This repository contains the DeepSeek CLI, a command-line AI coding assistant that leverages the powerful DeepSeek Coder models to accelerate your development workflows and enhance your coding experience.
With the DeepSeek CLI you can:
- Code Completion & Generation: Get intelligent code suggestions and generate complete functions across 100+ programming languages.
- Repository-Level Understanding: Analyze and work with large codebases using DeepSeek's advanced code comprehension capabilities.
- Code Refactoring & Migration: Modernize legacy code, migrate between frameworks, and implement best practices.
- Debugging & Code Review: Get help identifying bugs, security issues, and code quality improvements.
- Project Scaffolding: Generate new applications, components, and boilerplate code from descriptions or requirements.
- Multi-Language Support: Work seamlessly across Python, JavaScript, Java, C++, Go, Rust, and 90+ other languages.
- 
Prerequisites: - Node.js version 18 or higher
- Ollama for local AI models
 
- 
Install Ollama: # macOS brew install ollama # Linux curl -fsSL https://ollama.ai/install.sh | sh # Windows: Download from https://ollama.ai 
- 
Install the CLI: npm install -g run-deepseek-cli 
- 
Setup Local Environment: # Start Ollama service ollama serve # Install DeepSeek Coder model (choose one) ollama pull deepseek-coder:6.7b # Recommended (4GB) ollama pull deepseek-coder:1.3b # Lightweight (1GB) ollama pull deepseek-coder:33b # Most capable (19GB) 
- 
Start Using: deepseek 
- 
Install the CLI: npm install -g run-deepseek-cli 
- 
Configure API Access: export DEEPSEEK_API_KEY="your_api_key_here" export DEEPSEEK_USE_LOCAL=false Get your API key from DeepSeek Platform. 
- 
Start Using: deepseek 
Once the CLI is running, you can start interacting with DeepSeek Coder from your shell.
cd new-project/
deepseek
> Create a FastAPI web application with user authentication and a PostgreSQL databasegit clone https://github.com/your-repo/existing-project
cd existing-project
deepseek
> Analyze this codebase and suggest performance optimizationsdeepseek
> Write a Python function to implement binary search with proper error handlingdeepseek
> Convert this JavaScript function to TypeScript with proper type annotationsStart by cding into an existing or newly-cloned repository and running deepseek.
> Explain the architecture of this application and identify the main components.
> What design patterns are used in this codebase?
> Find potential security vulnerabilities in this code.
> Implement a REST API for user management with CRUD operations.
> Refactor this class to follow SOLID principles.
> Add comprehensive error handling to this module.
> Help me migrate this React class component to hooks.
> Convert this Python 2.7 code to Python 3.10 with type hints.
> Upgrade this Express.js app to use async/await instead of callbacks.
> Review this pull request and suggest improvements.
> Add unit tests for all functions in this file.
> Optimize this algorithm for better time complexity.
> Create a React TypeScript project with Redux Toolkit and Material-UI.
> Set up a Django project with Docker, PostgreSQL, and Redis.
> Generate a microservices architecture using Node.js and Docker Compose.
| Variable | Description | Default | 
|---|---|---|
| DEEPSEEK_USE_LOCAL | Use local Ollama instead of cloud API | true | 
| DEEPSEEK_MODEL | Model to use | deepseek-coder:6.7b | 
| OLLAMA_HOST | Ollama server URL | http://localhost:11434 | 
| DEEPSEEK_API_KEY | DeepSeek API key (cloud mode only) | - | 
| Model | Size | Memory | Best For | 
|---|---|---|---|
| deepseek-coder:1.3b | 1GB | 2GB RAM | Quick completions, lightweight tasks | 
| deepseek-coder:6.7b | 4GB | 8GB RAM | General coding, recommended | 
| deepseek-coder:33b | 19GB | 32GB RAM | Complex analysis, best quality | 
Create a .env file in your project root:
# Local mode (default)
DEEPSEEK_USE_LOCAL=true
DEEPSEEK_MODEL=deepseek-coder:6.7b
OLLAMA_HOST=http://localhost:11434
# Cloud mode (optional)
# DEEPSEEK_USE_LOCAL=false
# DEEPSEEK_API_KEY=your_api_key_hereDeepSeek CLI supports intelligent code generation and completion for 100+ programming languages including:
Popular Languages:
Python, JavaScript, TypeScript, Java, C++, C#, Go, Rust, PHP, Ruby, Kotlin, Swift, Scala, R, Julia, Dart, HTML, CSS, SQL
Systems & Scripting:
Shell, PowerShell, Bash, Dockerfile, Makefile, YAML, JSON, XML
Specialized:
CUDA, Assembly, Verilog, VHDL, Solidity, Protocol Buffer, Thrift
For the complete list, see Supported Languages.
