klyn-jarvis

v1.1.3

Published

a month ago

JARVIS AI Assistant — Intelligent voice-enabled AI with streaming search, natural speech synthesis, and a JARVIS-inspired interface

0High
0Medium
0Low

klyntech

KlynJarvis v1.1.0

JARVIS-inspired AI assistant with CLI, MCP, multi-provider LLM, parallel agents, and workflow automation

Node.js 18+ · TypeScript · Private License

What is KlynJarvis?

KlynJarvis is a JARVIS-inspired AI assistant featuring streaming chat, natural voice synthesis, and a cinematic dark interface. It offers dual interfaces — a Next.js web UI and a Commander.js CLI — backed by a 29-file agent subsystem with 25+ tools, parallel sub-agents, dynamic workflows, and MCP integration.

Features

| Category | Features | |---|---| | Web UI | Streaming SSE chat, Voice Engine v3 (prosody, emotional alignment), Wake word detection, Dark cinematic interface | | CLI | Interactive REPL, Single query mode, Web server mode, Configuration management | | MCP Server | 16 tools (code generation, analysis, debugging, refactoring), stdio transport, External tool integration | | Multi-Provider LLM | 30+ providers: OpenAI, Anthropic, Gemini, Mistral, Cohere, DeepSeek, Groq, Together, xAI, Qwen, Ollama, +20 more | | Agent System | 29-file subsystem, 25+ tools, Approval flow, Sandbox execution, Error recovery | | Parallel Agents | Sub-agent orchestration, Dependency graph resolution, 4 workflow templates | | Dynamic Workflows | Runtime script engine, Variable interpolation, 5 step types | | Permission Engine | Per-agent rulesets, Condition evaluation (path, command, time), Inheritance | | Event System | GlobalBus with middlewares, Event filtering, Replay, Performance monitoring | | Fork Sessions | Conversation branching, Session tree, 3 merge strategies | | Context Management | Token counting, Importance scoring, Automatic compression |

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                       KlynJarvis v1.1.0                         │
├────────────────────────────┬────────────────────────────────────┤
│        Web UI              │        CLI (Commander.js)          │
│   Next.js + React + Tailwind│  REPL / Chat / Serve / Config     │
├────────────────────────────┴────────────────────────────────────┤
│                    Agent System (29 files)                       │
│  ┌──────────┐ ┌───────────┐ ┌──────────┐ ┌──────────────────┐  │
│  │   Loop   │ │   Tools   │ │ Approval │ │   Permissions    │  │
│  │  (v3)    │ │  (25+)    │ │  Flow    │ │  (per-agent)     │  │
│  └──────────┘ └───────────┘ └──────────┘ └──────────────────┘  │
│  ┌───────────┐ ┌───────────┐ ┌──────────┐ ┌────────────────┐  │
│  │ Parallel  │ │ Workflows │ │  Events  │ │  Fork Sessions  │  │
│  │  Agents   │ │  Runtime  │ │ GlobalBus│ │  (branching)    │  │
│  └───────────┘ └───────────┘ └──────────┘ └────────────────┘  │
├─────────────────────────────────────────────────────────────────┤
│              MCP Server (16 tools, stdio transport)              │
├─────────────────────────────────────────────────────────────────┤
│   Multi-Provider LLM (30+ providers, unified abstraction)        │
│   OpenAI · Anthropic · Gemini · Mistral · Cohere · DeepSeek     │
│   Groq · Together · Fireworks · xAI · Qwen · Ollama · +20 more  │
└─────────────────────────────────────────────────────────────────┘

Tech Stack

| Layer | Technology | |---|---| | Framework | Next.js 16 (App Router, standalone) | | UI | React 19, Tailwind CSS 4, shadcn/ui (48 components) | | State | Zustand 5 with localStorage persistence | | Database | Prisma ORM (SQLite dev / PostgreSQL prod) | | Voice | Web Speech API (STT + TTS), Voice Engine v3 | | AI/LLM | 30+ providers: OpenAI, Anthropic, Gemini, Mistral, Cohere, DeepSeek, Groq, Together, xAI, Qwen, Ollama, +20 more | | CLI | Commander.js 12, tsx (dev), esbuild (bundle) | | MCP | @modelcontextprotocol/sdk | | Agent | 29-file subsystem, 25+ tools, Parallel orchestration |

Quick Start

Prerequisites

Node.js 18+
npm or bun

Installation

Global Install (Recommended)

npm install -g klyn-jarvis
klyn --help

From Source

git clone https://github.com/Klyntech/KlynJarvis.git
cd KlynJarvis
npm install
npx prisma generate
npx prisma db push
cp .env.example .env

