Added “Use as Context” action for AI Studio images

You can now hover over any generated image in your gallery and click a new "Context" button to instantly add that image as a reference for your next generation. This removes the need to re-upload images and streamlines the creative flow by using previous outputs as inspiration or input context.

Enhanced context handling with model metadata

Updated the internal data structures to store both a display-friendly model name and the underlying model ID used for generation. This extra detail helps with clearer UI presentation and better tracking of which model produced each creation.

Improved studio UI and grid card interactions

The studio grid cards now include the new "Context" action icon, making the feature accessible directly from each image card. The immersive studio interface also handles adding reference images smoothly, with success notifications to confirm the action.

Other improvements

• Refined how model names are extracted and displayed for creations, ensuring consistent naming.
• Minor cleanup and code adjustments in the immersive studio components for better maintainability.

Fixed partial sync bug in social post account statuses

I fixed an issue where updating the status of a single social media account (Instagram, Facebook, TikTok) during a partial sync would accidentally overwrite the statuses of other accounts tied to the same post. This was causing some accounts to get stuck in “partial” or “retry” states, blocking the scheduler from progressing normally. Now, partial updates only modify the intended account rows without clobbering siblings.

Added scheduler lease recovery for stuck partial states

The scheduler now recognizes “partial” states at the account level as recoverable leases, allowing it to retry and recover from unexpected partial syncs that previously caused stuck posts. This helps keep the publishing workflow moving smoothly even when transient issues occur.

Improved test coverage for lease status resolution

I added tests to verify how lease statuses are interpreted during scheduler processing, ensuring that only valid states are accepted and that “partial” states are treated correctly for lease recovery.

Other improvements

• Updated database function to only expand missing account entries when fully replacing the status map, preventing unintended overwrites during partial updates.
• Documented this bug risk and ongoing investigation around account health detection and reconnect UX for auth failures in the tech debt log.

Removed costly row-level materialized view refresh triggers for custom image packs

I disabled the triggers that caused full materialized view rebuilds on every insert, update, or delete to custom image packs and their images. Instead, pack counts now come directly from the images table, reducing write amplification and disk I/O significantly while keeping pack counts accurate in the app and Partner API.

Added comprehensive tests for custom image pack queries

To ensure the new approach for fetching pack image counts is solid, I added a suite of tests simulating various scenarios. This helps catch errors early and maintain confidence as we iterate on the database layer.

Launched Google Nano Banana 2 image generation model with web search support

Introduced the new Nano Banana 2 model that supports a toggle for enabling web search grounding. This lets the model incorporate fresh web information when generating images. The toggle is now available in the inline model options, and usage costs adjust based on resolution and web search usage.

Other improvements

• Updated default model in studio to Nano Banana 2 for a better baseline experience.
• Added web search toggle UI and integrated it into image generation parameters.
• Minor UI tweaks in model selector components for badge display and option toggles.
• Documented new technical debt items related to leftover materialized view artifacts and duplicated slideshow automation types.

Redesigned Analytics Onboarding and Add Account Dialog

I revamped the onboarding flow and the add-account dialog for analytics tracking. The new design clarifies that you can track any public account — whether it’s yours, competitors, or UGC creators — and improves the UI with updated icons, gradients, and clearer calls to action.

Refreshed Analytics Queue and Bulk Refresh Status Display

The analytics queue now shows a more informative and interactive status panel for bulk refreshes. You can see progress with collapsible details, and the refresh process runs multiple accounts in parallel to speed things up. This should make tracking and refreshing analytics smoother and more transparent.

Added Free Analytics Seat Allowance and Updated Billing Logic

Introduced free analytics seats based on subscription tiers (Scheduler, Creator, Professional, Business) with allowances ranging from 1 to 5 free seats. Billing now prioritizes these free seats before charging credits, effectively reducing the cost of tracking accounts. This update is reflected throughout the dashboard and onboarding components.

Minor UI and Usability Improvements

• Improved social account list to optionally hide platform text when the avatar already shows the platform icon, cleaning up the display.
• Updated various UI texts and labels for better clarity and consistency.
• Added platform-specific gradient backgrounds for empty states to enhance visual appeal.

Paginated Blog Listing

I revamped the blog page to support pagination. Now, instead of loading all posts at once, it displays 12 posts per page and lets users navigate through pages, which should improve load times and user experience.

New Virtual Accounts Post Service

Added a dedicated service module for handling virtual account posts. This separates concerns and improves the way we fetch and sync posts from external providers, making the codebase cleaner and more maintainable.

Refactored Virtual Accounts Service

Refined the main virtual accounts service by removing post-fetching logic and delegating it to the new posts service. This modular approach will help us scale and troubleshoot features related to virtual accounts more efficiently.

Configuration and Import Fixes

Fixed some broken import paths caused by merge conflicts and restored missing hostnames and settings in the Next.js config that were accidentally dropped. This ensures smooth deployment and asset loading without disruptions.

Other improvements

• Added a pnpm lockfile to stabilize dependency versions across environments.
• Minor code cleanup and formatting tweaks in the blog components and layout.
• Increased allowed body size for uploads in Next.js config to support larger files.

Added onboarding activation range filters

I introduced selectable time windows (7 days, 30 days, 90 days, and all time) for the onboarding activation analytics in the admin dashboard. This lets us easily switch the view to analyze activation trends over different periods, making it more flexible to track progress and spot changes.

Implemented activation rate trend line

The onboarding activation chart now includes a trend line showing the activation rate over time, calculated as the ratio of feature activations to onboarding completions. This visual helps quickly understand how activation is progressing relative to completed onboardings.

Switched to completion-based activation summary

The summary metrics now focus on users who completed onboarding rather than just started it, providing a more accurate picture of activation rates and user engagement after onboarding finishes.

Other improvements

• Updated targeted tests to cover the new range filters and activation rate calculations.
• Refined date handling utilities to support the new all-time range and ensure consistent UTC day calculations.
• Improved documentation with a dedicated plan file and changelog entry outlining the feature and troubleshooting tips.