Skip to content

specs.md

Latest commit

 

History

History
76 lines (49 loc) · 2.43 KB

specs.md

File metadata and controls

76 lines (49 loc) · 2.43 KB

Product Requirements Document (PRD) - MVP

Project Name: duvent minimal viable product. Version: 1.0 Status: Draft Owner: tuann04 Last Updated: 2025-12-22

1. Executive Summary

This is a web app for improving speaking foreign languague purpose. Right now, i don't find any speaking appp that focus on a particular topic so i feel bored to talk English in these app so i build one.

2. Problem Statement

An app to talk with AI chat bot on a specific topic.

  • The Problem: English speaking is so hard at intermiadiate level because we always talk about basic topics like hobby or school, etc.
  • Current Alternatives: Practice with colleague.
  • Our Solution: Make a chatbot that can role-play our colleague or expert in the same field.

3. Target Audience

For one who want to improve their speaking skill for a niche field.

  • Primary Persona: A neft developer
    • Needs: Talk like an engineer
    • Frustrations: I have no life bruh, why English speaking classes always talk about life.
    • Tech Savviness: Med

4. MVP Features (MoSCoW Prioritization)

🔴 Must Have (The MVP)

If we remove these, the product doesn't work.

  • Core Feature A: Record a sentence from user then craft and response. The conversation must be in a topic.
  • Core Feature B: Print mispronoun or unnatural speaking flow and suggest improment.

🟡 Should Have (Post-MVP)

Important but not vital for Day 1.

  • Push notifications
  • Social sharing
  • Dark mode
  • Auth

🟢 Could Have (Nice to Have)

  • Proper monitoring

⚪ Won't Have (Out of Scope)

  • Paid subscription tiers (for now)
  • Multi-language support

5. User Flow / User Journey

How does the user get from A to B?

  1. Onboarding: User signs up (optional) -> Sees "Welcome" -> lands on Home Screen.
  2. Main Action: User clicks [+] conversation, choose a topic. then start speaking, record their speaking. then send to server and wait for response.
  3. Result: User hear response, and also improvement for their questions.

6. Technical Stack

Optimized for "Home Lab" deployment with scalability in mind.

  • Frontend: Next.js (TypeScript)
  • Backend: Golang (Orchestrator, WebSocket Server)
  • AI Service: Python (gRPC Server with Whisper & Kokoro)
  • LLM: Ollama (Llama 3 via REST API)
  • Database: PostgreSQL
  • Infrastructure: Localhost (Docker for DB), ngrok for remote access.