strix Releases

Nov 15, 2025

Introduces the `DSQ` quantization format designed to reduce VRAM usage by patching heavy linear layers while preserving float components. It supports parallel loading and offers three precision tiers (Q4_K, Q6_K, Q8_0) for DeepSeek-OCR and PaddleOCR-VL models.

Nov 14, 2025

A maintenance release fixing a critical memory issue on Blackwell GPUs where the RoPE (Rotary Positional Embedding) cache was incorrectly expanding to an excessive size.

Nov 10, 2025

Adds the `PaddleOCR-VL` backend as an alternative to `DeepSeek-OCR`, implements lazy loading for weight initialization, and standardizes sampling controls across the CLI and server.

Nov 9, 2025

Strix v0.3.1 introduces headless CLI mode for seamless automation and CI/CD integration, multi-target scanning for full-stack assessments, and improved visibility with per-severity vulnerability breakdowns. The release also increases the default iteration limit to 300 for deeper scans and adds agent iteration limit warnings.

Nov 1, 2025

This release introduces advanced permissions similar to Postgres Row Level Security (RLS), finalizes Deno support with multiple asset serving options, and adds mode helpers to streamline development workflows.

Oct 31, 2025

A minor maintenance update focusing on fixing pagination logic when total counts are unavailable and adding documentation notes for Cloudflare Image Optimization.

Oct 13, 2025

Introduces three execution modes (UI-only, Code-only, Hybrid) and allows deep customization of the Admin UI. Also improves media handling and SQLite connection options.

Oct 1, 2025

A maintenance release fixing Cloudflare serve export, improving raw node execution, and bumping Docker and Postgres versions.

Sep 15, 2025

This release adds support for Ollama models and Excel/CSV file formats, along with a UI improvement to hide the Getting Started guide after the first visit.

Jul 14, 2025

Focuses on security hardening with XSS prevention and enhances developer experience through improved documentation and AI integration.

Jul 9, 2025

Adds Material Design icons support and fixes keyboard layout issues for search shortcuts, while removing the Docker image health check.

Jun 10, 2025

Introduces the Novita AI generator and fixes critical bugs related to spaCy language processing and OpenAI model identification.

Jan 27, 2025

This release introduces support for TimesFM-2.0 models, updates the default point forecaster to the median head, and includes bug fixes for PyTorch decoding.

Dec 31, 2024

This release introduces new AI integrations such as Upstage, AssemblyAI, and Groq, alongside updates to deployment configurations and the Weaviate client.

Dec 10, 2024

Updates the template engine to enable fixed positioning for headers and footers within the generated documents.

Nov 11, 2024

Fixes issues related to field editing permissions and resolves CSS styling errors in the date picker component.

Oct 26, 2024

This release adds PyTorch support for TimesFM inference and introduces separate dependency installation options for the Pax (JAX) and Torch versions.

Oct 18, 2024

This major release (v2.0.0) introduces a comprehensive overhaul of the GoldenVerba interface, featuring a rebuilt Admin Interface, DocumentExplorer for vector visualization, and async ingestion capabilities. It expands support for different file types and AI providers, offering granular control over document settings and chunking strategies.

Sep 6, 2024

Updates dependencies and refactors the database abstraction layer to a 'db' module, introducing new data processing functions.

Jul 11, 2024