Over the last few weeks we had a lot of fun running our first hackathon. We had several very solid submissions and many of you came up with great use cases for sync-powered AI. More on that in the feature story below.
It's also been another big month of shipping things. Our .NET SDK and SQL Server backend DB adapter both reached beta, meaning you can now build and ship production-ready PowerSync apps on the Microsoft stack. We also shipped query-driven sync for PowerSync's TanStack DB collection, a native Tauri SDK, and first-class Nuxt support.
Product updates shipped in March:
PowerSync Service:
- SQL Server Beta: SQL Server backend support is now production-ready. To get here, we needed to add automatic schema change handling. Now replication keeps working as your schema evolves, with minimal manual intervention. Release notes.
- v1.20.3: MongoDB now tracks the same replication metrics as Postgres (bytes, transactions, and chunks replicated), bringing the same observability to both. Changelog.
- v1.20.2: You can now define global CTEs so you don't have to repeat them across Sync Streams (thanks %%@Yahya%% for the nudge). Added %%instr()%% support in Sync Stream query expressions. SQL Server schema changes are now detected and handled automatically. Changelog.
- v1.20.1: Fixed a Sync Streams auth filter bug where certain auth subqueries could be silently ignored, potentially syncing data to users who shouldn't have access. It only applied in narrow circumstances, very few Cloud instances were affected and were patched automatically. Self-hosted users should upgrade to be safe. Security advisory.
Client SDKs:
- TanStack DB Query-Driven Sync: %%@tanstack/powersync-db-collection%% now supports on-demand sync mode — the collection loads only the rows your active live queries need, not the full SQLite table. No need to load 50,000 rows into memory if your list only shows 20. Release notes and Docs.
- .NET SDK Beta: The .NET SDK is production-ready. New since the alpha: full feature parity (Sync Streams, sync priorities, Rust-based sync client); concurrent reads so parallel queries no longer queue behind each other; faster query execution with significantly lower memory usage; and more idiomatic .NET APIs. Release notes and SDK Reference.
- Native Tauri SDK: Tauri apps can now use native Rust-backed SQLite instead of IndexedDB/OPFS (which reset on every app update). Sync state and watched queries are shared across all windows automatically. Release notes and SDK Reference.
- First-class Nuxt support: We shipped %%@powersync/nuxt%%. It includes auto-imported composables, built-in diagnostics and a SQLite inspector, and optional Kysely for type-safe queries. Release notes and Docs.
New demos & guides:
- Time-Based Sync Demo: Shows how to use Sync Streams to sync data by date ranges. Good reference for patterns like "sync the last N days" without redeploying sync config. Demo.
- SQL Grammar Reference: Railroad syntax diagrams showing the supported SQL when constructing queries in Sync Streams and Sync Rules. Sync Streams grammar and Sync Rules grammar.
- Sync Streams Bucket Troubleshooting: Run into %%PSYNC_S2305%% and not sure what to do? Sync Streams give you a lot of flexibility, but how your queries translate to sync bucket count isn't always obvious. We now explain how different patterns (subqueries, JOINs, hierarchies, subscription parameters) affect bucket count, and ways to reduce it. Docs.
Community spotlight
- Watch Dev build PowerSync apps from scratch in Codex: zero-shotting Supabase + PowerSync into a TanStack Start app using just the CLIs, then turning it into a full quest log from a single handwritten spec. Coding agents have come a long way, and so have we (AI accessible docs, overhauled CLI, official agent skills). A year ago these tools would hallucinate just trying to install our web SDK.
- Shoutout to %%@MrLightful%% who built a Tauri + PowerSync community example which helped motivate our native SDK.
- Thanks %%@qtpi-bonding%% for fixing the SQLCipher extension crash on iOS and macOS in the Flutter SDK. PR #381.
- Thanks %%@arcashka%% for catching the Android %%powersync-sqlite-core%% version mismatch that caused runtime crashes. PR #376.
- Built with PowerSync:
Feature story: AI Hackathon
The PowerSync AI Hackathon attracted a strong group of builders, and we received 53 submissions spanning a wide range of AI use cases. Across the final 8 winning projects, PowerSync and our partners awarded more than $9,000 in cash, credits, and prizes. Keep an eye on our socials - we'll be showcasing the winning projects there.
Huge shoutout to Christiaan, Manrich, Dev and Conrad for judging the submissions. It was a close competition, and many projects could easily have been placed.
Reviewing the submissions also gave us a clearer picture of how developers are building AI apps today. One recurring pattern was the heavy use of HTTP streaming for token output, even in cases where syncing outputs through PowerSync can be simpler and more reliable. We also saw that AI coding tools often struggled to use PowerSync correctly, with some of the same assumptions and API mistakes showing up across multiple submissions.
Those are useful lessons for us. They point to clear opportunities to improve our documentation, examples, and guidance so both developers and AI coding agents can build with PowerSync more effectively.
That’s it for this issue.