Home

Sylphx Flow - AI-Powered Development Platform

MEP (Minimal Effective Prompt) Architecture - 90% less prompt, 100% better results.

Welcome to Sylphx Flow, the first production platform built on StarCoder2 tokenization for 70+ language hybrid search, combining curated knowledge, TF-IDF semantic search, and AI agent orchestration through the Model Context Protocol (MCP).

🎯 The MEP Difference

Traditional Way ❌

User: "I'm using TypeScript + React + Next.js,
      project structure is src/app,
      using shadcn/ui,
      current time is 2025-10-30,
      implement authentication with JWT..."

MEP Way ✅ (Sylphx Flow)

User: "implement authentication"

// AI automatically gets:
✅ Project environment
✅ Current time
✅ Existing patterns (via codebase search)
✅ Best practices (via knowledge base)

Result: 90% less prompt, 100% more accurate

🚀 Core Innovations

1. 🌟 StarCoder2 Tokenization - 70+ Languages

First production use of StarCoder2 tokenization for search:

• 70+ Programming Languages - TypeScript to Assembly
• Hybrid Search - Search in any language, find code in any language
• Semantic Understanding - Find code by what it does
• Cross-Language - Understand concepts across languages

# Search in ANY language - finds same code!
flow codebase search "user login handling"     # English
flow codebase search "處理用戶登入嘅邏輯"       # Chinese
flow codebase search "ユーザーログイン処理"     # Japanese

# All find the same code:
✅ authenticateUser()
✅ loginHandler
✅ verifyCredentials

2. 🎯 Two-Command Architecture

Setup once, use forever:

# 1. Initialize once
npx @sylphx/flow init

# 2. Use anywhere, anytime
npx @sylphx/flow run "your task"

# Or enter interactive mode
npx @sylphx/flow run
# Then type your prompt interactively

Why two commands?

• ✅ Zero configuration - Set up once, never again
• ✅ Zero maintenance - Everything auto-updates
• ✅ Zero complexity - Just describe what you want

3. 📋 Curated Knowledge (By Design)

No custom knowledge support - intentionally:

• ✅ Quality Guaranteed - Professionally curated
• ✅ Zero Maintenance - Auto-updated with flow init
• ✅ Optimized Performance - Local tokenization + TF-IDF = <100ms search

Project-specific patterns? Use codebase search:

flow codebase search "our authentication pattern"
# AI finds YOUR actual patterns

4. 🏗️ Pure Functional Architecture

Built with FP principles:

• ✅ Composition over Inheritance - Tools compose naturally
• ✅ Pure Functions - Predictable, testable, parallel
• ✅ Immutable Data - No side effects
• ✅ Pipeline Architecture - Natural data flow

✨ Core Capabilities

Capability	What It Provides	Who Benefits
📚 Knowledge Base	Curated guidelines for React, Next.js, Node.js	AI assistants, developers
🔍 Semantic Search	Find code by meaning, not keywords	Development teams
🤖 Agent Framework	Orchestrated AI for complex tasks	Project managers, architects
⚡ Real-time Indexing	Always up-to-date search index	Large codebases
🔧 CLI Tools	Command-line control	DevOps, power users

🎯 Quick Navigation

Getting Started

• Installation & Setup - Get up and running in minutes
• CLI Commands - Complete command reference
• MEP Design Philosophy - ⭐ Why MEP changes everything

Core Features

• Knowledge Base - Curated development guidelines
• Codebase Search - ⭐ StarCoder2 70+ language search
• Agent Framework - Orchestrated AI agents

Technical Deep Dive

• Technical Architecture - ⭐ StarCoder2 + Functional design
• Configuration - Customize your setup
• Contributing - Help improve Sylphx Flow

🚀 Quick Start

Two commands, infinite possibilities:

1. Setup Once

npx @sylphx/flow init

2. Use Forever

Direct prompt:

# Just describe what you want
npx @sylphx/flow run "implement authentication"
npx @sylphx/flow run "review for security" --agent reviewer

Interactive mode:

# Enter interactive Claude mode
npx @sylphx/flow run

# Then type your prompt naturally:
User: implement authentication
User: add password reset
User: write tests
# Keep chatting until done!

That's it! AI automatically gets:

• ✅ Your environment and time
• ✅ Your existing code patterns
• ✅ Best practices from knowledge base
• ✅ Everything needed for accurate results

Zero extra prompting required.

