jan/docs/src/pages/changelog/2025-08-14-general-improvs.mdx
2025-09-19 14:10:59 +07:00

78 lines
3.0 KiB
Plaintext
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

---
title: "Jan v0.6.8: Engine fixes, new MCP tutorials, and cleaner docs"
version: 0.6.8
description: "Llama.cpp stability upgrades, Linear/Todoist MCP tutorials, new model pages (Lucy, Janv1), and docs reorganization"
date: 2025-08-14
ogImage: "https://catalog.jan.ai/docs/mcplinear2.gif"
---
import ChangelogHeader from "@/components/Changelog/ChangelogHeader"
import { Callout } from 'nextra/components'
<ChangelogHeader title="Jan v0.6.8: Engine fixes, new MCP tutorials, and cleaner docs" date="2025-08-14" ogImage="https://catalog.jan.ai/docs/mcplinear2.gif" />
## Highlights 🎉
v0.6.8 focuses on stability and real workflows: major llama.cpp hardening, two new MCP productivity tutorials, new model pages, and a cleaner docs structure.
### 🚀 New tutorials & docs
- Linear MCP tutorial: create/update issues, projects, comments, cycles — directly from chat
- Todoist MCP tutorial: add, list, update, complete, and delete tasks via natural language
- New model pages:
- Lucy (1.7B) — optimized for web_search tool calling
- Janv1 (4B) — strong SimpleQA (91.1%), solid tool use
- Docs updates:
- Reorganized landing and Products sections; streamlined QuickStart
- Ongoing Docs v2 (Astro) migration with handbook, blog, and changelog sections added and then removed
### 🧱 Llama.cpp engine: stability & correctness
- Structured error handling for llama.cpp extension
- Better argument handling, improved model path resolution, clearer error messages
- Device parsing tests; conditional Vulkan support; support for missing CUDA backends
- AVX2 instruction support check (Mac Intel) for MCP
- Server hang on model load — fixed
- Session management & port allocation moved to backend for robustness
- Recommended labels in settings; permodel Jinja template customization
- Tensor buffer type override support
- “Continuous batching” description corrected
### ✨ UX polish
- Thread sorting fixed; assistant dropdown click reliability improved
- Responsive left panel text color; provider logo blur cleanup
- Show toast on download errors; context size error dialog restored
- Prevent accidental message submit for IME users
- Onboarding loop fixed; GPU detection brought back
- Connected MCP servers status stays in sync after JSON edits
### 🔍 Hub & providers
- Hugging Face token respected for repo search and private README visualization
- Deep links and model details fixed
- Factory reset unblocked; special chars in `modelId` handled
- Feature toggle for autoupdater respected
### 🧪 CI & housekeeping
- Nightly/PR workflow tweaks; clearer API server logs
- Cleaned unused hardware APIs
- Release workflows updated; docs release paths consolidated
### 🤖 Reasoning model fixes
- gptoss “thinking block” rendering fixed
- Reasoning text no longer included in chat completion requests
## Thanks to new contributors
· @cmppoon · @shmutalov · @B0sh
---
Update your Jan or [download the latest](https://jan.ai/).
For the complete list of changes, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.6.8).