ZeyoYT/AiLama

a discord user installable app for ai response using ollama, JDA-5 and LangChain4J

AiLama

A User Installable Discord App, made using JDA-5 ( java discord api 5 ), LangChain4J, and using Ollama to provide Ai Response anywhere in discord

Please Read : Important Things to Consider

Table of Contents

• Features
• Prerequisite
• How-To
• Whitelisting Other People to use the App
• Important Things to Consider before Using the Project
• Limitations
• Run Using Docker
  • Using Docker Compose
  • Manual Method
• Build Docker Image Your self
• SearXNG Guide
  • Setting SearXNG
  • Adding Environment Variables to AiLama
• List of Commands

Features

• Simple To Use
• Web Search Capabilities using SearXNG
• Web Document for RAG Support using URL
• Function Calling on AI Command with DOT Notation ( example below )
• Normal Document Support
• can be used anywhere in discord where user installable apps are supported
• you can run you query by a different model on the fly using the model's field ( example below )

Prerequisite

Ollama :

Ollama is an open-source app that lets you run, create, and share large language models locally with a command-line interface

• Ollama 0.28 or Greater ( Download Here )
• Ai Model for Ollama ( Download Here ) [ llama3 Preferred ]

There are Two ways you can use AiLama, by using Docker Image or by Building the Project and Running it

Compile and Run

• Java 19 or Greater
• Gradle 7.6 or Greater ( Download Here )

Run without Compiling

• Docker ( Download Here )

Discord App

• A Discord Bot/App ( make New app from discord developer portal )

Optional Prerequisite

These are Optional Prerequisite, you can use the bot without them, but to get access to more features its recommended to install the following programs

Web Search

• SearXNG for Web Search ( Setup Here ) ( Guide by AiLama Here )

Image Generation

• Automatic1111 for Image Generation ( Download Here )

How-To

1. First Clone this Repo

https://github.com/ZeyoYT/AiLama.git

2. Change Directory

cd AiLama

3. Create Dot Env File ( check .env.example ) and Fill Details

mv ./.env.example ./.env

5. Build Project

./gradlew build

After Building the project, a new Jar File would be made in AiLama/build/shadow/, that jar file can now be used to run the bot using the following command

java -jar ./<jar file name here>.jar

How to Make a new Discord Installable App and Install it

1. Go to https://discord.com/developers/applications
2. Click 'New Application' button
3. Give it a Name and agree to TOS and Developer Policy
4. Go to Bot and Un-Select Public Bot and Enable Presence Intent & Server Members Intent
5. Click on Reset Token and Reset the bot's token
6. Copy the Shown Token and Paste it in the .env file
7. Save the changes
8. Go to Installation and Un-Select Guild Install, and Select User Install
9. in Install Link select Discord Provided Link, then copy the link and open it
10. Authorize the App

Whitelisting Other People to use the App

Using the Authorization Link, you can allow others to install the app, but to actually use the app you need to whitelist them

In the Dot Env File ( .env ), there is a field called WHITELISTED_USERS, it must be shown something like this

WHITELISTED_USERS='WHITELISTED_USER_ID_HERE, WHITELISTED_USER_ID_HERE, ...'

in this you need to add there User ID's inside a single quote seperated by coma ( , ), for Example

WHITELISTED_USERS='426802118683262976, 259214353931304963, 848561799132741652'

Note: After Authorization of app, they might need to restart discord for the app to show up, and the app must be running, or else the command wont be registerd

Important Things to Consider before Using the Project

1. This is a side project, i will try my best to maintain it
2. You Need to run the Bot to use the command and keep it running
3. Code Quality is Low because i am still new with Ai and Lang Chain
4. This project uses a unreleased feature of JDA-5 that Supports User Installable Apps
5. Because it's a unreleased feature, i didn't find any proper command framework, so a workaround manual command registration code was used ( will be fixed in future release )
6. The Ai does not have persistent memory, so it will forget the conversation after restart
7. It's my first time making a docker image, so there might be some issues with it
8. First Response might be slow because of the model loading

