phase 4 Grading

Fixed Phase 4 hallucination trap grading - agent that deflects all unknown questions honestly but is repetitive now scores 60-75% instead of 0% F, Added explicit grading instruction that knowledge boundary tests prioritize no-hallucination over conversation quality, Added repetition criterion (3+ identical asks = minor fail not auto-fail)

v16.5 Feb 11, 2026 · 1:32 PM EST UPDATE

Universal Agent Testing

Dynamic prompt-aware voice testing for ANY agent type, Prompt hash caching (24h TTL saves 20-40s on repeat runs), Agent-type-aware grading (sales/support/IVR/scheduling/collections/survey auto-detected), De-hardcoded caller prompts (removed Sofia-specific qualification rules), Dynamic expectedDurationSec from AI-generated scenarios, Universal default grading criteria (not sales-specific), Scenario preview cards in voice test confirmation dialog, Regenerate Scenarios button to clear cache, Dynamic phase names in auto-fixer

v16.4 Feb 11, 2026 · 1:15 PM EST UPDATE

Auto-Optimize After Tests

Auto-trigger prompt optimization when voice scenarios fail, test caller qualification fix (qualifying budget answers), duplicate pricing label cleanup

v16.3 Feb 11, 2026 · 1:00 PM EST FIX

Bug fixes

Fix copyTranscript not defined error (was trapped inside renderScenarioCards scope, moved to global)
Remove duplicate GPT-4o pricing entry with wrong Claude Sonnet label
Fix all-N/A voice metrics edge case in getUnifiedGrade

v16.2 Feb 11, 2026 · 12:31 PM EST UPDATE

Scoring System Overhaul

Weighted voice metrics (latency 20%, disconnect 15%, interruption 15%, silence 15%, depth 10%, duration 10%, turns 5%, TTS 5%, repetition 5%)
Missing data excluded from score instead of defaulting to 75 (was inflating grades)
Dynamic turn count scales with profile duration (not hardcoded 24)
Wider conversation depth thresholds (10-55 words PASS instead of 15-40)
Stronger auto-fail penalties (25/30/35pts, hallucination 40pts each)
Deterministic grading guidance in AI prompt
Collapsible scenario cards (click header to expand/collapse, default collapsed)
Increased poll timeout to 7.5min for S3 Confusion scenario

v16.1 Feb 11, 2026 · 12:15 PM EST FIX

Revert S3 voice to 11labs-James

Reverted S3 Silence and Confusion voice from 11labs-Valentino (non-existent) back to 11labs-James. This was causing S3 ERR on all voice tests.

v16.0 Feb 11, 2026 · 12:07 PM EST FIX

Remove auto-load results, clean voice test page

Removed auto-load of previous test results on voice page render - was expanding all results and destroying the clean layout. Results still accessible via View Full Results button in saved history section.

v15.9 Feb 11, 2026 · 12:00 PM EST FIX

Voice Test Layout Fix

Fixed voice test results displaying in wrong page section

v15.8 Feb 11, 2026 · 11:56 AM EST FIX

Smarter Auto-Optimization

Surgical prompt edits preserve what's already working
Voice test results persist across page refreshes
Prompt History always accessible from voice test page

v15.7 Feb 11, 2026 · 11:45 AM EST FIX

Prompt Surgeon + Realistic Callers

Prompt Surgeon now applies targeted edits per scenario
One-click Retest button on each scenario card
Randomized caller names and accented voices for realistic testing

v15.6 Feb 11, 2026 · 11:29 AM EST FIX

Auto-Optimizer Pipeline Fix

Fixed optimizer data source to read actual test results
Per-criterion failure extraction for targeted prompt fixes
Retest now uses live results instead of stale data

v15.5 Feb 11, 2026 · 11:25 AM EST FIX

Voice Test Stability Fix

Fixed crash affecting all 5 voice test scenarios
Admin users now get full feature access
Improved error handling across test pipeline

v15.4 Feb 11, 2026 · 11:17 AM EST FIX

Testing Engine Stability

Fixed stale results when switching between agents
Improved error detection for AI provider outages
Global error handler with copy-to-clipboard for support

