cybos Releases

Introduces a new publishing mechanism for npm, adds diagnostic tools, and refines embedding and search behaviors for better stability and performance.

May 20, 2026

Extends the accuracy reporting capabilities to cover the wavekat-zh backend.

May 11, 2026

Introduces a new smart-turn variant with Chinese fine-tuning capabilities.

May 11, 2026

Adds pipecat smart turn training capabilities.

May 10, 2026

Improves the homepage monitoring display and resolves various stability issues across file management, decompression, and system upgrades, specifically addressing OpenWRT compatibility.

May 8, 2026

Enhances AI capabilities with DeepSeek V4 support, improves container and file management features, and addresses a wide range of UI and functional bugs.

May 7, 2026

Updates to the Codex executor and user interface, alongside the retirement of project routes and the company's hiring banner.

Apr 24, 2026

Maintenance release focusing on CI infrastructure migration and cross-compilation fixes.

Apr 17, 2026

Initial release of the project.

Apr 14, 2026

Announcement regarding the cessation of operations for the company behind Vibe Kanban.

Apr 10, 2026

Integration of the Fish Speech S2 Pro TTS engine with significant performance optimizations using GPU kernel compilation.

Apr 8, 2026

This release introduces a new controller for managing simulation turns and exposes audio duration metadata in the prediction model.

Mar 31, 2026

This release updates the default Cohere rerank model to `rerank-v4.0-fast` and welcomes a new contributor to the project.

Mar 28, 2026

MiroFish is a multi-agent AI prediction engine that constructs high-fidelity parallel digital worlds from real-world seeds. It enables users to simulate social evolution and predict future outcomes by injecting variables into a digital sandbox.

Mar 7, 2026

Introduces VibeVoice LoRA training and inference, adds 7 built-in streaming voice presets, and upgrades to Qwen3.5 LLM support.

Mar 5, 2026

Adds support for MCP (Model Context Protocol) tools and enhances resource updating capabilities.

Mar 4, 2026

Implements CUDA Graph acceleration for Qwen3 TTS for 5-10x speedup and adds multi-GPU support for subsystems.

Mar 1, 2026

Introduces support for the `uv` package manager to accelerate the installation process.

Feb 27, 2026

Adds a Prompt Hub for quick access to prompts, introduces Ollama as a new LLM backend, and enables batch audio generation by paragraph.

Feb 22, 2026

Enhances audio output with multiple formats (WAV, FLAC, MP3), embeds metadata directly into audio files, and improves persistent settings management.

Feb 14, 2026