How It Works
Complete guide to understanding our three-tier LLM initialization system and automated setup process.
Overview
Our initialization scripts provide a fully automated approach to LLM setup that eliminates manual configuration, ensures consistency across environments, and gets you productive in seconds rather than hours.
The system uses an intelligent three-tier architecture that allows precise selection of your development stack:
Framework → Scope → LLM Specialization
Three-Tier Architecture
The Structure
Framework (e.g., blissframework)
├── Backend Scope
│ ├── Universal (works with any LLM)
│ ├── Claude (optimized for Claude)
│ └── Gemini (optimized for Gemini)
├── Frontend Scope
│ ├── Universal
│ ├── Claude
│ └── Gemini
└── Fullstack Scope
├── Universal
├── Claude
└── Gemini
Tier 1: Framework Selection
Choose your development framework that defines the overall architecture and tooling approach:
- Bliss Framework - Enhanced productivity framework with LLM-optimized patterns
- Additional frameworks can be added as the ecosystem grows
Tier 2: Scope Selection
Define what part of the application stack you’re working on:
- Backend - Server-side development, APIs, databases, microservices
- Frontend - Client-side development, UI/UX, user interfaces
- Fullstack - Complete application development combining both
Tier 3: LLM Specialization
Choose your AI model optimization level:
- Universal - Works with any LLM (Claude, Gemini, GPT, etc.) with generic optimizations
- Claude - Specifically optimized for Anthropic’s Claude with specialized prompts and workflows
- Gemini - Specifically optimized for Google’s Gemini with tailored configurations
The 3-Step Process
Step 1: Three-Tier Selection
# Interactive three-tier selection process
curl -sSL https://initai.dev/install.sh | bash
# Step 1: Framework Selection
# Which framework would you like to use?
# 1) blissframework
# Framework for enhanced development productivity
# Step 2: Scope Selection
# Which development scope for blissframework?
# 1) Backend - Server-side development, APIs, databases
# 2) Frontend - Client-side development, UI/UX
# Step 3: LLM Specialization
# Which LLM will you use with blissframework (backend)?
# 1) Universal - Works with any LLM
# 2) Claude (Anthropic) - Optimized for Claude AI
# 3) Gemini (Google) - Optimized for Gemini AI
Why Three Tiers?
Framework Level - Defines your development approach and tooling philosophy Scope Level - Tailors the setup for your specific development focus LLM Level - Optimizes configurations for your chosen AI model
This granular selection ensures you get exactly the right setup for your project without unnecessary bloat.
Step 2: Package Download & Setup
# Selected: blissframework (backend) for Claude
# Package downloaded and extracted to: initai/blissframework-backend-claude
# Configuration saved to .initai.json
# Created CLAUDE.md with project instructions
# The script handles everything automatically:
# ✓ Download framework-specific package (ZIP format)
# ✓ Extract to organized directory structure
# ✓ Generate CLAUDE.md with tailored instructions
# ✓ Protect existing CLAUDE.md files from overwriting
# ✓ Save user preferences for future use
# ✓ Optional Claude Code launch integration
What happens during installation:
- Package Selection - Retrieves the exact package for your framework/scope/LLM combination
- Directory Creation - Sets up organized project structure
- File Protection - Warns before overwriting existing CLAUDE.md customizations
- Context Setup - Creates instructions for static (CLAUDE.md) vs dynamic (PROJECT.md) context
