🤖 AI CLI Agent

A persistent, context-aware AI assistant that lives in your terminal

Your digital cognitive layer — remembers everything, works across devices, learns from your code

🎯 The Problem

Every time you ask ChatGPT or Copilot a question, you waste time explaining:

• What project you're working on
• What you've already tried
• What errors you encountered
• What decisions you made

This agent already knows. It watches your terminal, monitors your files, and remembers everything — across ALL your projects and devices.

✨ Key Features

🧠 Three-Layer Memory System

| Layer | Purpose | Persistence | |-------|---------|-------------| | Episodic | Recent terminal commands, file edits | Session | | Working | Current task, context, blockers | Project | | Semantic | Long-term knowledge via Supermemory | Permanent/Cloud |

🔄 Multi-Device Sync

Work on your laptop, continue on your desktop. Your context follows you via Supermemory cloud sync.

🕷️ Proactive "Spidey Sense"

Agent watches your code and interrupts you with suggestions:

• "This function is 50+ lines. Consider refactoring."
• "You've been stuck on this file for 20 min. Need help?"
• "Similar code found in 3 files. Extract to utility?"

🧠 Cross-Project Intelligence

Remembers solutions across ALL your projects:

"I see you solved a similar auth issue in project-x 6 months ago..."

📚 Personal Knowledge Base (Supermemory-Powered)

• Error Solutions — Never solve the same error twice
• Decision Log — Remember WHY you made architectural choices
• Learning Capture — Build your personal programming wiki
• Code Snippets — Reusable patterns across projects
• Smart Reminders — Context-aware notifications

📝 Auto-Documentation

Generates daily journals explaining:

• What you did
• Why you made those changes
• Gotchas for future-you

🚀 Quick Start

Installation

npm install -g @vaibhav_dangaich/ai-agent

Setup

ai init

This will:

1. Configure your LLM provider (OpenAI or Gemini)
2. Optionally set up Supermemory for cloud sync
3. Optionally connect your GitHub for knowledge expansion

First Command

ai ask "what can you do?"

📖 Commands Reference

Core Commands

| Command | Description | |---------|-------------| | ai ask "question" | Ask the AI with full context | | ai remember "fact" | Store in long-term memory | | ai task "description" | Set current task context | | ai status | Show current context and memory | | ai history | View conversation history |

Agent Commands

| Command | Description | |---------|-------------| | ai watch --proactive | Watch files with AI suggestions | | ai journal | Auto-generate daily development journal | | ai journal --weekly | Generate weekly summary | | ai projects --add | Register project for cross-project intel | | ai projects --index | Index all projects into Supermemory | | ai projects --search "query" | Search across all projects |

Knowledge Base Commands

| Command | Description | |---------|-------------| | ai error "msg" | Find past error solutions | | ai error "msg" --solve "fix" | Store an error solution | | ai decide "decision" --reason "why" | Log architectural decisions | | ai decide --search "keyword" | Search past decisions | | ai learned "insight" | Capture learnings | | ai learned --search "keyword" | Search learnings | | ai snippet save "name" --code "..." | Save code snippet | | ai snippet find "name" | Find snippets | | ai remind "msg" --when "context" | Context-aware reminders | | ai remind --file "pattern" | Remind when editing specific files | | ai knowledge | View knowledge base stats |

Integration Commands

| Command | Description | |---------|-------------| | ai github --repos | List your GitHub repositories | | ai github --index | Index all repos into memory | | ai github --repo owner/name | Index specific repo | | ai sync | Sync context to cloud | | ai handoff | Prepare for device switch |

Setup Commands

| Command | Description | |---------|-------------| | ai init | Setup wizard for API keys | | ai setup | Show setup instructions | | ai supermemory | Debug Supermemory connection |

