v0.2.1
THU-MAIC/OpenMAICv0.2.1Apr 26, 2026by wyuc
AI Summary
Focuses on advanced TTS features like voice cloning and per-model thinking configuration. It also improves the end-of-course experience and refactors orchestration prompts.
Key Highlights
- VoxCPM2 TTS provider with voice cloning
- Per-model thinking configuration (reasoning budget)
- End-of-course completion page with persistent quiz state
- Refactored orchestration prompts to file-based markdown templates
- Latest model support (GPT-5.5, DeepSeek-V4, MiMo, Hy3)
New Features
- Voice cloning
- Per-model thinking configuration
- End-of-course completion page
- Prompt refactoring
- Latest model support
Full Release Notes
### Features - **[VoxCPM2](https://github.com/OpenBMB/VoxCPM) TTS provider with voice cloning** — OpenMAIC adapts to user-managed VoxCPM backends (vLLM-Omni, Nano-VLLM, official Python API). Clone any voice from a reference audio clip you upload or record in the browser, or let Auto Voice generate a fitting voice from each agent's persona at synthesis time. Voice profiles are stored locally to keep the serverless setup model. The Agent Bar exposes a searchable, previewable voice picker that draws from the global VoxCPM voice pool [#496](https://github.com/THU-MAIC/OpenMAIC/pull/496) - **Per-model thinking configuration** — First-class metadata for each model's reasoning capability (effort levels, on/off toggle, adjustable budget, or fixed thinking) flows through chat and all generation paths and is mapped to the right provider-specific request fields (Anthropic `thinking`, OpenAI `reasoning`, etc.). The model selector becomes a unified provider/model/thinking popover with compact search and a much smaller toolbar footprint [#494](https://github.com/THU-MAIC/OpenMAIC/pull/494) - **End-of-course completion page with persistent quiz state** — When the outline is fully materialized, students see a course-complete view with quiz score card, scene-type stat cards, and a (motion-respecting) confetti celebration. Quiz answers persist on submit and grading results persist on completion, so navigating away and back restores the reviewing state with AI feedback intact instead of resetting [#484](https://github.com/THU-MAIC/OpenMAIC/pull/484) - Add latest released models including [GPT-5.5](https://github.com/THU-MAIC/OpenMAIC/pull/487), DeepSeek-V4 (`-pro`, `-flash`), Xiaomi [MiMo](https://github.com/XiaomiMiMo) (`mimo-v2.5-pro`, `mimo-v2.5`), Tencent [Hy3](https://github.com/Tencent-Hunyuan), and [OpenRouter](https://openrouter.ai/) as a multi-provider gateway [#481](https://github.com/THU-MAIC/OpenMAIC/pull/481) [#487](https://github.com/THU-MAIC/OpenMAIC/pull/487) - Add OpenAI image generation (GPT-Image-2) as a media provider [#481](https://github.com/THU-MAIC/OpenMAIC/pull/481) - Refresh built-in model registries across Anthropic, DeepSeek, Kimi, Qwen, MiniMax, Grok, OpenAI, GLM, SiliconFlow, and Ollama; persisted local settings now rehydrate in registry order so newly curated lists appear consistent without clearing state [#481](https://github.com/THU-MAIC/OpenMAIC/pull/481) - Add inline search for recent classrooms on the home page with deferred filtering by name and description, keyboard-driven open/clear/collapse [#476](https://github.com/THU-MAIC/OpenMAIC/pull/476) - Add Deep-Interactive badge on classroom thumbnails for sessions generated with Interactive Mode [#478](https://github.com/THU-MAIC/OpenMAIC/pull/478) - Replace always-included media instruction blocks in generation prompts with conditional snippet includes gated on `imageEnabled` / `videoEnabled` — disabled capabilities are removed from the prompt entirely instead of relying on negative-override directives the model often ignored [#490](https://github.com/THU-MAIC/OpenMAIC/pull/490) (by @YizukiAme) ### Bug Fixes - Fix language drift between outline and scene generation by unifying the languageDirective across the pipeline so the same target language flows from outline planning through every per-scene call [#474](https://github.com/THU-MAIC/OpenMAIC/pull/474) ### Other Changes - Refactor whiteboard role prompts to file-based markdown templates and add a geometry-conflict detector (overlap, line-through-bbox, canvas clipping) that surfaces problems back to the model. Eval (flash, repeat 3, gemini-3.1-pro scorer) shows overall quality 5.4 → 6.1 and overlap 6.3 → 8.1 from prompt + detector alone [#485](https://github.com/THU-MAIC/OpenMAIC/pull/485) - Migrate orchestration prompt builders (`buildStructuredPrompt`, `buildDirectorPrompt`, `buildPBLSystemPrompt`) from inline TS template literals to file-based markdown templates under `lib/prompts/`, sharing the loader infrastructure with the generation pipeline. `prompt-builder.ts` 890 → 314 lines; future content tweaks land as markdown edits [#459](https://github.com/THU-MAIC/OpenMAIC/pull/459) **Full Changelog**: https://github.com/THU-MAIC/OpenMAIC/compare/v0.2.0...v0.2.1