NanoGPT Discord Bot

A Discord bot powered by the NanoGPT API. It supports chat, memory, document context, image generation/editing, audio, video, embeddings, direct search, data extraction, detection, TEE metadata, and character workflows.

Features

• Slash-command AI chat with per-user and per-server model defaults
• Optional local conversation memory with /memory
• Reusable document context from PDF, TXT, Markdown, JSON, XML, CSV, HTML, and log files
• NanoGPT model catalogs for subscription, paid, canonical, personalized, image, video, audio, embedding, and character models
• Pay-as-you-go capable API access for supported NanoGPT endpoints
• Advanced json passthrough options on broad API commands for fast-moving NanoGPT fields
• Paginated JSON embeds for extraction, embedding, detection, and metadata responses, with an attachment fallback for extremely large payloads

Setup

1. Create Discord Application

1. Go to Discord Developer Portal
2. Create a new application
3. Go to "Bot" and create a bot
4. Copy the bot token
5. Go to "OAuth2" > "URL Generator"
   • Select scopes: bot, applications.commands
   • Select permissions: Send Messages, Use Slash Commands, Embed Links, Attach Files
6. Use the generated URL to invite the bot to your server

2. Get NanoGPT API Key

1. Sign up at NanoGPT
2. Generate an API key in settings
3. Add balance or subscribe to a plan depending on which endpoints you want to use

3. Configure Environment

Copy the example environment file and fill in your credentials:

cp .env.example .env

Minimum required .env:

DISCORD_TOKEN=your_discord_bot_token
DISCORD_CLIENT_ID=your_discord_application_id
NANOGPT_API_KEY=your_nanogpt_api_key
NANOGPT_BASE_URL=https://nano-gpt.com/api
SYSTEM_PROMPT=
DEFAULT_MODEL=moonshotai/kimi-latest
DATABASE_PATH=./data/bot.db

4. Deploy

# Build and start
docker compose up -d

# View logs
docker compose logs -f bot

# Register slash commands
docker compose exec bot bun run register

For local development:

bun install
bun run register
bun run dev

Commands

Command	Description
/help	Show available commands
/chat	Stateless chat completion with context, image input, web search suffixes, sampling, reasoning, provider, billing, suffix, and advanced JSON options
/memory chat	Chat with local per-user memory
/memory view	View recent local memory
/memory stats	Show memory statistics
/memory clear	Clear local memory
/models	List NanoGPT model catalogs by type, with optional detailed metadata
/setmodel	Set personal or server default model
/usage	Check NanoGPT subscription usage
/context add/list/view/remove	Manage reusable document contexts
/imagine	Backward-compatible image generation command
/scrape	Scrape up to five URLs with optional stealth mode and downloadable output
/responses create	Use NanoGPT's Responses API
/messages create	Use NanoGPT's Anthropic-compatible Messages API
/completion	Use legacy completions
/tokens count	Count tokens for a Messages API-style request
/extract firecrawl	Run Firecrawl scrape, map, or crawl
/extract maps	Extract Google Maps places with required result and cost caps
/extract maps-reviews	Extract Google Maps reviews with required review and cost caps
/extract facebook-ads	Extract Facebook ads from request JSON
/extract instagram-profile	Extract Instagram profile data from request JSON
/extract instagram-posts	Extract Instagram posts from request JSON
/extract reddit	Extract Reddit data from request JSON
/extract tiktok	Extract TikTok data from request JSON
/extract hunter	Use Hunter domain search, email finder, verifier, or discover
/image generate	Generate images through the image API
/image edit	Edit images with optional mask input
/image classify	Run NSFW image classification
/video generate/status/recover/extend/content	Use NanoGPT video generation and job utilities
/audio speech	Generate OpenAI-compatible speech/audio and return an audio file
/audio tts/tts-status	Use legacy async TTS/music endpoints
/audio transcribe/transcribe-status	Transcribe audio or check transcription status
/audio voice-clone	Create a voice clone from an audio sample
/search web	Run direct NanoGPT web search
/embed create	Create embeddings and return a JSON attachment
/detect ai	Run AI text detection
/detect plagiarism	Run plagiarism detection
/youtube transcribe	Transcribe a YouTube video
/tee attestation/signature	Fetch TEE verification metadata
/characters search/mine/view/create/edit/delete/review/report	Search and manage NanoGPT characters

Common Options

/chat

Option	Description
message	Required user message
context	Saved context name to inject into the system prompt
model	One-off model override
searchprovider	default, linkup, tavily, exa, kagi, brave, or valyu
searchvariant	Provider-specific variant such as deep, fast, auto, neural, web, or news
temperature, top_p, max_tokens	Sampling and output controls
reasoning	Reasoning effort, when supported by the model
provider	NanoGPT provider override header
billing	paygo or subscription
suffix	Exact model suffix override, for example :online/linkup-deep
json	Advanced NanoGPT chat request fields as a JSON object
image	Optional image attachment for multimodal models

/imagine

Option	Description
prompt	Required image prompt
model	Image model, with autocomplete
size	Any model-supported size string
count	Number of images, up to 10
guidance, steps, seed	Generation controls
image, strength	Image-to-image input and strength
mask	Mask image for edit/inpainting models
kontext_max_mode	Flux Kontext max mode
json	Advanced image request fields

Advanced JSON Passthrough

Most parity commands include a json option. This must be a JSON object and is merged into the request body or query before calling NanoGPT. Use it for fields that are model-specific, newly added in NanoGPT, or too large for Discord's slash-command option limits.

Example:

{"maxTotalChargeUsd":0.25,"includeReviews":true}

Data Extraction Safety

Extraction commands can incur pay-as-you-go costs. The Maps commands require explicit caps:

• /extract maps requires limit and max_cost
• /extract maps-reviews requires max_reviews and max_cost

Other extraction endpoints accept request JSON directly so you can pass NanoGPT's documented cost and result controls.

Feature Toggles

Feature toggles use this convention:

Value	Effect
false	Feature enabled for everyone
true	Feature disabled for everyone
admin	Feature available only to admin users or CONTEXT_ADMIN_USERS

Configured toggles:

• DISABLE_WEBSEARCH
• DISABLE_DEEPSEARCH
• DISABLE_IMAGEGEN
• DISABLE_SCRAPE
• DISABLE_PAYGO
• DISABLE_VIDEO
• DISABLE_AUDIO
• DISABLE_EMBEDDINGS
• DISABLE_DETECTION
• DISABLE_DIRECT_SEARCH
• DISABLE_ADVANCED_CHAT

Missing toggle variables default to false, which means enabled. DISABLE_PAYGO gates pay-as-you-go style commands and explicit billing:paygo chat requests; the broad media, embeddings, detection, direct search, and advanced chat commands also use their feature-specific toggles at command entry.

Document Context

Contexts can be personal or shared with the server:

• scope:user is personal to the user
• scope:server is shared with the server

Supported file types:

• PDF (.pdf)
• Plain text (.txt, .text)
• Markdown (.md, .markdown)
• Log files (.log)
• JSON (.json)
• XML (.xml)
• CSV (.csv)
• HTML (.html, .htm)

Notes

• Existing user or server defaults stored in SQLite override DEFAULT_MODEL.
• Register slash commands again after command changes with bun run register.
• Large JSON API results are paginated in Discord with Previous/Next buttons. Extremely large payloads are truncated in embeds and include a full JSON attachment fallback.