🤖 @Bott

A Discord bot, powered by Gemini.

Features

• Uses judgement on when to engage with server members, taking channel context into consideration.
• Views and discusses media posted in chat.
  • Supports reading JPEGs, PNGs, and most websites.
  • Experimental support for MP4s, GIFs, WAVs and MP3 files.
• Generates photos, movies, songs and essays as requested.
• Translates technical errors into user-friendly language, when appropriate.

Development

Note

Interested in contributing? See our Contribution Guide!

Getting started

Prerequisites

• Homebrew (https://brew.sh/)
• GCP Project (https://developers.google.com/workspace/guides/create-project)
• Discord Application (https://discord.com/developers/applications)

Instructions

1. Copy .env.example to .env.development:

cp .env.example .env.development

3. Get your GCP information and add it to .env.development.
4. Get your Discord information and add it to .env.development.
5. Set up the environment with deno task setup.
6. Start Bott with deno task start:dev.

Configuring Bott

Bott is configured via a series of environment variables.

Name	Description	Default
CONFIG_ASSESSMENT_MODEL	The AI model used for quality assessment of potential bot responses.	gemini-2.5-flash-lite
CONFIG_ASSESSMENT_SCORE_THRESHOLD	The minimum score (1-100) a potential response must achieve in quality assessments (e.g., novelty, relevance) to be sent.	70
CONFIG_ERROR_MODEL	The AI model used for generating user-friendly error messages.	gemini-2.5-flash
CONFIG_ESSAY_MODEL	The AI model used for generating essays and long-form text content.	gemini-2.5-pro
CONFIG_EVENTS_MODEL	The AI model used for generating responses to chat events and user messages.	gemini-2.5-flash
CONFIG_INPUT_EVENT_LIMIT	The maximum number of past chat events to include in the context for the AI model.	2000
CONFIG_INPUT_FILE_TOKEN_LIMIT	The maximum number of tokens to use for analyzing the content of input files (images, websites).	500000
CONFIG_MOVIE_MODEL	The AI model used for generating video content.	veo-3.0-generate-preview
CONFIG_PHOTO_MODEL	The AI model used for generating images.	imagen-4.0-generate-preview-06-06
CONFIG_RATE_LIMIT_IMAGES	The maximum number of images Bott can generate per month.	100
CONFIG_RATE_LIMIT_MUSIC	The maximum number of songs Bott can generate per month.	25
CONFIG_RATE_LIMIT_VIDEOS	The maximum number of videos Bott can generate per month.	10
CONFIG_SONG_MODEL	The AI model used for generating music and audio content.	lyria-002
DISCORD_TOKEN	The authentication token for your Discord bot application.	-
FILE_SYSTEM_ROOT	The root directory on the local file system for storing input and output files.	./fs_root
GOOGLE_ACCESS_TOKEN	An access token for authenticating with Google Cloud APIs (for local development).	-
GOOGLE_PROJECT_ID	The ID of your Google Cloud project.	-
GOOGLE_PROJECT_LOCATION	The GCP region where your Vertex AI resources are located.	-
LOG_TOPICS	Controls which log topics to display. Comma-separated list of topics: debug, info, warn, error, perf.	info,warn,error
PORT	The port of the health check server required for GCP Cloud Run.	8080

Deploying Bott

(TODO: need to enable the correct APIs)

Deploying Bott to Google Cloud Run involves two main steps: deploying the service itself, then configuring the necessary permissions.

1. Deploy the Service

Click this button to deploy the Bott service:

2. Configure Permissions (TODO: confirm this)

1. Navigate to the IAM & Admin page in your Google Cloud project.
2. Find the service account that was created for your new Cloud Run service.
3. Click the pencil icon to edit its permissions.
4. Click + ADD ANOTHER ROLE and add the following two roles:
   • Vertex AI User (allows the bot to access Gemini models)
   • Storage Object Admin (allows the bot to read/write temporary files)
5. Click SAVE.

Bott should now be running correctly. You may need to trigger a new revision deployment from the Cloud Run console for the permission changes to take effect immediately.

High-level Architecture

Tip

Review the code data model annotated in ./model/types.ts.

graph TD
  subgraph "@bott/model"
    BottDiscord["**@bott/discord**"]
    
    subgraph App["./app"]
      BottAppLayer["**@bott/task**"]
      BottDataLayer["**@bott/storage**<br>Persistence layer"]
      
      BottAppLayer --> BottDataLayer
      BottDataLayer --> BottAppLayer
    end
    
    BottGemini["**@bott/gemini**"]
  end

  %% Flow from Discord User to Bot and back
  Discord -- "User Message" --> BottDiscord
  BottDiscord -- "BottEvent" --> App
  App -- "BottEvent" --> BottGemini
  BottGemini -- "Calls Gemini API" --> Gemini
  Gemini -- "Generated Data" --> BottGemini
  BottGemini -- "BottEvent" --> App
  App -- "BottEvent" --> BottDiscord
  BottDiscord -- "System Message" --> Discord

  style App fill:darkblue;

Loading

---

Licensing

This project is dual-licensed. This model allows for free, open-source use for non-commercial purposes while requiring a separate license for commercial applications.

• For Non-Commercial Use: This software is free and open-source under the terms of the GNU Affero General Public License v3.0 (AGPLv3).

  • Read the full AGPLv3 license details in the LICENSE file.

• For Commercial Use: Use of this software for any purpose that is intended for commercial advantage or monetary compensation requires a Proprietary Commercial License. Please contact [email protected] to discuss licensing terms.

Copyright (C) 2025 DanielLaCos.se