Deepseek Codex CLI

An interactive command-line tool for AI-assisted system design. It provides a REPL that loads prompt templates and technical specifications from a local prompts/ directory, communicates with an inference platform API (defaulting to DeepSeek V4), and guides the user through the iterative workflow.

Installation

npx @opensassi/deepseek-codex

Requires Node.js ≥ 24.15.0 LTS.

Usage

Set your DeepSeek API key:

export DEEPSEEK_API_KEY=your-api-key-here

Run the CLI:

npx @opensassi/deepseek-codex

REPL Commands

| Command | Description | | -------------------- | ---------------------------------------------------------------------------------------------- | | /help | Display all available commands | | /load [name] | Load a prompt file (system-design-agent, npm-project-creation-prompt). If no name, loads both. | | /spec | Print the current technical specification | | /chat <message> | Send a message to the AI and get a response | | /append-spec | Append the last assistant response to the technical specification | | /run <prompt-name> | Load a prompt, prepend the technical spec, and send to AI | | /exit | Exit the REPL |

Command-Line Options

| Option | Description | | ---------------------- | ------------------------------------------------ | | --prompts-dir <path> | Path to prompts directory (default: ./prompts) | | --provider <name> | Provider to use (default: deepseek) | | --base-url <url> | Custom base URL for the API | | --model <name> | Model name to use | | --api-key <key> | API key (overrides environment variable) |

Development

# Install dependencies
npm install

# Build
npm run build

# Test
npm test

# Lint
npm run lint

License

MIT