- MagesticAI

v2.2.0 Breaking

This release includes 2 breaking changes for platform teams planning a safe upgrade.

Published 14d AI Agents & Assistants

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

ReleasePort's take

Light signal

editorial:auto 9d

Version v2.2.0 adds OpenAI‑compatible LLM endpoints and refreshes cloud‑provider model catalogs while removing outdated entries.

Why it matters: Affects developers integrating with /settings/openai-compat models; updates to Claude, Codex, Gemini catalogs now reflect current generations after dropping superseded GPT‑5.2‑Codex, GPT‑5.1‑Codex‑Max/Mini, and Gemini 3 Flash preview entries.

Summary

AI summary

Updates ✨ Added, 🛠️ Fixed, and 🔧 Changed across a mixed release.

Changes in this release

Type	Severity	Summary	CVE
Feature
Feature	Medium	Adds OpenAI‑compatible LLM endpoints supporting multiple servers. Adds OpenAI‑compatible LLM endpoints supporting multiple servers. Source: llm_adapter@2026-05-25 Confidence: high	—
Feature	Medium	Refreshes cloud‑provider model lists to current generations. Refreshes cloud‑provider model lists to current generations. Source: llm_adapter@2026-05-25 Confidence: high	—
Feature	Medium	Drops superseded model entries (GPT‑5.2‑Codex, GPT‑5.1‑Codex‑Max/Mini, Gemini 3 Flash preview). Drops superseded model entries (GPT‑5.2‑Codex, GPT‑5.1‑Codex‑Max/Mini, Gemini 3 Flash preview). Source: llm_adapter@2026-05-25 Confidence: high	—
Feature	Medium	Introduces auto‑release GitHub workflow that tags and creates releases on version change. Introduces auto‑release GitHub workflow that tags and creates releases on version change. Source: llm_adapter@2026-05-25 Confidence: high	—
Bugfix
Bugfix	Medium	Fixes NameError crash in settings routes by adding missing logger. Fixes NameError crash in settings routes by adding missing logger. Source: llm_adapter@2026-05-25 Confidence: high	—
Bugfix	Medium	Makes .magestic-ai/ subtrees reachable in self‑host mode for MagesticAI project. Makes .magestic-ai/ subtrees reachable in self‑host mode for MagesticAI project. Source: llm_adapter@2026-05-25 Confidence: high	—
Bugfix	Medium	Updates “Merge to Main” button label to show actual target branch. Updates “Merge to Main” button label to show actual target branch. Source: llm_adapter@2026-05-25 Confidence: high	—
Refactor	Medium	Restructures LLM Providers settings page to two panels, removing QA Provider and Local LLMs panels. Restructures LLM Providers settings page to two panels, removing QA Provider and Local LLMs panels. Source: llm_adapter@2026-05-25 Confidence: high	—
Refactor	Medium	Updates CLAUDE.md path references from legacy .auto-claude/ to canonical .magestic-ai/. Updates CLAUDE.md path references from legacy .auto-claude/ to canonical .magestic-ai/. Source: llm_adapter@2026-05-25 Confidence: low	—

Full changelog

✨ Added

OpenAI-Compatible LLM Endpoints (#42, closes #39): First-class support for any OpenAI-compatible server (LM Studio, vLLM, OpenRouter, Together, Groq, LocalAI, Anyscale, and Ollama at :11434/v1). New GET /settings/openai-compat/models and POST /settings/openai-compat/test endpoints. Provider-choice + OpenAI-compat steps in onboarding. openai_compat available across the settings model, agent profiles, and every provider dropdown.
Up-to-date cloud-provider model lists (#45): Refreshed Claude / Codex / Gemini catalogs to current generation — Opus 4.7, GPT-5.5, GPT-5.3-Codex, Gemini 3.5 Flash stable, Gemini 3.1 Flash-Lite. Dropped superseded entries (GPT-5.2-Codex, GPT-5.1-Codex-Max/Mini, Gemini 3 Flash preview).
Auto-release workflow (#48): New .github/workflows/release.yml auto-creates the git tag and GitHub release whenever package.json#version changes on a push to main. Validates CHANGELOG.md has a matching section, extracts release notes, and skips out cleanly on non-release merges.

🛠️ Fixed

NameError crash in settings routes (#41): Added the missing module-level logger in settings.py, fixing crashes in Ollama / OAuth error paths.
Files tab in self-host mode (#40): .magestic-ai/ subtrees are now reachable when the target project IS MagesticAI itself (dogfooding) — the _is_app_internal_path guard previously blocked legitimate spec data.
"Merge to Main" button label (#43): Footer button and progress indicator now show the actual target branch (e.g. "Merge to dev") instead of a hardcoded "main".

🔧 Changed

Simplified LLM Providers settings page (#46): Restructured from three confusing panels to two — Cloud Agents (Claude / Codex / Gemini CLI logins) and OpenAI-Compatible Endpoints (covers LM Studio, vLLM, Together, Groq, and Ollama at :11434/v1). Removed the QA Provider panel (dead UI — actual QA execution reads the model from Agent Settings phase config, not the dropped qaLlmProvider setting). Removed the Local LLMs panel and the top-header Ollama status badge (Ollama now consumed via OpenAI-compat). Anthropic icon replaced with a neutral Cloud icon on the multi-vendor Cloud Agents panel.

📦 Updated

CLAUDE.md path references (#44): Updated documentation from legacy .auto-claude/ to the canonical .magestic-ai/ everywhere — the code has been using the new paths for a while.

Breaking Changes

Removed QA Provider panel from settings UI.
Removed Local LLMs panel and Ollama status badge; Ollama usage now via OpenAI‑compatible endpoint.

View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track - MagesticAI

Get notified when new releases ship.

About - MagesticAI

All releases →