Table of Contents

• 🔬 Hypothesis
• 🧪 Experiment
• 📓 Observations
• 📊 Result
• 🔁 Next Step

Lab Logs #1: From Spec to Sandbox #

A Weekly Engineering Journal for TeenyOrb

🔬 Hypothesis #

If we could build a secure CLI tool that manages containerized coding sessions with LLM assistance - and demonstrate each phase of development with AI pair programming - then we could create both a useful tool and document effective AI-assisted development patterns.

This required:

• Comprehensive planning and architecture design
• CLI framework with container management
• File synchronization between host and containers
• Incremental development with clear milestones

🧪 Experiment #

Subject: TeenyOrb Phase 1 Implementation Environment:

• Go 1.24.2 with comprehensive planning spec
• LLM-assisted development (Claude/Copilot)
• Target: Complete CLI foundation with container management

Procedure:

1. Created comprehensive planning spec (docs/planning/v0.md) with 7-phase roadmap
2. Used LLM assistance to implement Phase 1 in a single focused session
3. Built complete CLI structure with Cobra framework integration
4. Implemented Docker-based container management with interfaces
5. Added file synchronization system using TAR-based transfers
6. Created working session management (create/list/stop commands)
7. Established configuration system and built functional binary

📓 Observations #

• Architecture Scope: LLM assistance helped implement far more than initially planned - complete CLI framework, container interfaces, and file sync
• Code Organization: Domain-driven design emerged naturally with clear separation between CLI, container, and sync concerns
• Interface Design: Abstract interfaces for container management appeared organically, enabling future backend flexibility
• Implementation Quality: Thread-safe session management and comprehensive error handling were implemented without explicit prompting
• Resource Management: Proper container resource limits and cleanup were included in initial implementation

Key Technical Achievements:

 1## Goal
 2Write a tiny Go CLI (“TeenyOrb”) that lets an LLM:
 31. plan coding tasks
 42. find/read/write files
 53. run commands in a scratch-built container (namespaces + cgroups)
 64. log every action to SQLite
 7
 8## Please expand:
 9- Features list
10- Non-Functional Requirements (security, resource limits, portability)
11- Architecture Sketch (modules + data flow)
12- Milestones (8 weekly posts)

• 🛠️ 1,396 lines of production-ready Go code in single session
• 🐳 Docker integration with advanced TAR-based file synchronization
• ⚡ Working binary (11.6MB) with help system and session management
• 🏗️ Clean architecture with interfaces, dependency injection, and error handling
• 📝 Complete configuration system with YAML and Viper integration

📊 Result #

Implementation Highlights:

 1// Complete CLI structure with Cobra
 2teeny-orb session create   // ✅ Working
 3teeny-orb session list     // ✅ Working  
 4teeny-orb session stop     // ✅ Working
 5teeny-orb generate "..."   // 🔄 Phase 2
 6teeny-orb review file.go   // 🔄 Phase 2
 7
 8// Container management with Docker
 9type Manager interface {
10    CreateSession(config SessionConfig) (Session, error)
11    ListSessions() ([]Session, error)
12    StopSession(id string) error
13}

🔁 Next Step #

With Phase 1 complete, we're ready for Phase 2: LLM Integration to add AI-powered capabilities:

• LLM provider interface with OpenAI/Anthropic implementations
• Configuration management for API keys and models
• Prompt templates and error handling
• Integration with existing generate and review commands

The CLI foundation and container management are solid - now we add the intelligence layer.

TeenyOrb has a pulse. Clone the repo, tweak the spec, and submit a PR with your Copilot-augmented ideas. 🔗 github.com/rcliao/teeny-orb