TokenRing Writer (tr-writer)

Overview

TokenRing Writer (tr-writer) is a unified platform for writing and managing news articles & blog posts. It provides a powerful interface where you can create, edit, and publish content, leveraging AI to assist with writing, editing, formatting, and publishing workflows.

Features

• Agent-based architecture: Multiple specialized AI agents for different content creation tasks (writer, editor, researcher, publisher).
• Interactive chat interface: Talk to agents in a conversational REPL for content creation assistance.
• Persistent content history: Sessions and content drafts are saved in a SQLite database.
• Command system: Issue commands prefixed with / to manage agents, content, and workflows.
• Publishing integration: Works directly with your content directory; can initialize with a TokenRing config.
• Multi-model support: Support for various AI models from different providers.
• Research capabilities: Built-in web search and Wikipedia integration for content research.
• File system integration: Direct integration with local and cloud file systems.

Available Agents

TokenRing Writer includes specialized AI agents for different content creation workflows:

• Content Writer: Expert content writer specializing in creating engaging, well-structured articles and blog posts. Excels at research, storytelling, and adapting writing style to different audiences.
• Managing Editor: Coordinates content creation by searching for trending news topics, evaluating newsworthiness, creating article assignments, and dispatching tasks to specialized writing agents.

Getting Started

Quick Start (NPM)

Run directly using npx without installation:

npx @tokenring-ai/writer --source ./path-to-your-content --initialize

Quick Start (Docker)

Pull and run from GitHub Container Registry:

docker pull ghcr.io/tokenring-ai/writer:latest
docker run -ti --net host $(env | grep '_KEY' | sed 's/^/-e /') -v ./path-to-your-content:/repo:rw ghcr.io/tokenring-ai/writer:latest

Prerequisites

• Bun (for local development)
• Git initialized content directory

Installation (local development)

1. Sync git submodules: This project uses git submodules that need to be initialized and updated before installation:

   git submodule update --init --recursive

2. Install dependencies: This project uses Bun as the package manager in a monorepo structure:

   bun install

3. Run the application: Use Bun to start the application:

bun src/tr-writer.ts --source ./path-to-your-content

Installation (As local docker container)

1. Sync git submodules: This project uses git submodules that need to be initialized and updated before installation:

   git submodule update --init --recursive

2. Build the docker container:

   # This command must be run in the root directory of the repo
   docker build -t token-ring/writer:latest -f docker/Dockerfile .

3. Run the docker container:

   docker run -ti --net host $(env | grep '_KEY' | sed 's/^/-e /') -v ./:/repo:rw token-ring/writer:latest

Container Registry

The Docker image is automatically built and published to GitHub Container Registry on version tags. Available tags:

• latest: Latest stable release
• v*.*.*: Specific version tags
• main: Latest build from main branch

docker pull ghcr.io/tokenring-ai/writer:latest

Initialization

To initialize your content directory with the necessary TokenRing configuration file, pass the --initialize flag after your content directory. This will initialize a new .tokenring directory in your project, which stores a writer-config.js config file for your project that you can customize, as well as a sqlite database which stores your content history.

tr-writer --source ./path-to-your-content --initialize

This copies a default .tokenring/writer-config.js into your content directory.

Chat and Commands

Once started, you enter the agent chat REPL:

• Type natural language queries to interact with AI agents for content creation.
• Use commands with / prefix, e.g., /help, /agent switch writer, /tools enable.
• Switch between different agents specialized for writing, editing, research, and publishing.
• Control commands: /quit to exit, /agent list to see available agents.

Commands

Some example commands:

• /help: Show available commands.
• /agent list: List available agents.
• /agent switch <name>: Switch to a specific agent.
• /tools list: Show available tools.
• /tools enable <pattern>: Enable specific tools.

Architecture

• CLI: Entry point with argument parsing and agent team initialization (tr-writer.ts).
• Agents: Specialized AI agents for different content creation tasks (writer, editor, researcher, publisher).
• Services: Core services for file system, web search, models, and database management.
• Configuration: Flexible configuration system supporting multiple models and services.

Data Persistence

Content data and sessions are stored in a SQLite database (writer-database.sqlite) managed through the Bun SQLite package.

Extensibility

The system supports:

• Custom agents: Define new agents with specific roles and capabilities.
• Multiple AI models: Support for various providers (OpenAI, Anthropic, Google, etc.).
• Service providers: Pluggable services for file systems, web search, and content publishing.
• Tool integration: Extensible tool system for agent capabilities.

Contributing

Contributions are welcome! Feel free to fork the repository and submit pull requests.

License

This project is MIT licensed

This README provides a high-level overview of the TokenRing Writer (tr-writer) functionality and usage. For detailed command usage, use /help inside the chat REPL.