Run Web UI

npm run dev

Open http://localhost:3000

Run CLI

npm run cli

Or build and install globally:

npm run build:cli:bundle
npm link
klyn

CLI Usage

Commands

| Command | Description | |---|---| | klyn | Start interactive REPL (default) | | klyn chat "query" | Single query mode with streaming | | klyn repl | Interactive chat session | | klyn serve | Start web server | | klyn config [key] [value] | Show/set configuration | | klyn tools | List available tools | | klyn version | Show version |

REPL Commands

| Command | Description | |---|---| | /exit | Exit the REPL | | /quit | Exit the REPL | | /help | Show help | | /clear | Clear conversation | | /tools | List tools | | /history | Show chat history | | /system <msg> | Set system prompt |

Supported Providers (30+)

KlynJarvis supports 30+ LLM providers through a unified provider abstraction. Most providers use the OpenAI-compatible API format, so adding new providers is trivial.

Tier 1: Major Providers

| Provider | Default Model | Env Variable | |---|---|---| | OpenAI | gpt-4o | OPENAI_API_KEY | | Anthropic | claude-sonnet-4-20250514 | ANTHROPIC_API_KEY | | Google Gemini | gemini-2.0-flash | GEMINI_API_KEY | | Mistral | mistral-large-latest | MISTRAL_API_KEY | | Cohere | command-r-plus | COHERE_API_KEY |

Tier 2: OpenAI-Compatible Aggregators

| Provider | Default Model | Env Variable | |---|---|---| | OpenRouter | openai/gpt-4o | OPENROUTER_API_KEY | | Together AI | meta-llama/Llama-3-70b-chat-hf | TOGETHER_API_KEY | | Fireworks AI | accounts/fireworks/models/llama-v3p1-70b-instruct | FIREWORKS_API_KEY | | Groq | llama-3.3-70b-versatile | GROQ_API_KEY | | DeepInfra | meta-llama/Meta-Llama-3.1-70B-Instruct | DEEPINFRA_API_KEY | | Novita AI | meta-llama/llama-3.1-70b-instruct | NOVITA_API_KEY |

Tier 3: Chinese AI Providers

| Provider | Default Model | Env Variable | |---|---|---| | DeepSeek | deepseek-chat | DEEPSEEK_API_KEY | | Moonshot (Kimi) | moonshot-v1-128k | MOONSHOT_API_KEY | | Zhipu (GLM) | glm-4 | ZHIPU_API_KEY | | Qwen (Alibaba) | qwen-max | QWEN_API_KEY | | Baichuan | Baichuan4 | BAICHUAN_API_KEY | | MiniMax | abab6.5s-chat | MINIMAX_API_KEY | | Yi (01.AI) | yi-large | YI_API_KEY | | StepFun | step-1-32k | STEPFUN_API_KEY | | SenseNova | nova-ptc-k5-pro | SENSENOVA_API_KEY |

Tier 4: Cloud Provider AI

| Provider | Default Model | Env Variable | |---|---|---| | Azure OpenAI | gpt-4 | AZURE_OPENAI_API_KEY | | Google Vertex AI | gemini-2.0-flash | GOOGLE_APPLICATION_CREDENTIALS | | AWS Bedrock | anthropic.claude-3-sonnet | AWS_ACCESS_KEY_ID | | Cloudflare Workers AI | @cf/meta/llama-3.1-8b-instruct | CLOUDFLARE_API_TOKEN |

Tier 5: Specialized Providers

| Provider | Default Model | Env Variable | |---|---|---| | Hugging Face | meta-llama/Meta-Llama-3.1-8B-Instruct | HUGGINGFACE_API_KEY | | xAI (Grok) | grok-2 | XAI_API_KEY | | Writer | palmyra-x-004 | WRITER_API_KEY | | Aleph Alpha | phamba-llama-3.1-8b | ALEPHALPHA_API_KEY | | Inflection | pi-4 | INFLECTION_API_KEY | | Voyage AI | voyage-3 | VOYAGE_API_KEY | | Nomic AI | nomic-embed-text | NOMIC_API_KEY |

Tier 6: Local / Self-Hosted

| Provider | Default Model | Endpoint | |---|---|---| | Ollama | llama3 | http://localhost:11434 | | LM Studio | default | http://localhost:1234 | | vLLM | default | http://localhost:8000 | | Text Generation WebUI | default | http://localhost:5000 | | KoboldCpp | default | http://localhost:5001 | | llamafile | default | http://localhost:8080 | | llama.cpp | default | http://localhost:8080 |

