@baryodev/cli

Your autonomous AI coding assistant powered by local LLMs 🤖

Interactive command-line interface for BaryoDev - a local-first, privacy-focused AI development agent that runs entirely on your machine using Ollama.

Status

Version: 0.1.0 Stability: ✅ Production Ready Test Coverage: 480+ unit tests passing Build Status: ✅ Clean (zero TypeScript errors) Last Updated: 2026-01-03

Implementation Status

| Feature Category | Status | Notes | |-----------------|--------|-------| | ✅ Core CLI | Complete | Version, help, error handling | | ✅ Model Management | Complete | List, info, pull, remove, default | | ✅ Session Management | Complete | List, show, rename, delete, export | | ✅ Tools Management | Complete | Info, test, enable, disable | | ✅ Configuration | Complete | Show, set, unset, edit, reset | | ✅ BARYO.md Init | Complete | Framework detection, protection | | 🚧 Interactive Chat | Beta | Core working, advanced features pending | | 📋 Multi-Model Support | Planned | Phase 6 (Q2 2025) | | 📋 Plugin System | Planned | Phase 8 (Q3 2025) |

Quick Start

# 1. Install dependencies
pnpm install

# 2. Build the CLI
pnpm --filter @baryodev/cli build

# 3. Test Ollama connection
node packages/cli/dist/index.js connect

# 4. Start coding with AI
node packages/cli/dist/index.js chat

Prerequisites

Required

• Node.js 20.0.0 or higher
• pnpm 8.0.0 or higher
• Ollama installed and running locally

Installing Ollama

# macOS
brew install ollama

# Linux
curl -fsSL https://ollama.com/install.sh | sh

# Start Ollama
ollama serve

Pull Recommended Models

# Fast, efficient coder (recommended)
ollama pull qwen2.5-coder:7b

# Alternative models
ollama pull deepseek-coder:6.7b
ollama pull codellama:13b
ollama pull llama3.2:3b

Architecture

System Overview

graph TB
    User[User]
    CLI[CLI Package]
    Core[Core Package]
    Ollama[Ollama Server]

    User -->|Commands| CLI
    CLI -->|ReAct Loop| Core
    Core -->|API Calls| Ollama
    Ollama -->|Streaming Response| Core
    Core -->|Tool Results| CLI
    CLI -->|Display| User

    style CLI fill:#e1f5ff
    style Core fill:#fff3e0
    style Ollama fill:#f3e5f5

Package Dependencies

graph LR
    CLI[cli]
    Core[core]
    Config[config]
    Parser[parser]
    Shared[shared]
    Tools[tools]

    CLI --> Core
    CLI --> Config
    Core --> Parser
    Core --> Config
    Core --> Tools
    Core --> Shared
    Config --> Shared
    Parser --> Shared
    Tools --> Shared

    style CLI fill:#4CAF50,color:#fff
    style Core fill:#2196F3,color:#fff
    style Shared fill:#FF9800,color:#fff

Command Flow

sequenceDiagram
    participant User
    participant CLI
    participant Core
    participant Ollama

    User->>CLI: baryodev chat "Create a REST API"
    CLI->>Core: Initialize ReAct engine
    Core->>Ollama: Stream LLM request
    Ollama-->>Core: Streaming response
    Core->>Core: Parse tool calls
    Core->>Core: Execute tools
    Core->>Ollama: Send tool results
    Ollama-->>Core: Continue response
    Core-->>CLI: Display results
    CLI-->>User: Show output

Features

🎯 Core Capabilities

1. Interactive AI Chat

Real-time conversations with local LLMs

• Streaming responses for immediate feedback
• Context-aware multi-turn conversations
• Automatic context management (7-tier priority system)
• Token usage tracking and visualization

2. Model Management

Complete control over your LLM models

• List all available models
• View detailed model information
• Set default model
• Guided model pull/remove

3. Session Persistence

Never lose your work

• Save conversations with custom names
• Resume previous sessions
• Export to JSON, Markdown, or plain text
• Session metadata tracking

4. Project Configuration (BARYO.md)

AI understands your project

• Automatic framework detection
• Custom coding conventions
• Protected file patterns
• Project-specific context

5. Tool Ecosystem

Extensible tool system

• File operations (read, write, list)
• Command execution
• Pattern searching (grep)
• Custom tool development

6. Smart Configuration

Flexible config management

• Global and local configuration
• Hierarchical overrides
• Dot notation access
• Safety guardrails

Commands Reference

Core Commands

baryodev chat [options]

Start an interactive chat session with the AI agent.

