Skip to content

Releases: bob798/speakeasy

v0.2b — Learning Loop

11 Mar 08:40
@bob798 bob798
v0.2b
f570e04

Choose a tag to compare

View all tags
v0.2b — Learning Loop Latest
Latest

Speakeasy v0.2b — Learning Loop / 学习闭环

Every conversation now closes with a review. Every mistake gets remembered.
每次对话结束后都有复盘。每个错误都会被记住。

What's New / 新功能

Conversation Debriefing / 对话复盘

After each session, Alex generates a structured review with three layers:
每次对话结束，Alex 自动生成三层复盘：

  • Summary / 摘要卡 — topics covered and fluency stats / 聊了什么、流畅度数据
  • Errors / 错误详情 — phrased as "a more natural way to say it", never as corrections / 用「更地道的说法」引导，不用「你错了」定性
  • Highlights / 亮点时刻 — moments where your English was particularly natural / 你表达得特别自然的那些句子

Graceful fallback to a "tip of the day" if LLM analysis fails.
LLM 分析失败时降级展示今日一句。

FSRS Memory Scheduling / 记忆调度

Grammar errors are tracked with spaced repetition:
语法错误通过间隔重复算法持续追踪：

  • First occurrence → pending card / 首次出现 → 创建 pending 卡片
  • Second occurrence → card graduates to active, FSRS scheduling begins / 第二次出现 → 升级为 active，开始 FSRS 调度
  • Alex naturally reinforces weak points in future conversations (max 3 cards injected per session) / Alex 在后续对话中自然强化薄弱点，每次最多注入 3 条
  • Rate cards 👍 / 🔁 from the review page to adjust your schedule / 在复盘页对卡片打分，调整下次复习时间

Click-to-Ask UI / 点击提问

Tap any word or phrase to ask Alex for a plain-language explanation.
点击任意词汇或短语，Alex 用简单语言重新解释。

API Changes / 接口变更

Endpoint Change
POST /chat Injects due grammar_cards into system prompt / 注入到期记忆卡片
POST /chat/summary Returns structured review + triggers FSRS update / 返回复盘数据并触发 FSRS
GET /review/{session_id} New — retrieve debriefing for any past session / 获取任意历史复盘
POST /review/{session_id}/rate New — submit card rating / 提交卡片评分

Database Changes / 数据表变更

Two new tables / 新增两张表：

  • grammar_cards — grammar error tracking with FSRS scheduling data / 语法错误记录，含 FSRS 调度数据
  • session_reviews — generated debriefing per session / 每次对话的复盘数据

Bug Fixes / Bug 修复

  • BUG-001 — Alex no longer echoes user's phrasing as a "natural alternative" / Alex 不再把用户的错误说法当作「更自然的表达」回传
  • BUG-002 — History view now shows "View Debriefing" for past sessions / 历史记录页正确显示「查看复盘」入口

Not in This Release / 不在本版本范围

Level assessment, preference extraction, memory management page, incognito mode, Plugin Registry → deferred to v0.3 (gated on retention data).

Level 评估、Preference 提取、记忆管理页面、无痕模式、Plugin Registry → 推迟至 v0.3（门控条件：留存数据达标）。

Full changelog: v0.2a...v0.2b

Assets 2
Loading

Speakeasy v0.2a — Voice & Memory

09 Mar 09:58
@bob798 bob798
v0.2a
dbfceb5

Choose a tag to compare

View all tags
Speakeasy v0.2a — Voice & Memory

Speakeasy v0.2a — Voice & Memory 🎙️

Talk to Alex. Out loud. And pick up where you left off. 和 Alex 开口说话，随时继续上次的对话。

What's New / 新功能

🎙️ Speech Input (STT) / 语音输入

Speak naturally — Alex listens. Powered by faster-whisper for low-latency, accurate transcription.

开口说话，Alex 会听。基于 faster-whisper，低延迟、高准确率的语音识别。

🔊 Voice Output (TTS) / 语音输出

Alex now talks back. Powered by edge-tts for natural, fluent English playback.

Alex 现在会开口回复你。基于 edge-tts，自然流畅的英文语音输出。

⚡ Streaming Response / 流式输出

Responses appear word by word — no more waiting for the full reply. Feels like a real conversation.

回复逐字出现，打字机效果，无需等待完整响应，对话感更真实。

💾 Persistent Conversation History / 对话历史持久化

Refresh the page, close the tab — your conversation with Alex is still there.