v15.3 Feb 11, 2026 · 10:52 AM EST FIX

Text Simulation Reliability

Fixed crashes in text simulation pipeline
A/B comparison results now handle edge cases gracefully
Self-learning engine wrapped with error recovery

v15.2 Feb 11, 2026 · 10:36 AM EST FIX

Text Sim Error Handling

Scenario cards now display errors cleanly with retry button
Cleaned up duplicate error display code

v15.1 Feb 11, 2026 · 10:23 AM EST UPDATE

Security Hardening

Full system security audit - clean bill of health
Hardened credential storage and access controls
Automated secret detection blocks accidental key exposure
Pre-deploy guardrails enforce version and changelog compliance

v15.0 Feb 11, 2026 · 9:23 AM EST MAJOR

Production Launch Ready

Complete auth system: signup, login, password reset with SHA-256 hashing
Stripe payment integration with checkout, webhooks, and transactional emails
Plan-based feature gating: Free 5 sims/month, Pro unlimited, Enterprise unlimited voice
Trial lifecycle: 14-day trial with auto-downgrade and 8-email drip sequence
User profile API with server-side data refresh on page load
Daily email-triggers cron for trial, re-engagement, and retention emails
Changelog auto-updates on every deployment via deploy.sh

v14.9 Feb 11, 2026 · 9:14 AM EST MAJOR

Plan-Based Feature Access

Free plan: 5 text simulations per month
Pro plan: Unlimited sims, voice testing, A/B comparison, auto-optimization
14-day trial gets full Pro access
Automated lifecycle emails for trial, engagement, and retention

v14.8 Feb 11, 2026 · 9:07 AM EST MAJOR

Production Stability

User profile syncs plan status on every page load
Expired trials auto-downgrade to free plan
Password reset and email triggers fully operational

v14.7 Feb 11, 2026 · 8:55 AM EST MAJOR

Stripe Payments

One-click upgrade to Pro and Enterprise from your account
Stripe checkout with automatic plan activation
Transactional emails for upgrades, cancellations, receipts, and failed payments

v14.6 Feb 11, 2026 · 8:37 AM EST MAJOR

User Authentication

Email and password signup with strength indicator
Secure login with password visibility toggle
14-day free trial starts automatically on signup
Login tracking and daily usage metrics

v14.5 Feb 11, 2026 · 3:55 PM EST MAJOR

Live WebRTC Call Testing + Call Explorer + Transcript System

Live web call testing via Retell WebRTC - test agents directly from browser with microphone
Real-time live transcript display during calls with role-based coloring (AI Agent vs Caller)
Call timer with elapsed time display during active calls
Call detail viewer with full transcript, recording playback, and call analysis breakdown
Auto-grade individual calls directly from Call Explorer
Universal transcript copy system - works across all 6 transcript formats (live, text sim, scenario, call explorer, recording, chat bubbles)
Recording audio player with inline playback controls
Copy transcript button on every transcript view across the entire app
Transcript toggle show/hide across all panels
Call library integration - save and review past calls
Stripe checkout endpoint for payment processing
Enhanced mobile responsiveness (v14.5 CSS overhaul)
Results dashboard consolidated view (v14.5 UX overhaul)

v14.4 Feb 11, 2026 · 3:30 PM EST UPDATE

Navigation + SEO + Deploy Automation

Updated nav: Features, Pricing, Use Cases, Docs, Blog, Login
Expanded footer with 10 links (pricing, use-cases, compare, blog, changelog, affiliate, docs, contact, privacy, terms)
JSON-LD SoftwareApplication schema for Google rich results
Canonical URL for SEO deduplication
Blog index page (was 404, now 200)
Sitemap expanded to 18 URLs (added /blog, /press, /va-checklist)
Auto-changelog: deploy.sh + git post-commit hook

v14.3 Feb 11, 2026 · 2:45 PM EST UPDATE

Enhanced Mobile Responsiveness + UX Overhaul

Complete mobile-responsive CSS overhaul across all app views
Consolidated Results Dashboard with unified view
UTM + affiliate referral tracking on landing page (localStorage + Plausible analytics)
Affiliate click tracking endpoint (no auth required)
Slack notification version tags updated