-
Preference Storage - Remembers your choices in
.initai.json
- IDE Integration - Optional Claude Code launch for seamless workflow
Step 3: Intelligent Development Workflow
# For Claude selections - automatic prompt:
# Spustit Claude? (y/n): y
# Spouštím Claude Code...
# Your project is now ready with:
# ├── initai/blissframework-backend-claude/ # Framework files
# ├── CLAUDE.md # Static instructions
# ├── PROJECT.md # Dynamic context (created by Claude)
# └── .initai.json # User preferences
Smart Context Management:
- CLAUDE.md - Contains initialization instructions (don’t modify)
- PROJECT.md - Your working context, decisions, and project evolution
-
Framework files - Organized in
initai/
directory with scope-specific templates
Platform-Specific Installation
Linux / macOS (Bash)
# Interactive three-tier selection
curl -sSL https://initai.dev/install.sh | bash
# The script will guide you through:
# 1. Framework selection (blissframework)
# 2. Scope selection (backend/frontend/fullstack)
# 3. LLM specialization (universal/claude/gemini)
Windows (PowerShell)
# Interactive three-tier selection
iwr -useb https://initai.dev/install.ps1 | iex
# Same three-tier selection process with Windows-optimized experience:
# - Native Windows path handling
# - PowerShell-style progress indicators
# - Windows Terminal color support
API Integration
# The system uses a RESTful API for package management:
GET /init/shared/list # Available frameworks
GET /init/shared/{framework}/{scope} # Universal packages
GET /init/shared/{framework}/{scope}/{llm} # LLM-specific packages
# Example API response structure:
{
"frameworks": [
{
"name": "blissframework",
"description": "Enhanced development productivity framework",
"scopes": [
{
"name": "backend",
"variants": [
{"name": "universal", "download_url": "/init/shared/blissframework/backend"},
{"name": "claude", "download_url": "/init/shared/blissframework/backend/claude"}
]
}
]
}
]
}
📋 Advanced: For detailed information about the multi-tenant architecture and future organization support, see Tenant System
Advanced Configuration
Custom Environment Setup
# Enterprise configuration
./install.sh --config enterprise \
--proxy http://proxy.company.com:8080 \
--cert /path/to/company.crt \
--no-telemetry
# Development team setup
./install.sh --team-config \
--shared-templates /shared/templates \
--git-hooks \
--pre-commit-checks
Framework-Specific Options
# Bliss Framework with extensions
./install.sh --provider claude \
--framework bliss \
--enable-extensions \
--install-models local
# Bliss Framework with enhanced features
./install.sh --provider gemini \
--framework bliss \
--enable-multimodal \
--optimize-performance
Supported Platforms & Requirements
System Requirements
- Operating System: Linux (Ubuntu 18+), macOS (10.15+), Windows (10+)
- Memory: Minimum 4GB RAM (8GB+ recommended)
- Storage: 2GB free space for full installation
- Network: Internet connection for initial setup
Supported Languages & Frameworks
- Python: 3.8+ (with pip, conda)
- Node.js: 16+ (with npm, yarn)
- Go: 1.19+ (experimental)
- Rust: 1.65+ (experimental)
Security & Enterprise Features
Built-in Security
# Secure by default configuration
✓ API keys stored in encrypted keyring
✓ Environment variables properly scoped
✓ Network requests use TLS 1.3
✓ No telemetry without explicit consent
✓ Local model caching encrypted
Enterprise Integration
- SSO Support: SAML, OAuth2, LDAP
- Proxy Configuration: Corporate firewall compatibility
- Audit Logging: Full installation and usage tracking
- Compliance: SOC2, GDPR, HIPAA ready configurations
Troubleshooting
Common Issues
Installation fails with permission error
# Fix permissions
sudo chown -R $USER:$USER ~/.local/share/initai
chmod +x ~/.local/bin/claude
# Or install to user directory
./install.sh --user-install
API connection timeouts
# Test connectivity
curl -I https://api.anthropic.com/health
curl -I https://generativelanguage.googleapis.com/health
# Configure proxy if needed
export HTTPS_PROXY=http://proxy.company.com:8080
./install.sh --use-proxy
Framework conflicts with existing installation
# Clean installation
./install.sh --clean --force
# Install in isolated environment
./install.sh --isolated --prefix /opt/llm-tools
Advanced Debugging
# Enable verbose logging
export INITAI_DEBUG=1
./install.sh --verbose
# Check installation logs
tail -f ~/.local/share/initai/logs/install.log
# Validate configuration
initai doctor --full-check
Next Steps
After successful installation:
- Configure Claude - Set up Anthropic Claude
- Configure Gemini - Set up Google Gemini
- Explore Examples - Practical usage examples
- Framework Guides - Deep dive into configurations
Performance Benchmarks
Our installation process is optimized for speed:
- Average installation time: 15-30 seconds
- Network usage: 50-150MB depending on selection
- First API call: <2 seconds after installation
- Framework loading: <500ms for most operations