# Basic usage
baryodev chat

# Use specific model
baryodev chat --model deepseek-coder:6.7b

# Configure context and iterations
baryodev chat --max-tokens 16384 --max-iterations 20

# Disable streaming
baryodev chat --no-stream

Options:

• -m, --model <name> - LLM model to use (default: qwen2.5-coder:7b)
• --session <name> - Load previous session
• --auto-approve - Auto-approve read-only tools
• --max-iterations <n> - Max agent iterations (default: 10)
• --context-size <n> - Max context tokens (default: 32768)
• --stream - Enable streaming (default)
• --no-stream - Disable streaming responses
• --temperature <n> - Model temperature 0-2 (default: 0.7)

Slash Commands (in chat):

• /help - Show available commands
• /model - List available models
• /model <name> - Switch model (requires restart)
• /clear - Clear conversation history
• /compact - Compress old messages
• /save <name> - Save current session
• /load <name> - Load saved session
• /tools - List available tools
• /exit - Exit chat

baryodev connect [options]

Test connection to Ollama server and list available models.

# Test default server
baryodev connect

# Test custom server
baryodev connect --url http://192.168.1.100:11434

# With timeout
baryodev connect --timeout 10000

Options:

• --url <url> - Ollama server URL (default: http://localhost:11434)
• --timeout <ms> - Connection timeout (default: 5000)

Output:

✔ Connected to Ollama

Server Information:
  URL: http://localhost:11434
  Models: 3

Available Models:
  • qwen2.5-coder:7b
  • deepseek-coder:6.7b
  • llama3.2:3b

✓ Ollama is ready!

Model Management

baryodev model list [options]

List all available models with optional details.

# Simple list
baryodev model list

# Detailed info
baryodev model list --detailed

# Custom Ollama server
baryodev model list --url http://localhost:11434

Options:

• --detailed - Show context length and descriptions
• --url <url> - Ollama server URL

baryodev model info <name>

Show detailed information about a specific model.

baryodev model info qwen2.5-coder:7b

Output:

Model: qwen2.5-coder:7b

Description: Ollama model: qwen2.5-coder:7b
Context Length: 32,768 tokens

ℹ For more details, use: ollama show qwen2.5-coder:7b

baryodev model default <name> [options]

Set default model in configuration.

# Set local default
baryodev model default qwen2.5-coder:7b

# Set global default
baryodev model default llama3.2:3b --global

Options:

• -g, --global - Set in global config (affects all projects)

baryodev model pull <name>

Guide for pulling a model from Ollama.

baryodev model pull codellama:13b

Output:

ℹ To pull a model, use the Ollama CLI:

  ollama pull codellama:13b

Then run this command again to verify

baryodev model remove <name>

Guide for removing a model.

baryodev model remove old-model:7b

Session Management

baryodev session list [options]

List all saved chat sessions.

# Simple list
baryodev session list

# Detailed view
baryodev session list --verbose

Options:

• --verbose - Show message counts, models, dates

Output:

Saved Sessions:

• feature-implementation
  Model: qwen2.5-coder:7b
  Messages: 42
  Created: 2026-01-03 10:30 AM

• bug-fix-session
  Model: deepseek-coder:6.7b
  Messages: 18
  Created: 2026-01-02 3:15 PM

ℹ Total: 2 sessions

baryodev session show <name> [options]

Show details of a saved session.

# Basic info
baryodev session show feature-implementation

# Full message history
baryodev session show feature-implementation --verbose

baryodev session rename <old> <new>

Rename a saved session.

baryodev session rename old-name new-name

baryodev session delete <name> [options]

Delete a saved session.

baryodev session delete unwanted-session --force

Options:

• --force - Confirm deletion (required for safety)

baryodev session export <name> [options]

Export session to file.