📖 Essential Reading

For First-Time Users

1. Installation & Setup - Install and configure
2. CLI Commands - Learn basic commands
3. Knowledge Base - Explore available guidelines

For AI Tool Users

1. MCP Integration - Connect your AI assistant
2. Knowledge Base - How AI uses knowledge
3. Codebase Search - Semantic search capabilities

For Developers

1. Architecture - System design and structure
2. Configuration - Advanced customization
3. Contributing - Contribute to the project

🎯 Real-World Use Cases

1. Onboarding New Developers

Your AI assistant has instant access to project patterns and architecture:

# AI searches knowledge base
knowledge_search("project architecture patterns")

# AI finds relevant code examples
codebase_search("authentication implementation")

2. Code Quality Maintenance

AI follows your team's standards automatically:

# Run code review with AI
flow run "review for security and performance" --agent reviewer

3. Complex Feature Development

Orchestrator breaks down and coordinates feature implementation:

# Orchestrator delegates to coder, reviewer, and writer
flow run "implement OAuth with tests and docs" --agent orchestrator

4. Semantic Code Discovery

Find code by what it does, not what it's called:

# Traditional search: Need exact keywords
grep -r "handlePayment"

# Semantic search: Describe what you're looking for
flow codebase search "payment processing logic"

🧠 How It Works

Knowledge Base System

Developer → AI Assistant → knowledge_search("react best practices")
                          ↓
                   Knowledge Base (curated guidelines)
                          ↓
                   Returns: React patterns, hooks, state management

Codebase Search

Developer → AI Assistant → codebase_search("authentication")
                          ↓
                   StarCoder2 Tokenization + TF-IDF (semantic understanding)
                          ↓
                   Returns: All auth-related code by meaning

Agent Orchestration

Complex Task → Orchestrator → Break down into subtasks
                             ↓
                  Delegate to: Coder → Reviewer → Writer
                             ↓
                  Synthesize results → Deliver to user

🔌 MCP Tools Available

When you run flow mcp start, AI assistants get these tools:

Knowledge Tools

• knowledge_search - Search guidelines and patterns
• knowledge_get - Get specific documents
• knowledge_list - List all resources

Codebase Tools

• codebase_search - Semantic code search
• codebase_reindex - Update search index
• codebase_status - Check index status

Time Tools

• time_get_current - Current timestamp
• time_format - Format times
• time_calculate - Time calculations

📊 Knowledge Base Content

Stacks (Framework-Specific)

• React App - Components, hooks, state management
• Next.js App - App router, server components, data fetching
• Node.js API - Express, middleware, error handling

Guides (Architecture & Design)

• SaaS Template - Multi-tenant patterns
• Tech Stack - Technology selection
• UI/UX - Design systems and accessibility

Universal (Cross-Cutting)

• Security - Auth, input validation, encryption
• Performance - Optimization, caching, monitoring
• Testing - Test strategies, coverage, integration
• Deployment - CI/CD, infrastructure, rollback

Data (Database Patterns)

• SQL - Query patterns, indexing, migrations

🤖 Agent Framework

Orchestrator

Role: Task coordination and delegation

Use When: Complex tasks requiring multiple specialists

Example:

flow run "implement feature with tests and docs" --agent orchestrator

Coder

Role: Code implementation and execution

Use When: Need to write or modify code

Example:

flow run "add user authentication" --agent coder

Reviewer

Role: Code review and quality assurance

Use When: Need code review for security, performance, or best practices

Example:

flow run "review this PR" --agent reviewer

Writer

Role: Documentation and technical writing

Use When: Need to create or update documentation

Example:

flow run "document the API endpoints" --agent writer

💡 Key Benefits

🔄 Consistency

"My AI assistants follow the same coding standards across all projects."

🧠 Intelligence

"AI understands my codebase by meaning, not just by keywords."

🤝 Collaboration

"Multiple AI agents work together on complex tasks."

🛠️ Control

"I have full control over what guidelines AI follows."

🎯 What's Next?

New Users

1. Install Sylphx Flow - Get set up
2. Learn CLI Commands - Master the basics
3. Explore Knowledge Base - See what's available

AI Tool Integration

1. MCP Integration - Connect AI tools
2. Configure Tools - Customize behavior
3. Use Case Examples - Real-world scenarios

Developers

1. Architecture Overview - Understand the system
2. Extend Knowledge Base - Add your own guidelines
3. Contribute - Help improve the project

🔗 Important Links

• GitHub Repository - Source code and releases
• Issue Tracker - Report bugs or request features
• Discussions - Community discussion

📊 System Status

• Current Version: 0.0.2
• Node.js Required: >= 18.0.0
• License: MIT
• Status: Active Development

---

💡 The Big Picture

Sylphx Flow isn't just another CLI tool or knowledge base. It's the missing infrastructure that transforms AI assistants from helpful tools into intelligent collaborators that understand your codebase, follow your standards, and coordinate complex tasks.

Transform your development workflow - Start with Installation & Setup.

---

Last Updated: 2025-10-30 | Edit this page | Report Issues