Screenshot

Cody Chat Shell in action

Overview

Cody Chat Shell is a Bash script that enhances the Cody CLI experience by creating a seamless interface between you and Sourcegraph's Cody AI assistant. This tool addresses several key needs for developers:

• Persistent Knowledge Management: Automatically logs all conversations to Markdown files, creating a searchable knowledge base of your AI interactions
• Contextual Continuity: Supports context files for better continuity between sessions
• Workflow Integration: Provides both interactive and non-interactive modes to fit into various development workflows
• Project-Specific Conversations: Allows directory-specific chats to maintain focus on relevant codebases
• Organized Documentation: Supports custom session naming for better organization of AI-assisted problem-solving

The primary purpose of this tool is to bridge the gap between ephemeral AI chat sessions and permanent documentation. By creating flat Markdown files for each conversation, you can:

1. Build a personal knowledge repository of solutions and explanations
2. Resume conversations by referencing previous chat logs
3. Share AI-generated solutions with team members
4. Track the evolution of your understanding and approach to problems
5. Maintain context across multiple development sessions

Installation

1. Install the Cody CLI:

   npm install -g @cody/cli

2. Clone this repository:

   git clone https://github.com/yourusername/cody-chat-shell.git

3. Make the script executable:

   chmod +x cody-chat-shell/cody-chat

4. Optionally, add the script to your PATH for easier access:

   ln -s "$(pwd)/cody-chat-shell/cody-chat" /usr/local/bin/chat

Authentication

You'll need a Cody API token to use this tool:

1. Authenticate through your company portal and complete the login flow for Cody
2. Generate a new API token in the portal
3. Set up your token in one of these ways:
   • Export as an environment variable: export CODY_AUTH_TOKEN="<your_token>"
   • Pass directly with each command: cody-chat --token "<your_token>" "Hello"

Usage

cody-chat [options] [initial_message]

Options

• -C, --dir <directory> - Change to this directory before starting chat
• --session <name> - Use a custom name for the session logfile
• --token <token> - Pass your API token explicitly for this session
• initial_message - Send this as the first message to Cody

Examples

Start an interactive chat session:

cody-chat

Ask a question and get a single response:

cody-chat "What is the capital of France?"

Use piped input:

echo "Explain this code:" | cody-chat

Chat about code in a specific directory:

cody-chat -C ~/projects/myapp "How can I optimize this code?"

Name your chat session:

cody-chat --session project-refactor "Let's discuss refactoring the auth module"

Chat Logs

By default, all conversations are logged to Markdown files in:

$HOME/dev/chat/log/

Files are named either:

• Using your custom session name: session-name.md
• With a timestamp: chat_YYYYMMDD_HHMMSS.md

Context Files

For better continuity, place a context.md file in your working directory. This file will be automatically passed to Cody to provide additional context for your conversation.

Resuming Sessions

One of the key benefits of this tool is the ability to resume previous conversations:

1. Start a new session with a meaningful name:

   cody-chat --session auth-refactor "Let's improve our authentication system"

2. Later, continue the same conversation by using the same session name:

   cody-chat --session auth-refactor "Now let's implement those changes we discussed"

The tool will append to the existing log file, maintaining the full conversation history.

Integration with Development Workflow

Cody Chat Shell is designed to integrate seamlessly with your development workflow:

• Problem Exploration: Use interactive mode to explore complex problems
• Quick Answers: Pipe in questions for immediate, non-interactive responses
• Code Reviews: Chat about specific directories to get contextual feedback
• Documentation: Generate explanations that are automatically saved for future reference
• Knowledge Transfer: Share chat logs with team members to transfer knowledge

License

MIT License