Hi-AI

Model Context Protocol Based AI Development Assistant

TypeScript + Python Support · 35 Specialized Tools · Knowledge Graph Memory · Session Context Auto-Injection

한국어

---

Table of Contents

• Overview
• Key Features
• v2.1.0 Updates
• Installation
• Tool Catalog
• Architecture
• Performance
• Development Guide
• License

---

Overview

Hi-AI is an AI development assistant that implements the Model Context Protocol (MCP) standard. It provides 35 specialized tools through natural language keyword recognition, helping developers perform complex tasks intuitively.

Core Values

• Natural Language Based: Automatically executes tools with Korean/English keywords
• Knowledge Graph Memory: Organizes memory relationships into a graph for exploring related information
• Multi-Language Support: TypeScript, JavaScript, Python code analysis
• Dependency Analysis: Visualizes code dependencies and detects circular references
• Enterprise Quality: 100% test coverage and strict type system

---

Key Features

1. Knowledge Graph Memory System

11 tools for organizing memory relationships into a graph for exploring related information:

• Session Context Auto-Injection: Automatically loads previous memories and knowledge graph on session start (v2.1 NEW)
• Relationship Linking: Sets semantic relationships between memories (related_to, depends_on, implements, etc.)
• Graph Traversal: Related memory exploration through BFS/DFS algorithms
• Multi-Strategy Search: Supports 5 search strategies (keyword, graph_traversal, temporal, priority, context_aware)
• Timeline: Time-ordered memory history visualization

Key Tools:

• get_session_context - 🚀 Auto-load context on session start (v2.1 NEW)
• save_memory - Store information in long-term memory
• recall_memory - Search stored information
• link_memories - Link relationships between memories
• get_memory_graph - Query knowledge graph
• search_memories_advanced - Multi-strategy search
• create_memory_timeline - Create timeline
• prioritize_memory - Manage memory priority

2. Semantic Code Analysis

AST-based code analysis and exploration tools:

• Symbol Search: Locate functions, classes, variables across the project
• Reference Tracking: Track all usages of specific symbols
• Dependency Graph: Code dependency visualization (v2.0 NEW)
• Circular Reference Detection: Automatic circular dependency detection (v2.0 NEW)
• Multi-Language: TypeScript, JavaScript, Python support

Key Tools:

• find_symbol - Search symbol definitions
• find_references - Find symbol references
• analyze_dependency_graph - Dependency graph analysis (v2.0 NEW)

3. Code Quality Analysis

Comprehensive code metrics and quality evaluation:

• Complexity Analysis: Cyclomatic, Cognitive, Halstead metrics
• Coupling/Cohesion: Module structure health evaluation
• Quality Score: A-F grading system
• Improvement Suggestions: Actionable refactoring recommendations

Key Tools:

• analyze_complexity - Complexity metrics analysis
• validate_code_quality - Code quality evaluation
• check_coupling_cohesion - Coupling/cohesion analysis
• suggest_improvements - Improvement suggestions
• apply_quality_rules - Apply quality rules
• get_coding_guide - Get coding guide

4. Project Planning Tools

Systematic requirements analysis and roadmap generation:

• PRD Generation: Automatic Product Requirements Document creation
• User Stories: Stories with acceptance criteria
• MoSCoW Analysis: Requirements prioritization
• Roadmap Creation: Phased development schedule planning

Key Tools:

• generate_prd - Generate Product Requirements Document
• create_user_stories - Create user stories
• analyze_requirements - Requirements analysis
• feature_roadmap - Feature roadmap generation

5. Sequential Thinking Tools

Structured problem solving and decision support:

• Problem Decomposition: Break down complex problems step by step
• Thinking Chain: Create sequential reasoning process
• Multiple Perspectives: Analytical/creative/systematic/critical thinking
• Execution Plan: Convert tasks into actionable plans

Key Tools:

• create_thinking_chain - Create thinking chain
• analyze_problem - Problem analysis
• step_by_step_analysis - Step-by-step analysis
• format_as_plan - Format as plan

6. Prompt Engineering

Prompt quality enhancement and optimization:

• Auto Enhancement: Convert vague requests into specific ones
• Quality Scoring: Clarity, specificity, context scoring
• Gemini Optimization: Google Gemini API prompting strategies

Key Tools:

• enhance_prompt - Enhance prompt
• analyze_prompt - Prompt quality analysis
• enhance_prompt_gemini - Gemini prompting strategies

7. Reasoning Framework

Systematic analysis of complex problems:

• 9-Step Reasoning: Problem decomposition, hypothesis exploration, risk assessment
• Logical Validation: Ensure completeness and precision

Key Tools:

• apply_reasoning_framework - 9-step reasoning framework

8. Usage Analytics (v2.0 NEW)

Tool usage statistics and analysis:

• Memory Statistics: Category distribution, time-based activity
• Graph Analysis: Connection statistics, cluster information

Key Tools:

• get_usage_analytics - Usage analytics query

9. UI Preview & Time

• preview_ui_ascii - ASCII UI preview
• get_current_time - Get current time

---

Hi-GCloud Integration

Hi-AI provides a powerful GCP operations + code modification workflow when used with hi-gcloud MCP.