Limitations

1. The Response will vary from model to model
2. URL RAG might not work with smaller models that support less context window
3. Only supports Ollama right now
4. Memory of previous conversation resets on restart

Note: Dot Query ( use function to get accurate response ) does not support memory, so it will not work with the previous conversation

Run Using Docker

Docker Image is Available at Docker Hub

Using Docker Compose :-

Docker Compose File :

AiLama/docker-compose.yaml

Lines 1 to 20 in 0f4dbfa

	version: '3.8'
	services:
	app:
	image: ailama:latest
	environment:
	- TOKEN=<YOUR_BOT_TOKEN>
	- OLLAMA_URL=<YOUR_OLLAMA_URL>
	- OLLAMA_PORT=<YOUR_OLLAMA_PORT>
	- OLLAMA_MODEL=<YOUR_OLLAMA_MODEL>
	- OLLAMA_EMBEDDING_MODEL=<YOUR_OLLAMA_EMBEDDING_MODEL>
	- OLLAMA_CHAT_MEMORY_UPTO=<NUMBER_OF_MESSAGES_TO_REMEMBER>
	- DEV_ID=<YOUR_DISCORD_ID>
	# Optional Parameters :-
	# Write inside single quotes and the user ids separated by commas, if you dont want to use this feature, just write '' (empty string enclosed by single quotes)
	- OLLAMA_TIMEOUT_SECONDS=<OLLAMA_TIMEOUT_SECONDS>
	- WHITELISTED_USERS='<user_id1>,<user_id2>,<user_id3> ... <user_idN>'
	- SEARXNG_URL=<YOUR_SEARXNG_URL>
	- SEARXNG_PORT=<YOUR_SEARXNG_PORT>
	- SEARXNG_ENGINES=<YOUR_SEARXNG_ENGINES>

Run Command :

docker-compose up -d

Manual Method :-

Pull Command :

docker pull zeyoog/ailama

Run Command :

docker run -e "TOKEN=<bot token>" -e OLLAMA_URL=<ollama_host_url> -e OLLAMA_PORT=<ollama_port> -e OLLAMA_MODEL=<ollama_model> -e OLLAMA_EMBEDDING_MODEL=<ollama_embedding_model> -e USER_ID=<developer_id> -e "WHITELISTED_USERS='<whitelisted_user_id>, <whitelisted_user_id>'" ailama:latest

Build Docker Image Your self

After Building you can also use the docker compose to run the bot

1. Clone Repo and Change Directory :

git clone https://github.com/zeyoyt/AiLama.git

cd AiLama

2. Run Docker Build Command :

docker build -t ailama:latest .

3. Run Docker Container :

docker run -e "TOKEN=<bot token>" -e OLLAMA_URL=<ollama_host_url> -e OLLAMA_PORT=<ollama_port> -e OLLAMA_MODEL=<ollama_model> -e OLLAMA_EMBEDDING_MODEL=<ollama_embedding_model> -e USER_ID=<developer_id> -e "WHITELISTED_USERS='<whitelisted_user_id>, <whitelisted_user_id>'" ailama:latest

SearXNG Guide

WARNING :- Using SearXNG you are scrapping results of different search engines and it might get you blocked from that search engine as its against there TOS

Setting SearXNG :-

1. First Install SearXNG ( Setup Here )
2. After Setting SearXNG, Open the settings.yml file ( Guide Here )
3. Find Search: ( Guide Here )
4. Add json to Formats like :-

# In settings.yml

search:
  formats:
    - html
    - json

Adding Environment Variables to AiLama :-

1. Add the Following Environment Variables to the .env file

SEARXNG_URL=<SEARXNG_HOST_URL>
SEARXNG_PORT=<SEARXNG_PORT>
SEARXNG_ENGINES=<SEARXNG_ENGINES>

2. Restart the Bot and Now you can use the Web Search Feature

List of Commands :-

• /ai
• /web
• /document
• /image
• /image-generate
• /info
• /model
• /models
• /reset_session
• /seturl

For More Information on Commands and Usage, Check the Wiki