A command-line tool exactly like Claude Code, powered by DeepSeek's API. Everything happens in interactive chat mode - analyze code, search web, run commands, read files - all in one conversational interface.

Major Feature Release - Deep Code now includes enterprise-grade features matching Claude Code:

• 🛠️ Structured Tool System - 8 dedicated tools: Read, Write, Edit, Glob, Grep, Bash, WebSearch, WebFetch
• 🔒 Advanced Security - Command validation, dangerous operation blocking, sensitive file protection
• 📊 Context Management - Token counting, smart truncation, conversation optimization
• ✏️ Precise Edit Tool - Exact string matching for surgical code changes
• 🔍 Glob & Grep Tools - Find files by pattern, search code with regex
• ✅ Comprehensive Tests - Full test suite with 50+ tests
• 📚 Enhanced Documentation - See FEATURES.md and QUICK_REFERENCE.md

See CHANGELOG.md for full details.

• 💬 Interactive Chat Mode - Everything happens here! (Default mode)
• 🤖 Auto Tool Detection - Automatically reads files, executes commands, searches web when you ask
• 🔍 Automatic Directory Analysis - Current directory context loaded automatically
• 📄 Smart File Reading - Just mention a file and it reads it
• 🌐 Web Search Integration - Search the web with @web or naturally in conversation
• 🌍 HTTP Requests - Make API calls with @curl or mention URLs
• 🖥️ Bash Execution - Run commands naturally: "run git status" or "@bash git status"
• 📚 Session Management - Conversations persist automatically
• 🔄 Pipe Support - cat file | deepcode -p for quick queries
• ⚡ Streaming Responses - Real-time AI responses
• 🎨 Beautiful Terminal UI - Rich formatting and syntax highlighting

# 1. Install dependencies
pip install -r requirements.txt

# 2. Install the package (installs 'deepcode' command globally)
pip install -e .

# 3. Set your API key
export DEEPSEEK_API_KEY=your_api_key_here

# Just install dependencies
pip install -r requirements.txt

# Set API key
export DEEPSEEK_API_KEY=your_api_key_here

# Run directly
python3 deepcode.py

If you get "No module named 'setuptools'" error:

pip install setuptools wheel

If you get "deepcode: command not found":

• Make sure you ran pip install -e .
• Or use python3 deepcode.py instead

Configuration via .env file: Create a .env file in the project directory:

DEEPSEEK_API_KEY=your_api_key_here
DEEPSEEK_API_BASE=https://api.deepseek.com
DEEPSEEK_MODEL=deepseek-chat

See INSTALL.md for detailed installation instructions.

# Start interactive chat - automatically loads current directory context
deepcode

# Start chat with initial question
deepcode "explain this project"

# With additional directories
deepcode --add-dir ../app --add-dir ../lib

# Everything is accessible in chat:
# - "Read app.py" → Automatically reads the file
# - "What does this project do?" → Uses directory context
# - "Run git status" → Executes the command
# - "Search for Python async best practices" → Searches web
# - "@curl https://api.github.com/users/octocat" → Makes HTTP request

# Query via SDK, then exit
deepcode -p "explain this function"

# Process piped content
cat file.txt | deepcode -p "explain"
cat log.txt | deepcode -p "analyze errors"

# JSON output
deepcode -p "query" --output-format json

# Continue most recent conversation
deepcode -c

# Continue with new query (non-interactive)
deepcode -c -p "check for type errors"

# Resume specific session by ID
deepcode -r "session-id" "query"

# Resume in interactive mode
deepcode -r "session-id"

# Automatically analyzes current directory
deepcode "what does this project do?"

# Add additional directories
deepcode --add-dir ../shared --add-dir ../utils "review codebase"

# Works in print mode too
deepcode -p --add-dir ../lib "analyze dependencies"

# Replace entire system prompt
deepcode --system-prompt "You are a Python expert" "review code"

# Append to default prompt
deepcode --append-system-prompt "Always use TypeScript" "generate component"

In interactive chat mode, you can ask naturally - tools are used automatically:

deepcode
> What does this project do?
> Read app.py
> Analyze the main function
> Run git status
> Search for latest Python async features
> Check https://api.github.com/repos/octocat/Hello-World
> Show me the README file

You can also use explicit commands:

• @web <query> or @search <query> - Perform web search
• @curl <url> - Make HTTP request
• @bash <command> or @exec <command> - Execute bash command

Examples in chat:

deepcode
> @web latest Python async features
> @bash git status
> @curl https://api.github.com/users/octocat

# Use specific model (default is deepseek-chat)
deepcode --model deepseek-chat "query"

# Override to use chat model
deepcode --model deepseek-chat "query"

# Interactive session
deepcode "explain this codebase"

# Quick query
deepcode -p "write a Python function to reverse a string"

# Analyze piped content
cat requirements.txt | deepcode -p "analyze dependencies"

# Continue previous conversation
deepcode -c

# Resume specific session
deepcode -r "abc123" "finish this PR"

# With custom prompt
deepcode --append-system-prompt "Use best practices" "review my code"

Sessions are automatically saved to ~/.deepcode/sessions.db. Each session is associated with a directory, allowing you to continue conversations contextually.

The tool uses DeepSeek API endpoints. Make sure you have a valid API key from DeepSeek.

Environment variables:

• DEEPSEEK_API_KEY - Your API key (required)
• DEEPSEEK_API_BASE - API base URL (default: https://api.deepseek.com)
• DEEPSEEK_MODEL - Default model (default: deepseek-chat - conversational chat model)

Note: The default model is deepseek-chat which is optimized for conversational interactions. You can override this by setting DEEPSEEK_MODEL or using the --model flag.

• FEATURES.md - Comprehensive guide to all features, tools, and usage examples
• QUICK_REFERENCE.md - Quick reference for common commands and patterns
• CHANGELOG.md - Version history and what's new
• INSTALL.md - Detailed installation instructions

Deep Code includes 8 powerful tools:

• Read, Write, Edit - File operations with precision
• Glob, Grep - Find and search files
• Bash - Execute commands safely
• WebSearch, WebFetch - Access web content

See FEATURES.md for detailed documentation.

Built-in security features:

• Dangerous command blocking
• Sensitive file protection
• System directory safeguards
• Permission management

• Token counting and optimization
• Smart message truncation
• Conversation summarization
• Memory-efficient operation

Run the comprehensive test suite:

python test_tools.py
python -m unittest test_tools -v