SynapCores v2.1.3-ce — Parallel analytics, on by default
Community Edition · free · drop-in upgrade from v2.1.2-ce
v2.1.3 is a performance-tuning release. The analytical query path now uses every core on your machine automatically, the GenAI inference scheduler right-sizes itself to the hardware it boots on, and the Linux binary ships smaller and stripped. No new SQL surface and no API changes — the same queries you run today get faster the moment you upgrade.
Your analytical queries now use every core
SELECT DISTINCT, GROUP BY, and aggregate scans are now executed in parallel
across your CPU, automatically — no flags, no tuning. On a 6-core box a 189k-row
DISTINCT dropped from 44 ms to 3.8 ms single-query (11.7× faster) and
sustained 7× the throughput under concurrency (121 → 851 queries/sec),
scaling cleanly out to every physical core.
Dashboards, analytics endpoints, and lake queries that fan out over large tables get faster the moment you upgrade.
Why it matters: the work that used to pin one core and make a report wait now spreads across all of them. Bigger box → faster analytics, with nothing to configure.
GenAI concurrency that fits your hardware
The inference scheduler now auto-tunes to the machine it boots on — thread
allocation, concurrency, and queue depth are derived from your CPU and memory
instead of fixed defaults. Bursts of EMBED/GENERATE (bulk embedding, RAG
ingestion, concurrent chat) are absorbed with graceful backpressure instead
of capacity errors, and we verified every embedding is actually written and
valid under a 32-way burst. Single-stream GENERATE latency is
unchanged-to-faster.
Why it matters: bulk embed and RAG-ingestion jobs ride out spikes instead of erroring, and the engine right-sizes itself from a laptop to a big server with no knobs.
Smaller, hardened binary
The Linux build now ships stripped — ~24% smaller (334 MB → 254 MB) and
without the internal symbol map. Crash-looping installs on resource-constrained
machines now stop cleanly at a failed status instead of restarting forever.
Fully compatible — and no OLTP cost
Point lookups, single-row writes, and transactional workloads perform identically to v2.1.2 (measured at parity) — the new parallelism and scheduling add zero overhead to your operational path. Same SQL, same API, same data; just upgrade the binary.
Caveat (kept in on purpose): the speedups above are for analytical queries (scans/aggregates/DISTINCT) and scale with physical core count — point/OLTP operations are at parity, not faster.
Upgrade
docker pull synapcores/community:v2.1.3-ce
Drop-in swap from v2.1.x. Tarball installs: upgrade the whole extracted folder
(the engine loads ONNX Runtime from lib/ beside the executable).