🏗️ Architecture

┌─────────────────────────────────────────────────────────────────┐
│                         CLI Interface                            │
│                         (bin/ai.js)                              │
├─────────────────────────────────────────────────────────────────┤
│                                                                  │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────────────────┐  │
│  │   LLM       │  │   Memory    │  │     Agent Modules       │  │
│  │  (LangChain)│  │  (3 Layer)  │  │                         │  │
│  │             │  │             │  │  • Proactive Watcher    │  │
│  │  • OpenAI   │  │  • Episodic │  │  • Cross-Project Intel  │  │
│  │  • Gemini   │  │  • Working  │  │  • Journal Generator    │  │
│  │             │  │  • Semantic │  │  • Knowledge Base       │  │
│  └─────────────┘  └──────┬──────┘  │  • Smart Reminders      │  │
│                          │         └─────────────────────────┘  │
│                          ▼                                       │
│  ┌─────────────────────────────────────────────────────────────┐│
│  │                    Supermemory SDK                          ││
│  │         (Cloud Sync, Semantic Search, Knowledge Store)      ││
│  └─────────────────────────────────────────────────────────────┘│
│                                                                  │
├─────────────────────────────────────────────────────────────────┤
│  ┌─────────────────────────────────────────────────────────────┐│
│  │                       Monitors                               ││
│  │                                                              ││
│  │    Terminal (zsh hook)  │  File Watcher  │  Git Monitor     ││
│  └─────────────────────────────────────────────────────────────┘│
└─────────────────────────────────────────────────────────────────┘

File Structure

cli_agent/
├── bin/
│   └── ai.js                 # CLI entry point (Commander.js)
├── core/
│   ├── agent/
│   │   ├── proactive.js      # Spidey Sense analyzer
│   │   ├── crossproject.js   # Cross-project intelligence
│   │   ├── journal.js        # Auto-documentation
│   │   ├── knowledge.js      # Error/Decision/Learning/Snippet store
│   │   └── reminders.js      # Context-aware reminders
│   ├── memory/
│   │   ├── episodic.js       # Recent activity (session)
│   │   ├── working.js        # Current task context
│   │   ├── conversation.js   # Chat history
│   │   ├── supermemory.js    # Supermemory SDK wrapper
│   │   ├── cloudsync.js      # Multi-device sync
│   │   └── index.js          # Memory orchestrator
│   ├── monitor/
│   │   ├── terminal.js       # Terminal command capture
│   │   ├── filewatcher.js    # File change detection
│   │   └── gitmonitor.js     # Git diff tracking
│   ├── integrations/
│   │   └── github.js         # GitHub API integration
│   ├── llm.js                # LangChain LLM wrapper
│   ├── prompt.js             # Prompt templates
│   └── config.js             # Configuration management
└── hooks/
    └── zsh-hook.sh           # Shell integration

🔧 Technical Deep-Dive

Supermemory Integration

The agent uses Supermemory as its semantic memory backbone. This is a key differentiator — Supermemory provides:

• Semantic Search — Find relevant memories by meaning, not just keywords
• Cloud Sync — Access your knowledge from any device
• Persistent Storage — Memories survive across sessions
• Scalable — Handles thousands of memories efficiently

What gets synced to Supermemory:

| Category | Tag | Description | |----------|-----|-------------| | Conversations | conversation-{project} | AI chat history | | Tasks | task-{project} | Current task context | | Activities | activity-{project} | Important file edits | | Errors | error-solutions | Error + solution pairs | | Decisions | decisions | Architectural decisions | | Learnings | learnings | Programming insights | | Snippets | snippets | Reusable code patterns | | Projects | project-{name} | Cross-project code index | | GitHub | github-{owner}-{repo} | Repository contents |

SDK Usage Example:

const { Supermemory } = require("supermemory");
const client = new Supermemory({ apiKey: process.env.SUPERMEMORY_API_KEY });

// Store with containerTags for organization
await client.add({
  content: "[ERROR SOLUTION]\nError: Cannot read property...\nSolution: Check for null...",
  metadata: { containerTags: ["error-solutions", "project-name"] }
});

// Semantic search across all memories
const results = await client.search.memories({
  query: "authentication error handling",
  limit: 10
});

LangChain Integration

Uses LangChain for provider-agnostic LLM calls:

const { ChatOpenAI } = require("@langchain/openai");
const { ChatGoogleGenerativeAI } = require("@langchain/google-genai");
const { ChatPromptTemplate } = require("@langchain/core/prompts");

// Switch between providers based on config
const llm = provider === "openai" 
  ? new ChatOpenAI({ openAIApiKey, modelName: "gpt-4o-mini" })
  : new ChatGoogleGenerativeAI({ apiKey, modelName: "gemini-pro" });

// Structured prompts with all context layers
const prompt = ChatPromptTemplate.fromMessages([
  SystemMessagePromptTemplate.fromTemplate(SYSTEM_TEMPLATE),
  HumanMessagePromptTemplate.fromTemplate("{question}")
]);