| Command | Description | 
|---|---|
| deepseek | Start interactive mode | 
| deepseek setup | Setup local Ollama environment | 
| deepseek chat "prompt" | Single prompt mode | 
| deepseek --local | Force local mode | 
| deepseek --model <model> | Specify model to use | 
| deepseek --help | Show help information | 
| deepseek --version | Show version | 
The setup command helps you install and configure the local environment:
deepseek setupThis will:
- Check if Ollama is installed
- Start Ollama service if needed
- Download the specified DeepSeek model
- Verify everything is working
The interactive mode provides a rich REPL experience:
- Syntax Highlighting: Color-coded responses for better readability
- Code Block Detection: Automatic language detection and formatting
- Session History: Navigate through previous commands
- File Context: Automatic inclusion of relevant project files
- Multi-turn Conversations: Maintain context across interactions
DeepSeek CLI automatically analyzes your repository structure and includes relevant context:
# Analyze entire repository
deepseek --include-all
> Suggest improvements for the overall code architecture
# Include specific files/directories
deepseek --include src/ tests/ README.md
> Update the documentation to reflect recent API changesGenerate common patterns and boilerplate:
> Generate a React component template with props validation
> Create a Python class with proper docstrings and type hints
> Set up a GitHub Actions workflow for CI/CD
> Analyze the git diff and write a commit message
> Generate API documentation from this OpenAPI spec
> Create database migration scripts from these model changes
npm install -g run-deepseek-cligit clone https://github.com/your-username/deepseek-cli.git
cd deepseek-cli
npm install
npm run build
npm linkdocker run -it -v $(pwd):/workspace -e DEEPSEEK_API_KEY=your_key deepseek/cliWe welcome contributions! Please see our Contributing Guide for details.
- Fork and clone the repository
- Install dependencies: npm install
- Run tests: npm test
- Start development server: npm run dev
npm run build
npm run package| Model | Speed | Context Window | Best For | 
|---|---|---|---|
| deepseek-coder-1.3b-instruct | Fastest | 16K tokens | Quick completions | 
| deepseek-coder-6.7b-instruct | Fast | 16K tokens | General coding tasks | 
| deepseek-coder-33b-instruct | Slower | 16K tokens | Complex analysis | 
This project leverages the DeepSeek Coder models:
- DeepSeek Coder Base: Pre-trained on 2T tokens (87% code, 13% natural language)
- DeepSeek Coder Instruct: Fine-tuned for instruction following and chat
- Multi-size Options: 1.3B, 6.7B, and 33B parameter models available
For more details about the models, visit the DeepSeek Coder GitHub repository.
API Key Issues:
export DEEPSEEK_API_KEY="your_actual_api_key"
# Verify with: echo $DEEPSEEK_API_KEYNode.js Version:
node --version  # Should be 18+
npm update -g deepseek-cliPermission Errors:
sudo npm install -g run-deepseek-cli
# Or use a Node version manager like nvmFor more troubleshooting help, see our Troubleshooting Guide.
This project is licensed under the MIT License. See LICENSE for details.
The use of DeepSeek Coder models is subject to the DeepSeek Model License.
If you use this tool in your research or projects, please cite:
@misc{deepseek-cli,
  title={DeepSeek CLI: Command-Line AI Coding Assistant},
  author={Your Name},
  year={2025},
  url={https://github.com/your-username/deepseek-cli}
}
@misc{deepseek-coder,
  author={Daya Guo, Qihao Zhu, Dejian Yang, Zhenda Xie, Kai Dong, Wentao Zhang, Guanting Chen, Xiao Bi, Y. Wu, Y.K. Li, Fuli Luo, Yingfei Xiong, Wenfeng Liang},
  title={DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence},
  journal={CoRR},
  volume={abs/2401.14196},
  year={2024},
  url={https://arxiv.org/abs/2401.14196}
}- Issues: GitHub Issues
- Discussions: GitHub Discussions
- DeepSeek Support: [email protected]
- Built on the foundation of Gemini CLI
- Powered by DeepSeek Coder models
- Thanks to the open-source community for contributions and feedback