v14.2 Feb 11, 2026 · 11:30 AM EST UPDATE

Commission Tiers + Auth Hardening

Affiliate commission tiers implemented: 30% Starter / 35% Growth (10+) / 40% Elite (25+)
Auto-tier upgrades based on referral count on dashboard load
Password reset fully functional with SHA-256 + salt hashing
Session invalidation on password change
Google Auth, login, signup endpoints hardened with proper CORS
Admin API expanded with affiliate management actions

v14.1 Feb 11, 2026 · 8:20 AM EST MAJOR

Affiliate Program + Password Reset + Landing Pages

Affiliate program: 30% Starter / 35% Growth (10+ referrals) / 40% Elite (25+ referrals) recurring commissions
Affiliate dashboard with click tracking, conversion rates, earnings breakdown
Auto-tier upgrades based on referral count
Password reset with SHA-256 hashing, 1-hour token expiry, session invalidation
26 landing pages: pricing, docs, demo, use cases, comparison, tutorials, affiliate, contact, terms, privacy, press, changelog, email preview, login, signup, forgot/reset password, verify email, 404, admin, VA checklist, project dashboard, launch, trailer, screencast
Sitemap.xml + robots.txt for SEO
Blog infrastructure with SEO-optimized templates

v14.0 Feb 11, 2026 · 5:42 AM EST MAJOR

Admin Dashboard

Admin API with user management and system controls
Admin dashboard page with analytics overview
User listing, plan management, and activity logs
System health monitoring from admin panel

v13.4 Feb 11, 2026 · 5:35 AM EST UPDATE

Email Preview + Template Testing

Email preview page for visual template testing
Render any of 74 templates with sample data
Mobile-responsive email preview mode

v13.3 Feb 11, 2026 · 5:28 AM EST UPDATE

Automated Email Triggers

Email triggers cron endpoint for automated sends
Trial reminders at day 3, 5, 7, 10, 12, 13
Re-engagement emails for inactive users
Weekly digest and monthly recap automation

v13.2 Feb 11, 2026 · 5:18 AM EST UPDATE

Retention + Referral Emails

Churn prevention email sequences
Referral program email templates
Post-purchase follow-up and upsell emails
Admin notification templates for system events

v13.1 Feb 11, 2026 · 5:08 AM EST UPDATE

Onboarding + Engagement Emails

Welcome email sequence for new signups
Getting started guide emails with setup steps
Engagement emails: feature highlights, tips, best practices
Conversion emails for trial-to-paid nudges

v13.0 Feb 11, 2026 · 4:58 AM EST MAJOR

Email Template Engine

72 branded email templates across 10 categories
Template library (lib/emails.js) with dynamic data injection
Email delivery service via Resend API
Transactional email support for password resets, verifications

v12.4 Feb 11, 2026 · 4:48 AM EST UPDATE

Stripe Webhook Handler

Stripe webhook endpoint for real-time subscription events
Handles plan upgrades, downgrades, cancellations
Payment failure handling with retry logic
Automatic user plan sync on billing changes

v12.3 Feb 11, 2026 · 4:38 AM EST UPDATE

Google Auth + Email Verification

Google OAuth integration for one-click signup
Email verification page with token validation
Verification email sent on registration

v12.2 Feb 11, 2026 · 4:28 AM EST UPDATE

Password Reset Flow

Password reset API with secure token generation
Forgot password page with email input
Reset password page with new password form
Expiring reset tokens for security

v12.1 Feb 11, 2026 · 4:18 AM EST UPDATE

Login System + Session Management

Login API with credential validation
Session management with KV-backed tokens
Auth middleware (lib/auth.js) for protected routes
Logout endpoint with session cleanup

v12.0 Feb 11, 2026 · 4:05 AM EST MAJOR

User Registration + 14-Day Trial

Signup API with user creation and trial activation
14-day free trial with automatic expiry tracking
KV storage layer (lib/kv.js) for user data
Signup landing page with registration form

v11.2 Feb 11, 2026 · 3:50 AM EST UPDATE

Vercel Routing + Clean URLs

vercel.json route configuration for all API endpoints
Clean URL rewrites for landing pages
CORS and security headers configuration

