LLM Chat via SSH

14 hours ago 1

Chat with AI over SSH.

English | 简体中文

# Replace username with your GitHub username ssh [email protected]

macOS: iTerm2, Ghostty
Linux: Awaiting your feedback
Windows: Awaiting your feedback

Backend: Node.js and SSH2
UI Framework: React and Ink
Database: PostgreSQL and Redis

We recommend using V.PS servers for Docker deployment.

Copy the contents of .env.example to a .env file and modify it according to the .env file description below.
Create a docker-compose.yml file with the following content. This is all you need to deploy SSH AI Chat. If you need to deploy PostgreSQL and Redis alongside it, refer to the docker-compose.dev.yml file.

services: ssh-ai-chat: image: ghcr.io/ccbikai/ssh-ai-chat ports: - 22:2222 volumes: - ./data:/app/data env_file: - .env mem_limit: 4g

Start with the docker compose up -d command.
Access using ssh username@host -p 22, making sure to replace the hostname and port number.

# Server name, optional, can be changed to your own domain SERVER_NAME=chat.aigc.ing # Whether it's a public server, required. If not configured, it defaults to private server and requires whitelist configuration PUBLIC_SERVER=false # Rate limiting settings, optional. TTL suffix is for time, LIMIT is for count. Strongly recommended for public servers RATE_LIMIT_TTL=3600 RATE_LIMIT_LIMIT=300 LOGIN_FAILED_TTL=600 LOGIN_FAILED_LIMIT=10 # Blacklist and whitelist, optional. Configure GitHub usernames separated by commas BLACK_LIST=alice WHITE_LIST=bob # Redis URL, optional. If not configured, it will use simulated Redis and data will be lost on restart REDIS_URL=redis://default:[email protected]:6379 # Database URL, optional. If not configured, it will use PGLite to store data in the /app/data directory DATABASE_URL=postgres://postgres:[email protected]:5432/ssh-ai-chat # Umami configuration, optional UMAMI_HOST=https://eu.umami.is UMAMI_SITE_ID=6bc6dd79-4672-44bc-91ea-938a6acb63a2 # System prompt, optional AI_MODEL_SYSTEM_PROMPT="You are an AI chat assistant that..." # Model list, **required**, separated by commas AI_MODELS="DeepSeek-V3,DeepSeek-R1,Gemini-2.5-Flash,Gemini-2.5-Pro" # Models that support chain of thought, use `<think>` tags to return reasoning chain. Optional, if not configured will display reasoning content AI_MODEL_REASONING_MODELS="DeepSeek-R1,Qwen3-8B" # System reasoning model, optional, used for generating conversation titles. Only one model can be configured. If not configured, uses the first model from the model list AI_SYSTEM_MODEL="Qwen3-8B" # Model configuration file, configures API call information for models in `AI_MODELS` and `AI_SYSTEM_MODEL` lists. # Name format: prefix `AI_MODEL_CONFIG_`, model name in all caps, `-` and `.` replaced with `_`. Conversion relationships will be shown in startup logs. # Value format: type, model ID, BaseURL, APIKey. API format must support OpenAI-compatible format. Type is currently unused AI_MODEL_CONFIG_GEMINI_2_5_FLASH=fast,gemini-2.5-flash,https://api.example.com/v1,sk-abc AI_MODEL_CONFIG_GEMINI_2_5_PRO=pro,gemini-2.5-pro,https://api.example.com/v1,sk-abc

# Install dependencies pnpm i # Develop CLI interface pnpm run dev:cli # Develop SSH Server pnpm run dev

Special thanks to V.PS for sponsoring our servers.

Read Entire Article