mesh-llm Releases

Rebases the llama.cpp fork on the latest upstream, incorporating security patches and new capabilities while retaining mesh-specific patches.

Mar 24, 2026

Adds CLI control commands and improves documentation with CI testing integration.

Mar 24, 2026

Updates documentation, diagrams, and repository references, including a repo rename to `michaelneale/mesh-llm`.

Mar 24, 2026

Finalizes documentation and install fixes, ensuring MCP server docs and correct asset names.

Mar 24, 2026

Introduces the Blackboard MCP server and changes default reasoning behavior for performance.

Mar 24, 2026

This release introduces the 'Blackboard' feature, enabling shared agent collaboration across the mesh network. It allows agents and users to post status updates, findings, and questions on an ephemeral blackboard that propagates across nodes without relying on cloud or external services.

Mar 23, 2026

Enhanced the Chat UI with support for rendering Mermaid diagrams and KaTeX math equations. Both features use CDN lazy-loading to ensure the bundle size remains low at 926KB.

Mar 23, 2026

Fixed 400 errors occurring during long conversations by increasing the small-host context window. Router logic was updated to stop preferring small models and tighten the random spread window.

Mar 23, 2026

Introduced an expandable thinking stream to visualize reasoning tokens in real-time and implemented auto-routing load spreading to distribute concurrent requests across top-scoring models.

Mar 22, 2026

Implemented an expandable thinking stream to visualize reasoning tokens and auto-routing load spreading to handle concurrent requests more effectively across models.

Mar 22, 2026

Updated the project description and chat placeholder text, along with adding a dashboard banner and adjusting wording for public versus private mesh configurations.

Mar 21, 2026

This release adds vision and multimodal support to the mesh, allowing models to process images alongside text. It introduces three new vision-native models and updates the default 'Mini' model, while retiring a legacy API application.

Mar 20, 2026

Improved download reliability by implementing infinite retries with exponential backoff and added a disk space check to prevent failures due to insufficient storage.

Mar 20, 2026

Polished the UI with new empty states and documentation links, fixed Fly API routing issues, and reverted unnecessary client connection changes.

Mar 20, 2026

Fixed desktop chat layout issues and improved mobile responsiveness, including a 16px font size fix for iOS Safari and added documentation links throughout the UI.

Mar 19, 2026

Cleaned up the demand tracking logic to remove pollution from auto and split GGUF suffixes, and fixed mobile rendering issues for code blocks and tables.

Mar 19, 2026

Significant performance optimization for Claude Code by disabling attribution headers that were invalidating the KV cache, along with general optimizations for local inference.

Mar 19, 2026

Fixes the publish watchdog to ensure the mesh remains visible to new nodes, while incorporating changes from v0.37.0 regarding model selection and startup logic.

Mar 18, 2026

Introduces smart model selection logic and fixes routing for tools schemas, alongside the ability to auto-start a mesh client from launchers.

Mar 18, 2026

Adds a visual 'Thinking' indicator to the Chat UI that displays a spinner during model reasoning and collapses into an expandable accordion.

Mar 18, 2026