# Export as JSON
baryodev session export my-session --format json

# Export as Markdown
baryodev session export my-session --format md --output docs/chat.md

# Export as plain text
baryodev session export my-session --format txt

Options:

• --format <fmt> - Export format: json, md, txt (default: json)
• --output <path> - Output file path (default: .)

Configuration Management

baryodev config show [key]

Display configuration.

# Show all config
baryodev config show

# Show specific value
baryodev config show ollama.defaultModel

# Show nested object
baryodev config show ui

Output:

{
  "ollama": {
    "baseUrl": "http://localhost:11434",
    "defaultModel": "qwen2.5-coder:7b",
    "timeout": 120000
  },
  "safety": {
    "autoApproveReadOnly": true,
    "blockedCommands": ["^rm\\s+-rf\\s+/", "^sudo\\s+rm"],
    "maxExecTimeout": 30000
  },
  "ui": {
    "theme": "auto",
    "showTokenCount": true,
    "streamResponses": true,
    "useColors": true
  }
}

baryodev config set <key> <value> [options]

Set configuration value.

# Set local config
baryodev config set agent.maxIterations 20

# Set global config
baryodev config set ollama.defaultModel llama3.2:3b --global

Options:

• -g, --global - Set in global config
• -l, --local - Set in local config (default)

baryodev config unset <key>

Remove configuration value.

baryodev config unset test.value

baryodev config edit [options]

Open config in default editor.

# Edit local config
baryodev config edit

# Edit global config
baryodev config edit --global

baryodev config reset [options]

Reset configuration to defaults.

# Reset local config
baryodev config reset --local

# Reset global config
baryodev config reset --global

Tools Management

baryodev tools list

List all available tools.

baryodev tools list

Output:

No tools registered yet

ℹ Tools will be available when running the agent

baryodev tools info <name>

Show detailed tool information.

baryodev tools info read_file

Output:

Tool: read_file

Description: Read contents of a file

Parameters:

  • path (required)
    Type: string
    File path to read

Risk Level: READ-ONLY
Requires Approval: No

baryodev tools test <name> [options]

Test tool execution with parameters.

baryodev tools test read_file --params '{"path":"test.txt"}'

baryodev tools enable <name>

Enable a tool (makes it available to the agent).

baryodev tools enable read_file

baryodev tools disable <name>

Disable a tool.

baryodev tools disable execute_command

Project Initialization

baryodev init [options]

Initialize BARYO.md for current project.

# Interactive initialization
baryodev init

# Minimal setup
baryodev init --minimal

# Force overwrite existing
baryodev init --force

Options:

• --minimal - Create minimal BARYO.md
• --force - Overwrite existing BARYO.md

Output:

✔ Analyzing project structure...

✓ Project initialized successfully!

Framework: react
BARYO.md: /Users/me/project/BARYO.md
Config dir: /Users/me/project/.baryodev

Next steps:
1. Review and customize BARYO.md
2. Add important files to the "Important Files" section
3. Define project-specific conventions
4. Run baryodev chat to start coding with AI

Help System

baryodev help [topic]

Show help for specific topic.

# List all topics
baryodev help

# Specific topic
baryodev help getting-started
baryodev help models
baryodev help context

Available Topics:

• getting-started
• models
• context
• sessions
• baryo
• tools
• config
• troubleshooting
• slash-commands

Use Cases

1. Feature Development

$ baryodev chat

> Create a REST API with Express.js for managing user profiles.
> It should have endpoints for CRUD operations.

💭 Agent analyzes requirements...

✅ I'll create a REST API for you. Here's my plan:
   1. Set up Express.js server
   2. Create user model
   3. Implement CRUD endpoints
   4. Add input validation

🔧 Creating package.json...
✔ File created

🔧 Creating server.js...
✔ File created

🔧 Creating routes/users.js...
✔ File created

✅ Done! Your REST API is ready.
   - Run: npm install && node server.js
   - API will be at http://localhost:3000

2. Bug Fixing

> There's a bug in calculateTotal function - it returns NaN for empty arrays

💭 Let me read the file and investigate...

🔧 Reading src/utils/calculator.js...
✔ File read

💡 Found the issue: The reduce function doesn't handle empty arrays.

🔧 Updating src/utils/calculator.js...
✔ File updated

✅ Fixed! Added default value of 0 to reduce function.
   The function now correctly returns 0 for empty arrays.