Integration Method

When hi-gcloud detects errors, it automatically recommends hi-ai tools:

📋 Cloud Run logs: my-api
🔴 3 errors detected.

💡 hi-ai integration available: Use analyze_problem tool to analyze the cause,
   find_symbol to locate related code, and suggest_improvements for fixes.

Workflow Example

User: "Deployment failed"

[hi-gcloud]
→ Query error logs with gcp_run_logs
→ Found 3 errors, provide hi-ai integration hint

[hi-ai auto-integration]
→ Analyze error cause with analyze_problem
→ Locate related code with find_symbol
→ Propose fixes with suggest_improvements
→ Save solution with save_memory (prevent recurrence)

Installation

Installing both MCPs together enables automatic integration:

{
  "mcpServers": {
    "hi-ai": {
      "command": "npx",
      "args": ["-y", "@su-record/hi-ai"]
    },
    "hi-gcloud": {
      "command": "npx",
      "args": ["-y", "@polin-go/hi-gcloud"]
    }
  }
}

Integration Tool Mapping

hi-gcloud Situation	hi-ai Recommended Tools
Error log detected	analyze_problem, find_symbol
Deployment failed	step_by_step_analysis, suggest_improvements
Performance issue	analyze_complexity, check_coupling_cohesion
Cost increase	format_as_plan

---

v2.1.0 Updates

Major Changes

Hi-AI v2.1.0 is a minor release introducing Session Context Auto-Injection functionality.

New Features

Feature	Description
get_session_context tool	Query previous memories, knowledge graph, and timeline at once on session start
hi-ai://context/session resource	Automatically provides context when client reads the resource
Tool description improvement	Encourages LLM to automatically understand context on session start

Change Summary

Item	v2.0.0	v2.1.0	Change
Tool count	34	35	+1
Resource count	3	4	+1
Session context	Manual	Auto recommended	Improved

---

v2.0.0 Updates

Major Changes

Hi-AI v2.0.0 is a major release introducing knowledge graph-based memory system and advanced code analysis features.

New Features (6 tools)

Tool	Description
link_memories	Link memory relationships (knowledge graph)
get_memory_graph	Query/visualize knowledge graph (Mermaid diagram support)
search_memories_advanced	Multi-strategy search with 5 strategies
create_memory_timeline	Time-ordered memory timeline
analyze_dependency_graph	Code dependency analysis and circular reference detection
get_usage_analytics	Usage statistics/analysis

Architecture Improvements

• index.ts: 37 switch cases → Dynamic dispatch pattern
• MemoryManager: Knowledge graph feature added (395 lines → 823 lines)
• Code Optimization: Removed unnecessary dependencies (puppeteer-core)

---

Installation

System Requirements

• Node.js 18.0 or higher
• TypeScript 5.0 or higher
• MCP compatible client (Claude Desktop, Cursor, Windsurf)
• Python 3.x (for Python code analysis)

Installation Methods

NPM Package

# Global install
npm install -g @su-record/hi-ai

# Local install
npm install @su-record/hi-ai

Smithery Platform

# One-click install
https://smithery.ai/server/@su-record/hi-ai

MCP Client Configuration

Add to your Claude Desktop or other MCP client's configuration file:

{
  "mcpServers": {
    "hi-ai": {
      "command": "hi-ai",
      "args": [],
      "env": {}
    }
  }
}

---

Tool Catalog

Complete Tool List (35 tools)

Category	Count	Tools
Memory - Basic	6	save_memory, recall_memory, list_memories, delete_memory, update_memory, prioritize_memory
Memory - Graph	4	link_memories, get_memory_graph, search_memories_advanced, create_memory_timeline
Memory - Session	1	get_session_context 🚀
Code Analysis	3	find_symbol, find_references, analyze_dependency_graph
Thinking	4	create_thinking_chain, analyze_problem, step_by_step_analysis, format_as_plan
Code Quality	6	analyze_complexity, validate_code_quality, check_coupling_cohesion, suggest_improvements, apply_quality_rules, get_coding_guide
Planning	4	generate_prd, create_user_stories, analyze_requirements, feature_roadmap
Prompt	3	enhance_prompt, analyze_prompt, enhance_prompt_gemini
Reasoning	1	apply_reasoning_framework
Analytics	1	get_usage_analytics
UI	1	preview_ui_ascii
Time	1	get_current_time

Keyword Mapping Examples

Memory Tools

Tool	Korean Keywords	English Keywords
save_memory	기억해, 저장해	remember, save this
recall_memory	떠올려, 기억나	recall, remind me
get_session_context	세션 시작, 컨텍스트	session start, context
link_memories	연결해, 관계	link, connect
get_memory_graph	그래프, 관계도	graph, relations
search_memories_advanced	고급 검색, 찾아	advanced search, find

