We have deployed version 1.1.0 to the adGPT core infrastructure. This patch focuses heavily on autonomous system recovery, expanded media generation, and structural upgrades to the mobile canvas.
All active workspaces have been updated automatically.
1. Autonomous Execution & Self-Healing
The core routing engine has been upgraded to audit its own workflows in real-time. The system no longer relies on basic success signals; it mathematically verifies that assets and campaigns exist after deployment.
- Self-Healing Architecture: If a complex workflow (like a multi-ad group build) partially fails due to platform API limits, the OS will automatically detect the missing components and resume the build without requiring human intervention.
- Zero-Conversion Rescue Strategy: The engine can now actively detect when Smart Bidding campaigns stall due to a lack of data. It will automatically draft and propose a recovery playbook to revive the campaign.
- Programmatic GA4 Control: adGPT has been upgraded from read-only reporting to full administrative control over Google Analytics 4. You can now prompt the OS to programmatically construct custom audiences, define key events, and adjust attribution models directly from the canvas.
2. Multi-Modal Creative Foundry
We have reduced the latency between data analysis and creative deployment by upgrading the native media synthesis nodes.
- Continuous Voice Interaction: You can now interact with the workspace completely hands-free using Voice Activity Detection (VAD). The system processes commands and speaks back with natural pacing.
- Surgical Audio Sanitization: The Text-to-Speech engine has been heavily refined. It autonomously strips out code snippets, raw URLs, and markdown formatting, ensuring executive summaries read aloud sound like a human analyst, not a machine reading code.
- Companion Mode: Media generation nodes (Video, Audio, Image) now support Picture-in-Picture execution, allowing you to continue auditing ad accounts while heavy media renders in the background.
3. Compute Ledger & System Integrity
Transparency in infrastructure costs is non-negotiable. We have upgraded the billing architecture to provide real-time visibility into compute usage.
- Live Compute Ledger: Your active workspace now features a real-time WebSocket connection to the credit service. Compute costs are calculated and displayed instantly upon task completion, with zero page reloads.
- Pre-Flight Predictions: The OS now calculates the exact GPU footprint of complex requests (like generating a 60-second video or executing a 10,000-row audit) and verifies your available compute before initiating the job, preventing mid-task failures.
- Credit Watchdog: The system actively monitors execution trajectories and provides soft visual alerts (Amber/Red) if your automated workflows approach your compute limits.
4. Tactile Mobile Workspaces
We have aggressively optimized the mobile interface to behave like native hardware rather than a responsive website.
- Bottom Sheet Architecture: Complex settings, asset managers, and media editors have been migrated to tactile "Bottom Sheets" to keep core actions perfectly aligned with the natural thumb zone on mobile devices.
- Intelligent Presentation Layer (IPL): Mobile charts have been upgraded. The canvas now intelligently auto-formats multi-metric datasets into Dual-Axis, Funnel, or Radar charts without requiring explicit formatting instructions.