3. Code Review

> Review the authentication middleware and suggest improvements

💭 Analyzing code...

🔧 Reading middleware/auth.js...
✔ File read

📋 Security Review:

✅ Good practices:
   - Using bcrypt for password hashing
   - JWT token validation

⚠️ Issues found:
   1. No rate limiting on login attempts
   2. Token expiry not checked
   3. Missing CORS configuration

💡 Recommendations:
   1. Add express-rate-limit for login endpoint
   2. Verify token expiration in middleware
   3. Configure CORS for production

Would you like me to implement these fixes?

4. Test Generation

> Write comprehensive tests for the user service

💭 Analyzing user service...

🔧 Reading src/services/userService.js...
✔ File read

🔧 Creating __tests__/userService.test.js...
✔ File created with 15 test cases

✅ Created tests for:
   - User creation (5 tests)
   - User retrieval (4 tests)
   - User update (3 tests)
   - User deletion (3 tests)

Run: npm test

Configuration

Configuration File Locations

Global: ~/.config/baryodev/config.json Local: ./.baryodev/config.json

Precedence: Local overrides Global

Configuration Options

interface BaryoDevConfig {
    // Ollama configuration
    ollama: {
        baseUrl: string;           // Default: "http://localhost:11434"
        defaultModel: string;      // Default: "qwen2.5-coder:7b"
        timeout: number;           // Default: 120000 (ms)
    };

    // Safety configuration
    safety: {
        autoApproveReadOnly: boolean;    // Default: true
        blockedCommands: string[];       // Regex patterns
        maxExecTimeout: number;          // Default: 30000 (ms)
    };

    // UI configuration
    ui: {
        theme: "auto" | "light" | "dark";  // Default: "auto"
        showTokenCount: boolean;            // Default: true
        streamResponses: boolean;           // Default: true
        useColors: boolean;                 // Default: true
    };

    // Context management
    context: {
        maxTokens: number;           // Default: 32768
        evictionThreshold: number;   // Default: 0.8 (80%)
        autoCompact: boolean;        // Default: true
        preserveRecentExchanges: number;  // Default: 3
    };

    // Agent configuration
    agent: {
        maxIterations: number;       // Default: 10
        thinkingEnabled: boolean;    // Default: true
        verboseLogging: boolean;     // Default: false
    };

    // Session configuration
    session: {
        autoSaveInterval: number;    // Default: 300000 (5 min)
        maxSessions: number;         // Default: 100
        compressionEnabled: boolean; // Default: true
    };
}

Known Issues & Limitations

Current Limitations

1. Interactive Chat Testing Pending

   • Status: Beta
   • Impact: Some advanced chat features need more testing
   • Workaround: Basic chat works perfectly
   • Fix: E2E testing framework planned for v0.2.0

2. Single Model Per Session

   • Status: By Design
   • Impact: Can't switch models mid-conversation
   • Workaround: Use /model <name> and restart chat
   • Fix: Multi-model support planned for Phase 6

3. Tool Registration at Runtime

   • Status: Working, needs docs
   • Impact: Custom tools require code changes
   • Workaround: Edit tool registry manually
   • Fix: Plugin system planned for Phase 8

Known Bugs

None reported! 🎉

Found a bug? Open an issue

Performance

Benchmarks (macOS, M1 Pro, Node 22)

| Command | Time | Target | Status | |---------|------|--------|--------| | --version | 33ms | 500ms | ✅ 15x faster | | --help | 33ms | 1000ms | ✅ 30x faster | | config show | 120ms | 1000ms | ✅ 8x faster | | model list | 34ms | 3000ms | ✅ 88x faster | | connect | 350ms | 5000ms | ✅ 14x faster |

All commands execute well under performance targets!

