Changelog

Track our rapid iteration. We ship updates constantly to bring you the best local AI developer experience.

Robust Downloader Integrity

v0.6.5June 2026
  • ●Replaced strict model size validation checks against recipe metadata with dynamic server-side validation and file magic-bytes checks, preventing false-positive download failures on rounded or approximate size values.
  • ●Documentation & Templates
  • ●Exposed optional sha256 fields and updated size validation notes in TEMPLATE.yaml and recipes.mdx to make recipe creation easier.

Reliable Downloads & Cache Protection: Added automatic retries for download failures and ensured files don't conflict or get corrupted, automatically cleaning up bad cache files.

v0.6.4June 2026
  • ●Improved App Stability & Crash Detection: The app now shuts down cleanly, prevents leftover background tasks, and immediately detects engine crashes to keep you informed.
  • ●Smoother Command Line Experience: Increased the update file size limit to 500MB, fixed progress bar crashes, and added helpful error messages if you mistype a configuration path.
  • ●Conflict Prevention: Added checks to prevent starting on a port already in use and delayed exit logs so you can read what went wrong before the interface closes.

Port Configuration Fix: Fixed a bug where setting the custom port flag (--port) caused the backend server to immediately exit. You can now reliably configure custom ports.

v0.6.3June 2026

Host Configuration Fix: Fixed a startup crash caused by configuring the backend server host address (--host), ensuring port and host flags are correctly interpreted.

v0.6.2June 2026

Improved Testing Stability: Stabilized internal tests on Windows and macOS/Linux by using native components instead of brittle system scripts.

v0.6.1June 2026
  • ●Faster Windows Shutdown: Fixed a persistent 5-second shutdown delay on Windows when terminating processes.

Core Engine Rebuild & Runtime System Redesign

v0.6.0June 2026
  • ■Unified all inference backends under a common Engine interface (llama.cpp, vLLM, and SGLang), allowing seamless plug-and-play capability.
  • ■Standardized core CLI execution via a stage-based execution pipeline (Validate ➔ Download ➔ Resolve ➔ Probe ➔ MapFlags ➔ Launch ➔ Await ➔ Present).
  • ■Implemented the process.Supervisor shared execution layer to unify container/subprocess lifecycles, signal handling, watchdog-monitored startups, active health polling (/health), and concurrent persistent logging under ~/.cache/bloc/logs/.
  • ■Added capability-first flag mapping, dynamically resolving flag naming and syntax at runtime (e.g., -fa vs --flash-attn on) via help-probing instead of hardcoded string arrays.
  • ●Security Hardening
  • ■Removed dangerous shell execution patterns (sh -c) in pre-run script stages, replacing them with a secure allowlist-filtered system execution.
  • ■Mitigated path traversal risks by introducing strict containment checking on all user-supplied paths via filepath.Abs and prefix validation.
  • ■Enhanced API token privacy by preventing Hugging Face tokens from escaping into error messages or system logs.
  • ■Implemented environment filtering to scrub sensitive environment variables (such as LD_PRELOAD) from child subprocesses.
  • ●Performance Tuning & Correctness
  • ■Cached CPU and VRAM capabilities probing via sync.Once to prevent expensive redundant python execution loops.
  • ■Gated CUDA-based Docker testing behind a host-level nvidia-smi presence check, preventing automatic ~100MB downloads on CPU-only machines.
  • ■Resolved nil-map panic vectors in local download operations and fixed deprecated flag compatibility issues with vLLM 0.10.0+.

Modern Model Reasoning & Performance Support

v0.5.4June 2026
  • ●Updated compatibility flags for modern local model engines (like llama-server and vLLM) to prevent startup failures, and enabled native reasoning/chain-of-thought extraction for reasoning models like DeepSeek-R1.

Flash Attention Support

v0.5.3June 2026
  • ●Fixed a crash when running models with Flash Attention enabled by updating the settings flags to match the latest changes in the underlying model engine (llama.cpp).

Instant Screen Sync

v0.5.2June 2026
  • ●Fixed a visual delay in the interface where changes in the Chat or History tab did not update across both screens instantly.

Contextual UI Shortcuts

v0.5.1June 2026
  • ●Replaced static tips in the Studio with helpful, dynamic keyboard shortcuts and linked the "Recent Chats" list to your actual chat history.

Introduced the Bloc Studio

v0.5.0June 2026
  • ●A major update bringing a full-fledged Terminal User Interface (TUI) to Bloc for managing models, chatting, and viewing history.

Seamless Self-Updates

v0.4.1June 2026
  • ●Improved update checks on macOS for Homebrew users, and added a smooth fallback update mechanism if the primary update fails.

SGLang Engine Support

v0.4.0June 2026
  • ●Added native integration for the highly optimized SGLang engine, enabling massive throughput improvements for production deployments.

Simplified CLI Commands

v0.3.3June 2026
  • ●Renamed the deployment command from bloc deploy to a simpler and more intuitive bloc run, while keeping the old command working for backward compatibility.

Enhanced Security Safeguards

v0.3.2June 2026
  • ●Expanded the list of blocked or unsafe settings flags (now covering 36 flags) to protect your system from dangerous model behavior.

Resolved Freeze Bugs

v0.3.1May 2026
  • ●Fixed a critical freeze (deadlock) that occurred when deleting a model from your local cache.

Update Pointer Migration

v0.3.0May 2026
  • ●Successfully migrated the repository pointer in the update module to the new Bloc-ai/Bloc organization.

Hardened Security Certificates

v0.2.1May 2026
  • ●Updated secure connection certificates to maintain safe, encrypted communications.

Windows Cross-Compilation Support

v0.2.0May 2026
  • ●Resolved cross-compilation toolchain errors, stabilizing the Windows build pipeline.

Initial Release

v0.1.0May 2026
  • ●The first stable release of Bloc, featuring the core CLI, dynamic recipe system, and native llama.cpp execution capabilities.