Lingua Journey
Practise what you’d say in a real conversation.
Help a learner complete a realistic language task, receive useful feedback, and bring weak expressions back into later practice.
- 128
- commits in 6 weeks
- 273
- iOS tests passing
- 12
- practice scenes
- Swift 6
- strict concurrency
As of 2026-09-04, from the repository and releases






Captured from the development build
How it works
- 01
Choose a mission
A briefing sets the situation, language and completion goals.
- 02
Have a conversation
Microphone audio is converted into transport-ready packets; the active live path uses the Qwen relay.
- 03
Review the attempt
Conversation turns and mission context become feedback and candidate learning material.
- 04
Practice again
Saved cards can become future review targets and return to the next live conversation.
The hard parts, and what I did
Make one conversation prepare you for the next.
Practise a specific situation, then bring useful expressions into the next attempt. The idea is to make practice build on itself.
A conversation at human speed
- The problem
- Scripted drills feel nothing like ordering a coffee. A live conversation needs audio going both ways with turn-taking that doesn’t stall.
- What I did
- Microphone audio is converted to PCM packets and streamed through a small WebSocket relay; the app coordinates who is speaking and plays the reply as it arrives.
- Result
- AVAudioEngine capture and playback, a Node.js relay, and a mission-based session flow.
Practice that comes back
- The problem
- Feedback you read once is forgotten by the next day.
- What I did
- Each attempt turns into learning cards; a swipe deck brings them back for spaced review, and a listening puzzle lets you practise without turning on the microphone.
- Result
- Card library, review deck and listening puzzle all shipped in the 2026-09 build.
Independent projectiOS practice flow and live voice are implemented. Learning outcomes still need user validation.
Stack
SwiftUI · SwiftAVAudioEngine · PCM audioWebSocket voice transportNode.js · ws relayStructured evaluationLocal learning-card storeWeChat mini-program cloud functions
Implementation details
iOS learning flow with a live Qwen route and local rehearsal. Other adapters vary.
- Your voice
- Practice a concrete goal, such as ordering a coffee.
- Mission space
- Keep the setting, goals and conversation visible.
- Live session
- Capture, convert and play audio while coordinating turns.
- Learning feedback
- Turn the completed attempt into feedback and useful expressions.
- Learning cards
- Bring eligible saved expressions into future conversations.
- Voice relay
- The inspected live route connects through the Qwen relay.