v11.1 Feb 11, 2026 · 3:40 AM EST UPDATE

Mobile Responsive + Data API

Enhanced mobile responsiveness across all views
New data API endpoint for external integrations
Server-side performance improvements

v11.0 Feb 11, 2026 · 3:00 AM EST MAJOR

Server Architecture + API Layer

Rebuilt server architecture with Express.js routing
Vercel serverless API endpoints for all backend operations
Modular lib/ directory for shared utilities
Retell API integration module

v10.2 Feb 10, 2026 · 11:39 PM EST UPDATE

One-Click Prompt Fixes

One-click prompt fixes with live re-scan after applying
Audit timestamp tracking for historical fix records

v10.1 Feb 10, 2026 · 11:33 PM EST UPDATE

Actionable Audit Items

Actionable audit items with auto-fix recommendations
Dashboard quick-links to flagged issues
Priority-ranked recommendations engine

v10.0 Feb 10, 2026 · 11:05 PM EST MAJOR

Multi-Platform Support + Auto-Apply + Phone Management

Auto-apply prompt optimizations pushed directly to Retell
Vapi platform support alongside Retell AI
Phone number manager with call routing configuration
Cron-based scheduled testing (4h, 8h, 12h, 24h intervals)
Marketing landing page launch

v9.7 Feb 10, 2026 · 10:35 PM EST UPDATE

Shareable Reports + Realistic Callers

Shareable QA reports with premium dark mode design
Dashboard analytics cards with score trends
Overhauled scenario generation for realistic caller behavior
Fixed caller realism: no unprompted financial details
Auto-stop audio playback on tab switch

v9.6 Feb 10, 2026 · 9:59 PM EST UPDATE

A/B Prompt Comparison

A/B prompt comparison: clone, modify, compare side-by-side
Mobile responsive layout improvements

v9.5 Feb 10, 2026 · 9:55 PM EST UPDATE

PDF Export + Batch Testing

PDF report export with branded formatting
Batch testing across multiple scenarios
Dedicated settings page for configuration

v9.4 Feb 10, 2026 · 8:58 PM EST UPDATE

Scheduled Monitoring + Agent Health Checks

Scheduled monitoring with custom grading criteria
Agent health check dashboard
Floating space background for login page
Safety wrappers for all widget components
Fixed text testing crash from cached localStorage scenarios
Fixed name randomization using global replace

v9.3 Feb 10, 2026 · 8:01 PM EST UPDATE

Prompt Version History + Slack Alerts

Prompt version history with diff view between versions
Slack webhook alerts for failed tests

v9.2 Feb 10, 2026 · 7:50 PM EST UPDATE

Dashboard Score Trends

Dashboard score trends over time with charts
Notification system wired to key events
Quality gate tracking and pass/fail history

v9.1 Feb 10, 2026 · 7:41 PM EST MAJOR

Dynamic Caller AI + Self-Learning Engine

Dynamic caller AI with random personas and conversation variety
Self-learning engine that adapts to test patterns
Database hydration and persistence hooks for all data types
Audit tab redesign with actionable insights

v9.0 Feb 10, 2026 · 6:53 PM EST MAJOR

Database Persistence Layer (Upstash KV)

Upstash KV database integration for persistent storage
All test results, reports, and settings now persist across sessions
Text Testing UI cleanup and performance improvements

v8.9 Feb 10, 2026 · 6:13 PM EST UPDATE

Single-Agent Focus Redesign

Redesigned Overview, Settings Audit, and Text Testing for single-agent focus
Streamlined UI to center workflows around one agent at a time

v8.8 Feb 10, 2026 · 6:01 PM EST UPDATE

UI Simplification + Context Bar Redesign

Removed Compare tab in favor of dedicated A/B testing
Moved Costs tab from Testing Lab to top-right badge
Redesigned context bar and agent selector
Fixed duplicate caller names, enforced unique scenarios

v8.7 Feb 10, 2026 · 5:38 PM EST UPDATE

Cost Analytics + Layout Overhaul

Cost analytics page with historical spend tracking
Center-aligned all navigation, titles, headers, and context bar
Instant loading feedback for voice tests
Fixed voice test caller talking over agent at call start