Code Analysis Tools

Tool	Korean Keywords	English Keywords
find_symbol	함수 찾아, 클래스 어디	find function, where is
analyze_dependency_graph	의존성, 관계	dependency, relations
analyze_complexity	복잡도, 복잡한지	complexity, how complex
validate_code_quality	품질, 리뷰	quality, review

---

Architecture

System Structure

graph TB
    subgraph "Client Layer"
        A[Claude Desktop / Cursor / Windsurf]
    end

    subgraph "MCP Server"
        B[Hi-AI v2.1.0]
    end

    subgraph "Core Libraries"
        C1[MemoryManager + Graph]
        C2[ContextCompressor]
        C3[ProjectCache]
        C4[PythonParser]
    end

    subgraph "Tool Categories"
        D1[Memory Basic x6]
        D2[Memory Graph x4]
        D2b[Memory Session x1]
        D3[Code Analysis x3]
        D4[Thinking Tools x4]
        D5[Quality Tools x6]
        D6[Planning Tools x4]
        D7[Prompt Tools x3]
        D8[Reasoning x1]
        D9[Analytics x1]
        D10[UI/Time x2]
    end

    subgraph "Data Layer"
        E1[(SQLite Database)]
        E2[Project Files]
    end

    A <--> B
    B --> C1 & C2 & C3 & C4
    B --> D1 & D2 & D2b & D3 & D4 & D5 & D6 & D7 & D8 & D9 & D10
    C1 --> E1
    C3 --> E2
    C4 --> E2
    D1 --> C1 & C2
    D2 --> C1
    D3 --> C3 & C4
    D5 --> C4
    D9 --> C1

Core Components

MemoryManager (v2.0 extended)

• Role: Persistent memory storage and knowledge graph management
• Technology: SQLite, better-sqlite3
• Features: CRUD, search, priority, graph relationships, BFS/DFS traversal
• Optimization: WAL mode, indexing, Prepared Statements

ContextCompressor

• Role: Context compression management
• Algorithm: Priority-based compression
• Features: Selective preservation based on importance

ProjectCache

• Role: ts-morph project caching
• Strategy: LRU algorithm
• Features: Improved repeated analysis performance
• Limits: 100MB/project, 200MB total

PythonParser

• Role: Python code AST analysis
• Method: subprocess execution
• Features: Symbol extraction, complexity calculation
• Safety: Timeout, automatic cleanup

Database Schema (v2.0)

-- memories table
CREATE TABLE memories (
  key TEXT PRIMARY KEY,
  value TEXT NOT NULL,
  category TEXT NOT NULL DEFAULT 'general',
  timestamp TEXT NOT NULL,
  lastAccessed TEXT NOT NULL,
  priority INTEGER DEFAULT 0
);

-- memory_relations table (v2.0 NEW)
CREATE TABLE memory_relations (
  id INTEGER PRIMARY KEY AUTOINCREMENT,
  sourceKey TEXT NOT NULL,
  targetKey TEXT NOT NULL,
  relationType TEXT NOT NULL,
  strength REAL DEFAULT 1.0,
  metadata TEXT,
  timestamp TEXT NOT NULL,
  UNIQUE(sourceKey, targetKey, relationType)
);

---

Performance

Major Optimizations

Project Caching

• Improved repeated analysis performance through LRU cache
• 5-minute TTL to maintain fresh state
• Resource management through memory limits

Memory Operations

• SQLite transaction for batch operation optimization
• Time complexity improvement: O(n²) → O(n)
• Fast lookup through indexing

Graph Traversal (v2.0)

• Efficient traversal with BFS/DFS algorithms
• Cluster detection with Union-Find
• Path finding optimization

---

Development Guide

Environment Setup

# Clone repository
git clone https://github.com/su-record/hi-ai.git
cd hi-ai

# Install dependencies
npm install

# Build
npm run build

# Development mode
npm run dev

Testing

# Run all tests
npm test

# Watch mode
npm run test:watch

# UI mode
npm run test:ui

# Coverage report
npm run test:coverage

Code Style

• TypeScript: strict mode
• Types: Use src/types/tool.ts
• Tests: Maintain 100% coverage
• Commits: Conventional Commits format

Adding New Tools

1. Create file in src/tools/category/ directory
2. Implement ToolDefinition interface
3. Register in src/index.ts toolHandlers
4. Write tests in tests/unit/ directory
5. Update README

---

Contributors

Special Thanks

• Smithery - MCP server deployment and one-click installation platform

---

License

MIT License - Free to use, modify, and distribute

---

Citation

If using this project for research or commercial purposes:

@software{hi-ai2025,
  author = {Su},
  title = {Hi-AI: Knowledge Graph-Based MCP Server for AI-Assisted Development},
  year = {2025},
  version = {2.1.0},
  url = {https://github.com/su-record/hi-ai}
}

---

Star History

Hi-AI v2.1.0

Knowledge Graph Memory · Session Context Auto-Injection · Dependency Analysis · 35 Specialized Tools

Made with ❤️ by Su

🏠 Homepage · 📚 Documentation · 🐛 Issues · 💬 Discussions