CONVERSATION MEETS CAPABILITY
Rael is an AI-powered Discord bot built for natural conversation, multi-model flexibility, and useful utilities like vision, web search, live stock data, and usage tracking.
Note
Rael v2 focuses on a leaner, faster, and more reliable experience with a complete internal overhaul.
- Migrated the entire project to TypeScript and Bun for improved performance and maintainability
- Introduced automatic model fallbacks to keep requests running even when a provider is unavailable
- Reworked the conversation context pipeline for more consistent responses
- Reduced the base system prompt to improve efficiency and lower token usage
- Restructured the project's file and folder organization for a cleaner, more maintainable codebase
- Removed personas and stock cards to simplify the overall experience
- Removed DevHub-specific prompt information, making Rael more general-purpose
- Added automatic ping suppression to prevent unintended user or role mentions
Rael uses a prefix based interface with two modes of interaction.
,for AI-first input mode$for command-first mode
Start your message with a comma to send it directly to the AI without needing a command. You can also simply ping the bot instead of using the prefix.
Example:
,what is a JavaScript promise?
,explain async/await in simple terms
Use this prefix for structured commands such as configuration, status checks, or utility actions.
Example:
$help
$ping
$usage
$stats
$leaderboard
The AI system supports both prefixes and tagging:
,what is event loop in Node.js?
$ai what is event loop in Node.js?
@Rael what is event loop in Node.js?
To reset context:
$resetctx
or
$resetai
or
$clearctx
Resetting clears conversation history and restores default behavior.
Rael can search the web when you ask it to do so, to provide up to date information when needed, rather than relying solely on its training data.
Search is powered by the TinyFish Search API. Create a key at agent.tinyfish.ai/api-keys and set TINYFISH_API_KEY in your .env — see .env.example for every variable Rael reads. If the key is missing, the rest of the bot still works and only the search tool reports that it is unconfigured.
Users can view their token usage at any time through a clean, image based visual breakdown rather than plain numbers.
$usage
$leaderboard
Rael is designed to:
- Provide fast and reliable access to AI-assisted conversation
- Support multiple models and personas without complicating the interface
- Extend usefulness beyond chat and search
- Maintain a minimal and predictable command system
Rael is licensed under GPL-3.0. See the full license in the LICENSE file