NevAI

A Discord bot that lets admins set an AI channel per server, mention it and it responds using free AI models and Exa.ai (optional), remembering the last 15 messages for context.

Features

• Only the Discord server administrator can use /ai, /ai_settings, /language, and /logs.
• The bot responds using the AI models listed below.
• The bot supports vision models (image understanding).
• Responses are limited to 1500 characters.
• Pre-built Docker image available.
• Custom success and error emoji.
• The bot reacts with an emoji when responding in a Discord chat.
• The server administrator can set a custom AI prompt.
• Multi-language support (add your own language).
• The bot saves the last 15 messages for chat context.
• Live logs support (/logs command).
• The bot can use custom Discord server emoji.
• Responds using Exa.ai.
• Generate infographic (only Anthropic models).
• Running local models using Ollama.
• Generate and edits images.
• Manage AI providers and tools from /ai_settings

How to use?

1. Add bot and set the AI channel using the /ai command.
2. Mention the bot, then ask your question.

3. You can customize the bot's options using administrator commands (server administrator only).

AI models list (free models)

Gemini

gemini-2.5-flash
gemini-2.5-flash-lite

Groq

openai/gpt-oss-120b 
openai/gpt-oss-20b

HackClub

openai/gpt-6-luna

Openrouter

Free models (openrouter/free)

Warning

The model list may change at any time. Free models can be renamed, limited or removed by their providers.

AI models list (paid models)

Anthropic

Claude Haiku 5.5

Generate image AI (paid model)

OpenRouter

google/gemini-3.1-flash-lite-image

Gemini

gemini-3.1-flash-lite-image

---

Ollama (local models)

Ollama is detected automatically. Make sure it is running and has at least one chat model downloaded (e.g. ollama pull gemma3:12b). By default, the bot looks for Ollama at http://127.0.0.1:11434 if no address is set in .env under OLLAMA_URL=. If no AI model is set under OLLAMA_MODEL= in .env, the first detected model is used. Embedding models are skipped.

Warning

Internet search and infographic generation are not available when using Ollama.

Models web search (Exa.ai)

The model decides for itself whether it wants to use Exa.ai. When the model uses Exa.ai, information from the internet is returned. The model uses this information to respond in the Discord chat. When the model decides it doesn’t need to use Exa.ai, it won’t. If you don’t add an EXA_API token, the bot will continue to function, but it will not search for information on the web. Exa returns images too.

How infographic generation works (Required Anthropic Api and Exa.ai)

flowchart LR
    A["User asks<br/>on Discord"] --> B["Claude calls<br/>create_infographic"]
    B --> C["Exa returns 5 pages:<br/>title, URL, text,<br/>image links"]
    C --> D["Claude picks portrait<br/>& draws SVG from facts"]
    D --> E["Bot renders<br/>SVG → PNG"]
    E --> F["PNG sent<br/>to Discord"]

    classDef discord fill:#5865F2,stroke:#3C45A5,color:#FFFFFF
    classDef claude fill:#D97757,stroke:#A0522D,color:#FFFFFF
    classDef exa fill:#1E88E5,stroke:#0D47A1,color:#FFFFFF
    classDef bot fill:#43A047,stroke:#1B5E20,color:#FFFFFF

    class A,F discord
    class B,D claude
    class C exa
    class E bot

Loading

If both ANTHROPIC_API and EXA_API are set in .env, every provider except Gemini, Ollama and vision models gets access to the create_infographic tool. The tool generates custom infographics based on the data returned by Exa. The infographic itself is always drawn by an Anthropic model (tested with Haiku 5.5).

For example, if you are currently using the Hack Club provider, its model (as long as it supports tools) can request an infographic. The bot then searches Exa for facts and images, and the Anthropic model picks the character portrait and draws the infographic. The finished image is attached to the reply on Discord, and the main model only receives a confirmation that the infographic was created, so it can add a short message.

Image generation (OpenRouter or Google Gemini, paid)

The generate_image tool generates new images and edits images sent by the user. The request is sent to the model defined in providers_list.js. To use image generation, you need at least one of the following API keys set in .env under GENERATE_IMAGE_API=. The provider is detected automatically from the key you paste:

• OpenRouter (key starts with sk-or-)
• Google Gemini (key starts with AIza)

The models used for each provider are defined in providers_list.js:

image: {
    openrouter: 'google/gemini-3.1-flash-lite-image', // Paid
    google: 'gemini-3.1-flash-lite-image',            // Paid
},

Warning

Image generation is a paid feature. Every generated or edited image is charged to the API key owner. Set a spending limit on your key before enabling it.

Administrator commands

Commands	Description
/ai	Set the channel where the AI will respond to mentions.
/ai_settings	Open the settings panel to customize the AI prompt, AI providers and reaction emoji.
/language	Change the bot's response language.
/logs	Set the channel where AI response logs will be sent.
/help	List of commands in embed format

Manage AI providers (using /ai_settings)

Managing AI providers lets you set which provider answers first by changing the order in the list. You can also enable or disable each provider. If a provider's API has not been added to .env, the ⚠️ Not configured label is shown. In the AI provider settings you can also enable or disable web search with Exa, infographic generation (which uses Exa and Claude models), and image generation or editing with Gemini models.

Requirements

At least one of the listed AI providers:

Groq | Gemini | HackClub | OpenRouter | Anthropic | Ollama (Ollama running locally, no API key needed)

You'll also need a Discord Application to get your bot token and optionally an Exa.ai API key (so the models can search for information on the internet).

Env

GROQ_API=
GEMINI_API=
HACKCLUB_API=
OPENROUTER_API=
EXA_API=
DISCORD_API=
ANTHROPIC_API=
DEBUG_MODE= #true/false
OLLAMA_MODEL= # By default, the first model returned by Ollama
OLLAMA_URL= # Default http://127.0.0.1:11434
GENERATE_IMAGE_API= # Paid google model

Note

DEBUG_MODE adds an advanced logging option to the debug.log file.

Running

Docker with install.sh

curl -sSL https://raw.githubusercontent.com/Nahida4479/NevAI-Bot/main/install/install.sh | bash

Docker manual (no automatic updates)

mkdir NevAI
cd NevAI
curl -fsSL -o .env https://raw.githubusercontent.com/Nahida4479/NevAI-Bot/main/example/env.example 
nano .env
chmod 600 .env
touch data.json
docker run -d --env-file .env -v $(pwd)/data.json:/NevAI/data.json --network host --name nevai --restart unless-stopped ghcr.io/nahida4479/nevai:latest

License

Released under the MIT License © 2026 Nahida4479. Free to use, modify and share, as long as the license notice is kept. Provided "as is", without warranty.