scira Releases

This release focuses on dependency updates and reliability improvements for the model gallery and API documentation. It introduces a new backend integration and fixes for CI workflows.

Aug 7, 2026

This release addresses stability issues, including fixes for VRAM handling, vLLM-CPP engine ABI compatibility, and React UI crashes. It also adds new gallery models and updates various dependencies.

Aug 6, 2026

A major release introducing vLLM-CPP (alpha), 3D generation modality, and a new audio.cpp backend. It also features significant UI performance improvements, gallery variant support, and VRAM budgeting.

Aug 5, 2026

This is a backport release for version 2.34 that includes dependency updates, dashboard improvements, and security hardening measures like OAuth2 validation.

Jul 28, 2026

This is a mainline release featuring dependency updates and core security fixes, specifically regarding OAuth2 redirect validation and log viewer performance.

Jul 27, 2026

Fixes configuration injection logic for llama.cpp serving options.

Jul 14, 2026

Major release adding managed voice cloning, local video generation, interleaved reasoning, and new audio engines.

Jul 14, 2026

Fixes bugs related to gvisor module paths, deployment workspace permissions, and dashboard redirects.

Jul 14, 2026

Fixes routing for gpt-5.6 responses, gvisor directives, and dashboard activity editing capabilities.

Jul 14, 2026

Aligns the Kotlin CodeInterpreter SDK with Sandbox SDK v1.0.16, introducing new isolation capabilities and fixing Maven repository URLs.

Jul 13, 2026

Introduces isolation convenience APIs, endpoint caching, and security patches while removing obsolete features and unifying logging.

Jul 13, 2026

Adds isolation convenience APIs, list sessions, bind mounts, and pool destroy support to the Kotlin Sandbox SDK, aligning with Python SDK features.

Jul 13, 2026

Introduces Credential Vault placeholder substitutions and security hardening, requiring specific configuration changes for DNS enforcement.

Jul 12, 2026

Fixes security vulnerabilities by preventing transitive dependencies from exposing endpoints and improves metrics normalization.

Jul 12, 2026

This release introduces a major update to the template creation process with a new builder UI and backend support, alongside several security hardening measures for OIDC authentication and chat infrastructure.

Jul 7, 2026

A runtime update providing prebuilt binaries for the FunASR llama.cpp engine. This release adds explicit AVX2 assets for faster performance while maintaining conservative x64 assets for maximum CPU compatibility.

Jun 29, 2026

A portable rebuild of the runtime that disables high-ISA CPU flags and native tuning to prevent crashes caused by illegal instruction errors on mismatched hardware.

Jun 28, 2026

A patch release addressing various fixes in the CSS formatter, JSON reporting, migration tooling, and Svelte linting. Notably, the 'Go-to definition' scanner feature is now disabled by default to prevent memory leaks.

Jun 23, 2026

Introduces prebuilt binaries for running local ASR (SenseVoice, Paraformer) using the llama.cpp runtime. This release introduces support for q8 GGUF models which offer half the file size with equivalent accuracy.

Jun 21, 2026

The initial release of the FunASR llama.cpp runtime, providing prebuilt binaries to run SenseVoice, Paraformer, and Fun-ASR-Nano locally without requiring Python or build tools.

Jun 21, 2026