Provider Aliases

| Alias | Resolves To | |---|---| | gpt4, gpt | openai | | claude, sonnet, haiku, opus | anthropic | | geminiPro, palm | gemini | | mixtral | mistral | | command | cohere | | llama, codellama | ollama | | grok | xai | | palmyra | writer | | glm | zhipu | | qwenMax | qwen |

Configuration

# Set provider via CLI
klyn config --provider gemini --model gemini-2.0-flash

# Set provider via environment variable
export GEMINI_API_KEY=your-key-here
klyn

# Use OpenRouter for access to multiple providers
klyn config --provider openrouter
export OPENROUTER_API_KEY=your-key-here
klyn chat "Hello, which model are you?"

MCP Server

Available Tools

| Tool | Description | |---|---| | klynjarvis_chat | Chat with KlynJarvis | | generate_code | Generate code from description | | analyze_code | Analyze code for issues | | debug | Debug errors | | refactor | Refactor code | | explain | Explain code | | create_file | Create a file | | read_file | Read a file | | edit_file | Edit a file | | search_files | Search files | | run_command | Run shell command | | web_search | Search the web | | git_operations | Git operations | | database_query | Database queries | | test_generation | Generate tests | | documentation | Generate docs |

Agent System

The agent subsystem consists of 29 files providing:

Core Loop — Agentic iteration with tool execution and context management
25+ Tools — File operations, code analysis, web search, git, database
Approval Flow — Human-in-the-loop for sensitive operations
Sandbox — Safe code execution environment
Error Recovery — Automatic retry and fallback strategies
Parallel Agents — Sub-agent orchestration with dependency resolution
Workflows — Dynamic script engine with variable interpolation
Permissions — Per-agent rulesets with condition evaluation
Events — GlobalBus for decoupled communication
Fork Sessions — Conversation branching and merging

API Reference

| Endpoint | Method | Description | |---|---|---| | /api/chat | POST | Streaming chat (SSE) | | /api/chat/approve | POST | Tool approval | | /api/health | GET | Health check | | /api/providers | GET | List LLM providers | | /api/search | POST | Web search | | /api/sessions | GET/POST | Session management | | /api/tools | POST | Tool execution | | /api/voice/asr | POST | Speech-to-text | | /api/voice/tts | POST | Text-to-speech | | /api/zai/[...path] | ALL | Z-AI proxy |

Project Structure

src/
├── app/                    # Next.js App Router (14 files)
│   ├── page.tsx            # Main JARVIS UI
│   ├── layout.tsx          # Root layout
│   └── api/                # 10 API endpoints
├── components/
│   ├── jarvis/             # Custom JARVIS components (5)
│   └── ui/                 # shadcn/ui components (48)
├── hooks/                  # React hooks (4)
├── lib/
│   ├── agent/              # Agent subsystem (29 files)
│   ├── events/             # Event system
│   ├── context-manager.ts  # Context management
│   ├── store.ts            # Zustand store
│   ├── voice-engine.ts     # Voice Engine v3
│   └── ...                 # Other libraries
├── cli/                    # CLI application (5 files)
│   ├── index.ts            # Entry point
│   ├── config.ts           # Config loader
│   ├── session.ts          # Chat session
│   ├── ui.ts               # Terminal UI
│   └── providers/          # LLM providers (4)
└── mcp/                    # MCP integration (5 files)
    ├── server.ts           # MCP server
    ├── client.ts           # MCP client
    ├── tools.ts            # 16 tool definitions
    └── integration.ts      # Agent bridge

Deployment

Render (Primary)

Fork the repo
Create a Web Service on Render
Connect your GitHub repo
Render auto-detects render.yaml
Set environment variables
Deploy

Environment Variables

| Variable | Description | Required | |---|---|---| | ZAI_BASE_URL | Z-AI API base URL | Yes | | ZAI_API_KEY | Z-AI API key | Yes | | ZAI_CHAT_ID | Chat session ID | No | | ZAI_TOKEN | Auth token | No | | ZAI_USER_ID | User identifier | No | | DATABASE_URL | Database connection string | Yes | | GEMINI_API_KEY | Google Gemini API key | Optional | | OPENAI_API_KEY | OpenAI API key | Optional | | ANTHROPIC_API_KEY | Anthropic API key | Optional | | NODE_ENV | Environment mode | Yes | | PORT | Server port | No | | LLM_PROVIDER | Default LLM provider | No |