v8.6 Feb 10, 2026 · 3:52 PM EST MAJOR

Prompt-Based Voice Testing

Prompt-based voice testing with AI-driven call scripts
Center-aligned navigation and mode toggle
Critical bug fixes for testing pipeline

v8.5 Feb 10, 2026 · 2:59 PM EST UPDATE

Caller Realism Overhaul

Complete overhaul of caller conversation scripts for natural behavior
Callers now behave like real humans with realistic pauses and responses

v8.4 Feb 10, 2026 · 2:26 PM EST UPDATE

Prompt Auditor + Quality Gates

Prompt auditor tabs with spend breakdown per prompt version
Quality gates with grade feedback and self-analysis
Conversation history tracking across test runs
Loading spinner during AI script generation
Clean SVG notification bell icon
Fixed voice tests using hardcoded scripts instead of AI-generated ones

v8.3 Feb 10, 2026 · 1:48 PM EST FIX

Voice Testing Stability

Forced AI script generation on every test run
Model upgrade for more reliable voice test responses
Role fix and visual loader improvements

v8.2 Feb 10, 2026 · 1:35 PM EST FIX

Voice Testing 404 Fix

Fixed voice testing 404 errors caused by invalid ElevenLabs voice IDs

v8.1 Feb 10, 2026 · 1:16 PM EST FIX

Agent Not Found Fix

Fixed voice testing agent-not-found bug
Removed legacy phases 6-8 from test pipeline

v8.0 Feb 10, 2026 · 9:22 AM EST MAJOR

QA Engine Overhaul

Removed phases 6-8 and renamed remaining phases to scenarios
Removed legacy costs tab from testing interface
Added progress indicator for test runs
Streamlined testing flow for faster iteration

v7.3 Feb 10, 2026 · 9:11 AM EST FIX

Caller Agent Quality Fix

Strict caller prompts with self-analysis and lower temperature
Fixed null reference crash in generateTestScenarios

v7.2 Feb 10, 2026 · 8:49 AM EST FIX

Voice Test Quality

Reduced to 5 phases for focused testing
Auto-generate test scripts per run instead of static scripts
Removed pulsing logo animation for cleaner UI

v7.1 Feb 10, 2026 · 8:41 AM EST FIX

Test Script Personas

Unique test personas across all 8 phases
Shorter, more focused test scripts

v7.0 Feb 10, 2026 · 8:39 AM EST MAJOR

Major QA Overhaul

Complete QA system overhaul: shorter calls, faster feedback
Per-test cost tracking for budget visibility
UX polish across all testing interfaces

v6.0 Feb 10, 2026 · 7:31 AM EST UPDATE

Live Transcript + Phone Assignment

Live call transcript with real-time updates
Stop button to end calls mid-test
4-minute safety timeout for runaway calls
Phone number assignment with dynamic phone map loading

v5.0 Feb 10, 2026 · 5:45 AM EST UPDATE

QA Leaderboard + Call Filters

QA Leaderboard tab ranking agents by test performance
Call filters and recording visibility controls
Select All checkbox for bulk agent selection
Audio player improvements and bigger login logo

v4.0 Feb 10, 2026 · 4:40 AM EST UPDATE

Simulation Engine Upgrades

5 simulation optimizations: model upgrade, smarter grading, new metrics
Voice simulation auto-retry with validation and stronger prompts
Raised call duration thresholds for better QA quality

v3.0 Feb 10, 2026 · 3:58 AM EST UPDATE

Call Recordings + MP3 Download

Call recordings for all agent-to-agent test calls
MP3 download and unified recording player
Fixed recording playback with stronger call duration

v2.0 Feb 10, 2026 · 3:28 AM EST UPDATE

Auth + Login Redesign

Branded login page with VoxGrade logo and brand colors
Password protection with auth restore
Redesigned topbar with prominent agent selector

v1.0 Feb 10, 2026 · 3:21 AM EST FOUNDATION

Initial Launch

VoxGrade deployed to Vercel
Retell AI integration for real voice agent calls
VoxGrade branding and initial dashboard UI