Memory Architecture

Layer 1: Episodic (Session)

• Stores: Terminal commands, file edits, git operations
• Backend: In-memory + JSON file
• Scope: Current session, per project

Layer 2: Working (Project)

• Stores: Current task, blockers, decisions
• Backend: JSON files in ~/.config/ai-agent/working/
• Scope: Persistent per project

Layer 3: Semantic (Cloud)

• Stores: Long-term facts, cross-project knowledge
• Backend: Supermemory cloud
• Scope: Permanent, multi-device

Proactive Agent Architecture

// The proactive watcher analyzes code in real-time
async function analyze(filePath, content, options) {
  const issues = [];
  
  // Check for code smells
  issues.push(...checkLargeFunctions(content));
  issues.push(...checkDeepNesting(content));
  issues.push(...checkDuplicateCode(recentChanges));
  
  // Check for stuck developer
  if (timeOnFile > 20 minutes) {
    issues.push({ type: "stuck", message: "Need help?" });
  }
  
  // Get AI-enhanced suggestion
  const suggestion = await getAISuggestion(issues);
  return { issues, suggestion };
}

📊 Resume/LinkedIn Highlights

Project Summary

AI CLI Agent — A context-aware AI assistant with persistent memory, multi-device sync, and proactive code analysis. Built with Node.js, LangChain, and Supermemory.

Key Technical Achievements

1. Three-Layer Memory System

   • Designed and implemented hierarchical memory (episodic, working, semantic)
   • Integrated Supermemory SDK for semantic search and cloud persistence
   • Achieved sub-second memory recall across 1000+ stored items

2. Multi-Device Synchronization

   • Built seamless handoff between devices using Supermemory cloud
   • Implemented conflict-free sync with device-tagged containers
   • Zero-config setup — just use same API key on any device

3. Proactive Code Analysis

   • Real-time code smell detection (large functions, deep nesting, duplicates)
   • Context-aware suggestions using LLM inference
   • Non-intrusive notification system

4. Personal Knowledge Base

   • Error solutions database with semantic search
   • Decision log with reasoning capture
   • Cross-project code pattern detection

5. LLM Orchestration

   • Provider-agnostic design (OpenAI, Gemini)
   • Streaming responses for better UX
   • Context-aware prompt engineering

Technologies Used

| Category | Technologies | |----------|-------------| | Runtime | Node.js 18+ | | CLI | Commander.js | | LLM | LangChain, OpenAI, Gemini | | Memory | Supermemory SDK | | Monitoring | Chokidar (file watcher), zsh hooks | | Storage | JSON files, SQLite (planned) | | API Integration | GitHub REST API |

Metrics

• 20+ CLI commands implemented
• 6 memory/knowledge modules integrated with Supermemory
• Real-time file watching across any editor
• Cross-device sync with zero manual intervention

🔑 Environment Variables

| Variable | Required | Description | |----------|----------|-------------| | OPENAI_API_KEY | One of these | OpenAI API key | | GEMINI_API_KEY | required | Google Gemini API key | | SUPERMEMORY_API_KEY | Recommended | Enables cloud sync, semantic search, knowledge base | | GITHUB_TOKEN | Optional | Index GitHub repos for knowledge expansion | | LLM_PROVIDER | Auto-detected | Force openai or gemini |

🛠️ Local Development

# Clone
git clone https://github.com/VaibhavDangaich/cli_agent.git
cd cli_agent

# Install
npm install

# Link globally
npm link

# Create .env
cp .env.example .env
# Edit .env with your API keys

# Test
ai ask "hello"

🚀 Roadmap

• [ ] Ghost Mode — Autonomous background agent that works while you sleep
• [ ] Phone Bridge — Control laptop agent from phone
• [ ] VS Code Extension — Deep IDE integration
• [ ] Team Features — Shared knowledge base across team
• [ ] Voice Interface — Speak to your agent
• [ ] Auto-PR — Agent creates PRs for tasks

📝 License

MIT License — see LICENSE

👨‍💻 Author

Vaibhav Dangaich

• GitHub: @VaibhavDangaich
• npm: @vaibhav_dangaich

Built with ❤️ for developers who hate repeating themselves

If you find this useful, consider giving it a ⭐ on GitHub!

Report Bug · Request Feature