v0.2.0

magnitudedev/magnitudev0.2.0Oct 2, 2026by alexchenzl

Upgrading?

AI Summary

Major overhaul of the extension architecture using WXT and an AI SDK, replacing LangChain and removing legacy API providers.

Key Highlights

  • Switched build system to WXT for single-package architecture with instant reloads
  • New AI engine using Vercel AI SDK for more reliable structured output
  • Added 'reasoning effort' setting to models that support it
  • Updated default models for OpenAI, Anthropic, Gemini, Grok, DeepSeek, and Ollama

Breaking Changes

  • Removed Groq, Cerebras, and Llama API providers (settings and keys deleted)
  • Minimal reasoning effort option removed (treated as Low)

New Features

  • WXT-based development and build system
  • Vercel AI SDK integration for LLM calls
  • Ollama now connects via OpenAI-compatible API

Full Release Notes

  What's new

  - Built with WXT: the extension is now a single package built with WXT instead of a multi-package
    monorepo. Development mode reloads pages instantly and reloads the extension automatically.
  - New AI engine: LLM calls now run on the Vercel AI SDK instead of LangChain. Structured output is more
    reliable with Anthropic, Gemini 3, DeepSeek and GPT-6 models.
  - Reasoning effort: models that support it now use a reasoning effort setting instead of temperature
    and top P.
  - Updated default models for OpenAI, Anthropic, Gemini, Grok, DeepSeek and Ollama.
  - Ollama now connects through its OpenAI-compatible API. Existing Ollama settings keep working.

  Breaking changes

  - Groq, Cerebras and Llama API providers were removed. Their saved settings and API keys are deleted
    when you upgrade. To keep using one, add it again as a custom OpenAI-compatible provider, and choose
    a new model for any agent that used it.
  - The Minimal reasoning effort option was removed. If you had saved Minimal, it is now treated as Low.

  For developers

  - Building the extension requires Node.js 24 LTS and pnpm 10.

Related Repositories