Roadmap: What's coming in the next release / 下个版本路线图

[wyuc]

We've been reading through all the issues and PRs over the past week — thank you to everyone who filed bugs, requested features, and submitted code. Here's what we're working on for the next release.

过去一周我们仔细看了所有的 issue 和 PR，感谢每一位提交 bug、功能建议和代码的朋友。下面是下个版本的计划。

Discussion TTS / 讨论语音

The most requested feature so far. Currently discussions are text-only while lectures have full TTS. We're adding voice playback to the discussion phase so every agent can speak.

目前呼声最高的功能。讲课环节有语音，但讨论环节只有文字。我们正在给讨论环节加上语音播报，让每个 Agent 都能开口说话。

• Distinct voice per agent / 每个 Agent 有独立声音
• Skip button to jump past current speech / 跳过按钮
• Works with all existing TTS providers including browser-native / 支持所有现有 TTS 供应商，包括浏览器原生（无需 API Key）

Tracking: #39, #27, #109

Presentation Mode / 演示模式

Full-screen immersive view for projecting lectures. Keyboard navigation, auto-hide controls. Big thanks to @YizukiAme for PR #133 which lays the foundation.

全屏沉浸式教学视图，支持键盘导航、控件自动隐藏。感谢 @YizukiAme 的 PR #133 提供了基础。

Tracking: #102

Discussion UX / 讨论体验增强

• Buffer-level pause: freeze text reveal without killing the stream, based on @YizukiAme's PR feat: discussion buffer-level pause (freeze text reveal without aborting SSE) #129 / 缓冲级暂停：冻结文字显示但不中断 AI 生成，基于 @YizukiAme 的 PR feat: discussion buffer-level pause (freeze text reveal without aborting SSE) #129
• Speed control for discussion text / 讨论文字速度控制
• Navigate between agent turns after discussion ends / 讨论结束后可在各 Agent 发言间跳转

Tracking: #73, #37

---

If you have thoughts on any of these, let us know here or in the linked issues.

如果对以上内容有想法，欢迎在这里或对应 issue 中反馈。

[bernard-thewagonltd]

Hi, first of great project. This will further put AI on the map, as this creates actionable content which people can benefit from. A missing feature which should be included is a way to reiterate over a created lesson module. This way one can review and find missing elements and re iterate.

[YizukiAme]

This has been requested before — see #29 and #72. Would be great to have!

[wyuc]

Thanks @bernard-thewagonltd! Post-generation editing is definitely on our radar. See #68 for the existing feature request — we are planning to tackle it in a follow-up release after this one.

感谢反馈！生成后编辑迭代确实是下一步要做的，可以关注 #68 的进展。

[nkmohit]

one more related feature I wanted to mention is pre-generation editing, currently after the outline finishes streaming, the generation doesn't give user any chance to review the outline it has produced and immediately calls /api/generate/scene-content with no user confirmation. so any “this isn’t what I wanted” moment is too late before the generation has already begun.

Also, there’s a OutlinesEditor component in the repo that looks intended for exactly this gap, but it isn’t connected to this flow right now so users currently can’t remove/reorder/update outline items before committing.

and beyond manual editing, a conversational loop would be more powerful where users describe changes in plain language and the AI regenerates the outline until they confirm. especially useful for course curators and saves unnecessary api calls.

opened #173 for this if it's worth tracking separately.

[wyuc]

Thanks for the write-up and for opening #173.

We'll consider outline editing for a future release. The current flow skips the review step on purpose. We tried adding human-in-the-loop confirmation steps in our hosted version, including outline review. It added friction without actually improving the experience. Turns out keeping the flow simple works better in practice.

Tracking in #173.

[upandey0]

Exciting roadmap! Looking forward to Discussion TTS especially.
One thing I noticed is missing — support for open/free LLM providers. Currently OpenMAIC requires API keys from OpenAI, Anthropic, or similar paid providers, which is a barrier for students and independent developers who want to explore the platform.
I'd like to contribute Hugging Face Inference API support as an additional LLM provider. This would let users run OpenMAIC with models like Mistral, Llama 3, Qwen etc. — either through HF's free tier or self-hosted via text-generation-inference.
I see there's also an open discussion on "Local LLM or Local TTS" (#thread) which aligns with this. Happy to take this on if the team is open to it.

[YizukiAme]

Hi @upandey0! Just wanted to clarify — OpenMAIC actually already supports using alternative LLM providers, including Hugging Face Inference API, ollama, vLLM, and essentially any service that exposes an OpenAI-compatible API.😊

The key is that most open-source model serving solutions (HF's text-generation-inference, ollama, LM Studio, etc.) all implement the OpenAI-compatible API format (/v1/chat/completions). So rather than needing a dedicated provider for each platform, you can simply configure the base URL in the existing OpenAI provider settings to point to your desired endpoint.

[lizhong006]

为什么开源版生成的PPT样式，相较于官方在线的，差距很大，美观性、布局都差很多？是配置的模型原因，还是代码逻辑不一样造成的。

[YizukiAme]

Model

[wyuc]

All three items from this roadmap have shipped in v0.1.0! 🎉

• Discussion TTS — Per-agent voice playback in discussions, with all TTS providers supported including browser-native (no API key needed). #211
• Immersive Mode — Full-screen view with speech bubbles, keyboard navigation (T/V/Esc/Space/M/S/C), auto-hide controls. #195, #256
• Discussion UX — Buffer-level pause #129, 1.25x speed #131

Plus many community-contributed features and fixes — see the full changelog.

---

After extensive (and hopefully thorough enough) testing, we've finally decided to tag the first official versioned release. This version is far from perfect — there are almost certainly bugs lurking in corners we haven't reached yet. Bug reports, feedback, and contributions are all very welcome.

One thing we'd especially love input on: testing strategy. The UX has a large number of logical branches (TTS providers × playback states × interaction modes × …), which makes manual testing incredibly painful. If you have experience with or ideas about GUI agents, visual regression testing, or other approaches that could help us scale test coverage, we'd love to hear from you.

---

Thanks to everyone who contributed! We'll open a new discussion for the next roadmap soon.

[YizukiAme]

Finally! Congrats on v0.1.0!!!!!!!!!!!!! 🎉
It's been an honor to contribute to this project~
Wishing OpenMAIC continued growth and success —
excited to see where the next roadmap takes us!