Resource Usage

• Memory: ~25-50MB baseline
• CPU: Minimal when idle
• Disk: <1MB for CLI, sessions vary
• Network: Only to Ollama (localhost)

Troubleshooting

"Failed to connect to Ollama"

Cause: Ollama server not running

Solution:

# Check if Ollama is running
curl http://localhost:11434/api/tags

# Start Ollama
ollama serve

# Verify in another terminal
baryodev connect

"Model not found"

Cause: Model not pulled yet

Solution:

# List available models
ollama list

# Pull missing model
ollama pull qwen2.5-coder:7b

# Verify
baryodev model list

"Context filling up too quickly"

Cause: Long conversation or large files

Solution:

# In chat, use compaction
/compact

# Or reduce context size
baryodev chat --max-tokens 16384

# Or adjust config
baryodev config set context.maxTokens 16384

"Command not found: baryodev"

Cause: Not globally installed

Solution:

# Run directly
node packages/cli/dist/index.js <command>

# Or link globally
cd packages/cli
npm link

# Then use
baryodev <command>

Development

Build & Test

# Install dependencies
pnpm install

# Build CLI
pnpm --filter @baryodev/cli build

# Watch mode (auto-rebuild)
pnpm --filter @baryodev/cli build:watch

# Run tests
pnpm --filter @baryodev/cli test

# Run directly without building
pnpm --filter @baryodev/cli dev chat

Project Structure

packages/cli/
├── src/
│   ├── commands/          # Command implementations
│   │   ├── chat.ts       # Interactive chat
│   │   ├── connect.ts    # Connection testing
│   │   ├── config.ts     # Config management
│   │   ├── init.ts       # BARYO.md initialization
│   │   ├── model-subcommands.ts     # Model management
│   │   ├── session-subcommands.ts   # Session management
│   │   └── tools-subcommands.ts     # Tools management
│   ├── ui/               # UI components
│   │   ├── colors.ts     # Color utilities
│   │   ├── prompt.ts     # Prompt formatting
│   │   ├── spinner.ts    # Loading indicators
│   │   ├── help.ts       # Help system
│   │   └── error-display.ts  # Error formatting
│   ├── config/           # Configuration
│   │   ├── types.ts      # Config types
│   │   └── manager.ts    # Config CRUD
│   └── index.ts          # Main entry point
├── dist/                 # Built JavaScript
├── __tests__/            # Test files
└── README.md            # This file

Future Roadmap

v0.2.0 (Q1 2025)

• ✅ Complete E2E testing for chat
• 📋 Slash command improvements
• 📋 Enhanced session management
• 📋 Tool approval UI improvements

v0.3.0 (Q2 2025) - Multi-Model Support

• 📋 Switch models mid-conversation
• 📋 Model comparison mode
• 📋 Cloud provider integration (Claude, GPT-4, Gemini)
• 📋 Smart routing (cost/speed/quality)

v0.4.0 (Q2 2025) - Advanced Features

• 📋 Skills system (custom agent behaviors)
• 📋 Codebase indexing (AST parsing)
• 📋 Multi-file editing (atomic operations)
• 📋 Agentic workflows

v0.5.0 (Q3 2025) - Integration

• 📋 VS Code extension
• 📋 JetBrains plugin
• 📋 GitHub Actions integration
• 📋 CI/CD pipelines

v1.0.0 (Q4 2025) - Production Ready

• 📋 Plugin ecosystem
• 📋 Custom tool marketplace
• 📋 Team collaboration features
• 📋 Enterprise features

See ROADMAP.md for complete roadmap.

Security

Security Features

• ✅ Command injection prevention - Blocked dangerous patterns
• ✅ Path traversal protection - Sandboxed file operations
• ✅ BARYO.md protection - Enforced file restrictions
• ✅ Audit logging - All operations logged
• ✅ Local-first - No data sent to cloud

Security Best Practices

1. Review BARYO.md - Define protected files
2. Use safety config - Block dangerous commands
3. Review tool approvals - Check before executing
4. Keep Ollama updated - Latest security patches
5. Audit logs - Review .baryodev/audit.log

Security issue? Report to [email protected]

Contributing

We welcome contributions! See CONTRIBUTING.md for:

• Code style guide
• Testing requirements
• Pull request process
• Development workflow

License

GPL-3.0-or-later

Copyright © 2024-2025 BaryoDev Contributors

See Also

• Core Package - ReAct engine implementation
• Shared Types - Type definitions
• Parser - Tool call parsing
• Ollama - Local LLM runtime
• Documentation - Full documentation

Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions
• Documentation: docs/
• Examples: examples/

Made with ❤️ by the BaryoDev community

Local AI, global impact 🌍