刷新页面、关闭标签，与 Alex 的对话记录依然保留。

Getting Started / 快速开始

git clone https://github.com/your-username/speakeasy.git
cd speakeasy

python -m venv venv

source venv/bin/activate       # Windows: venv\Scripts\activate


pip install -r requirements.txt


cp .env.example .env


Add your API key / 填入你的 API Key


uvicorn app.main:app --reload


Open / 打开 http://localhost:8000

What's Next / 下一版本预告 — v0.2b

  • 📋 Session Review — grammar highlights, key phrases, conversation summary
  • 🧠 FSRS Memory Scheduling — smart flashcard review for your mistakes
  • 💬 Click-to-Ask — tap any sentence to ask Alex to explain it

Version History / 版本历史

Version Status Highlights
v0.1 ✅ Text chat · Tip of the Day · Multi-model
v0.2a ✅ STT · TTS · Streaming · History
v0.2b 🔄 Session review · FSRS · Click-to-ask
v0.3 ⏳ Level assessment · Preference · Memory dashboard

Feedback / 反馈

Found a bug or have a suggestion? Open an issue or start a discussion.

发现 Bug 或有建议？欢迎提 issue 或在 discussion 中留言。

# Speakeasy v0.2a — Voice & Memory 🎙️

Talk to Alex. Out loud. And pick up where you left off.
和 Alex 开口说话，随时继续上次的对话。

What's New / 新功能

🎙️ Speech Input (STT) / 语音输入

Speak naturally — Alex listens. Powered by faster-whisper for low-latency, accurate transcription.

开口说话，Alex 会听。基于 faster-whisper，低延迟、高准确率的语音识别。

🔊 Voice Output (TTS) / 语音输出

Alex now talks back. Powered by edge-tts for natural, fluent English playback.

Alex 现在会开口回复你。基于 edge-tts，自然流畅的英文语音输出。

⚡ Streaming Response / 流式输出

Responses appear word by word — no more waiting for the full reply. Feels like a real conversation.

回复逐字出现，打字机效果，无需等待完整响应，对话感更真实。

💾 Persistent Conversation History / 对话历史持久化

Refresh the page, close the tab — your conversation with Alex is still there.

刷新页面、关闭标签，与 Alex 的对话记录依然保留。

Getting Started / 快速开始

git clone https://github.com/your-username/speakeasy.git
cd speakeasy

python -m venv venv
source venv/bin/activate       # Windows: venv\Scripts\activate

pip install -r requirements.txt

cp .env.example .env
# Add your API key / 填入你的 API Key

uvicorn app.main:app --reload
# Open / 打开 http://localhost:8000

What's Next / 下一版本预告 — v0.2b

  • 📋 Session Review — grammar highlights, key phrases, conversation summary
  • 🧠 FSRS Memory Scheduling — smart flashcard review for your mistakes
  • 💬 Click-to-Ask — tap any sentence to ask Alex to explain it

Version History / 版本历史

Version Status Highlights
v0.1 ✅ Text chat · Tip of the Day · Multi-model
v0.2a ✅ STT · TTS · Streaming · History
v0.2b 🔄 Session review · FSRS · Click-to-ask
v0.3 ⏳ Level assessment · Preference · Memory dashboard

Feedback / 反馈

Found a bug or have a suggestion? Open an [issue](../../issues) or start a [discussion](../../discussions).

发现 Bug 或有建议？欢迎提 [issue](../../issues) 或在 [discussion](../../discussions) 中留言。

Loading

v0.1.0

07 Mar 15:28
@bob798 bob798
v0.1.0
c85efc9

Choose a tag to compare

View all tags
v0.1.0

Speakeasy v0.1.0 — First Release 🎉

Talk to Alex. Practice English without the pressure.

Speakeasy is an AI-powered English conversation practice app. This is the first public release.

What's included

Alex, your AI conversation partner

A zero-anxiety English practice experience powered by large language models. Alex keeps conversations natural and encouraging — no red pens, no grammar lectures mid-sentence.

Multi-model support

Switch between providers without changing your workflow:

  • Anthropic Claude (claude-haiku-4-5)
  • DeepSeek (deepseek-chat)
  • Volcengine / Doubao (doubao-1-5-pro-32k)
  • Zhipu AI (glm-4-flash)

Clean chat UI

A focused, distraction-free interface to keep you in the conversation.

"Tip of the Day"

Each session ends with a practical English tip from Alex — something small you can use right away.

Getting started

git clone https://github.com/your-username/speakeasy.git
cd speakeasy

python -m venv venv
source venv/bin/activate       # Windows: venv\Scripts\activate

pip install -r requirements.txt

cp .env.example .env
# Add your API key and set MODEL_PROVIDER in .env

uvicorn main:app --reload
# Open http://localhost:8000

See [README.md](./README.md) and [FAQ.md](./FAQ.md) for proxy configuration and troubleshooting.

Known limitations

  • Conversation history is not persisted (resets on page refresh) — coming in v0.2
  • No voice input or output — coming in v0.2
  • Single user, no accounts

What's next — v0.2 roadmap

  • 🎙️ Voice input (STT) and voice output (TTS)
  • ⚡ Streaming responses with typewriter effect
  • 💾 Persistent conversation history
  • 📋 Session review with grammar feedback

Feedback

Found a bug or have a suggestion? Open an [issue](../../issues) or start a [discussion](../../discussions).

Loading