Skip to content

Releases: appatalks/Synapse

Release list

SYNAPSE v0.1.0rc2

SYNAPSE v0.1.0rc2 Pre-release
Pre-release

Choose a tag to compare

View all tags
@appatalks appatalks released this 18 Oct 07:50
v0.1.0rc2
3933bfb

Highlights

  • Measured Auto Tune in Model Controls: Balanced and Maximum Context goals, optional experimental hybrid trials, cancellation/restoration, saved reports, and recommended client limits.
  • Clearer model-loading phases, elapsed time, pending backend checks, and cancellation.
  • A quick Reasoning Effort selector synchronized with Model Controls.
  • Explicit Q8 K/V cache plus Flash Attention with backend confirmation; FP16 remains the default and silent precision fallback is rejected.
  • Context allocations up to 262,144 tokens where the model/backend and hardware support them.
  • Updated monitoring screenshot, shorter README, and detailed application notes.

Downloads

  • SYNAPSE.pyz: standalone application; run with Python 3.10+ on Linux, macOS, or Windows.
  • synapse_local-0.1.0rc2-py3-none-any.whl: Python wheel.
  • SHA256SUMS: checksums for both application assets.

Requirements and limitations

Bring your own GGUF models and a compatible llama-server backend. Python, the inference backend, GPU drivers/toolkits, and model weights are not bundled in these application assets. NVIDIA GPU monitoring uses NVML; AMD/Metal sensor support is unavailable.

This is a prerelease, not a production-qualified inference engine. Optional Linux/NVIDIA hybrid execution and Q8 cache precision remain experimental and are not broadly answer-quality-qualified. Local capacity and timing checks are not universal fit, speed, or correctness guarantees. Keep the unauthenticated API on loopback; do not expose it through a tunnel or reverse proxy.

Releases are now prepared and published manually. Application CI and code scanning remain enabled.

See installation instructions and application notes.

Loading