Wed, Oct 08, 25, DEVELOPMENT ISSUES - Auto-imported from uconGPT project
- kor2Unity Development Issues
Auto-imported from:
D:/repos/aiegoo/uconGPT/eng2Fix/kor2fix/DEVELOPMENT-ISSUES.md
Original filename:DEVELOPMENT-ISSUES.md
Import date: Wed, Oct 08, 25
kor2Unity Development Issues
π― Current Development Focus
Korean language learning platform with self-hosted LLM integration
π Open Issues Summary
High Priority Issues
Issue #1: LLM Integration Strategy
- Status: π In Progress
- Priority: High
- Focus: Self-hosted model integration with existing infrastructure
-
Key Decisions:
- Primary environment:
minigpt4
conda environment - Model choice: Llama 2 7B-HF with MiniGPT-4 multimodal capabilities
- Port allocation: Migrate Ollama service to port 8203
- Primary environment:
Issue #2: UI Architecture Decision - TUI vs Web UI
- Status: π€ Decision Required
- Priority: High
- Focus: Choose between Terminal UI and Web UI for Korean learning
-
Current Recommendation:
- Phase 1: TUI MVP leveraging existing infrastructure
- Phase 2: Web UI for broader user adoption
- Decision Matrix Score: Web UI (38/50) vs TUI (35/50)
Issue #3: Conda Environment Configuration
- Status: β³ Ready to Start
- Priority: High
-
Focus: Activate and validate
minigpt4
environment for model deployment -
Next Action:
conda activate minigpt4
and environment validation
Infrastructure Status
β Completed
- FastAPI backend foundation (
/home/hsyyu/llm_api.py
) - Docker service orchestration (MongoDB, Ollama, backend)
- Dedicated port allocation strategy (8200-8299 range)
- Model inventory and hardware compatibility assessment
π In Progress
- Conda environment activation for model integration
- TUI vs Web UI architecture decision
- Korean language learning feature specification
π Planned
- Model loading and inference pipeline
- Korean tokenization and prompt engineering
- Frontend interface implementation (TUI or Web)
- Integration testing and validation
ποΈ Technical Architecture
Current Stack
βββββββββββββββββββ βββββββββββββββββββ βββββββββββββββββββ
β Frontend β β Backend β β AI Models β
β (TUI/Web) ββββββ FastAPI ββββββ minigpt4 β
β Port: 8200 β β Port: 8201 β β + Llama2-7B β
βββββββββββββββββββ βββββββββββββββββββ β Port: 8203 β
βββββββββββββββββββ
β β β
β β β
βββββββββββββββββββ βββββββββββββββββββ βββββββββββββββββββ
β Database β β Services β β Environment β
β MongoDB β β Docker β β Conda β
β Port: 8202 β β Compose β β minigpt4 β
βββββββββββββββββββ βββββββββββββββββββ βββββββββββββββββββ
Available Models
-
Llama 2 7B-HF:
/home/hsyyu/llama2-7b-hf/
(Machine appropriate) -
MiniGPT-4:
/home/hsyyu/checkpoints/
(Multimodal capabilities) - Model Size: 277MB pretrained + 47MB stage2 checkpoints
Development Environments
-
Primary:
minigpt4
conda environment (recommended) -
Fallback:
ai-cplus-dev
conda environment -
Future:
llama4-env
(requires high-end hardware)
π Development Roadmap
Week 1: Environment & Model Setup
- Activate
minigpt4
conda environment - Validate model loading and inference
- Test Korean text processing capabilities
- Benchmark performance on current hardware
Week 2: UI Decision & Implementation
- Finalize TUI vs Web UI architecture decision
- Begin frontend implementation based on choice
- Integrate Korean IME and text input handling
- Create basic Korean learning interface
Week 3: Backend Integration
- Extend FastAPI for Korean learning endpoints
- Implement model inference pipeline
- Add Korean tokenization and prompt engineering
- Create learning session management
Week 4: Testing & Validation
- Integration testing across full stack
- Korean language learning feature validation
- Performance optimization and tuning
- Documentation and deployment preparation
π Success Metrics
- Successful conda environment activation
- Model loading and inference validation
- Korean text processing accuracy
- UI responsiveness and usability
- End-to-end learning experience testing
π Next Immediate Actions
-
Environment Activation:
conda activate minigpt4
- Model Validation: Test model loading capabilities
- UI Decision: Choose TUI vs Web UI approach
- Development Sprint: Begin implementation based on decisions
Last Updated: August 7, 2025 Repository: aiegoo/kor2unity Branch: environment-setup-v1