Skip to main content

Getting started

eLLMental is a toolset for building AI-powered applications written in Kotlin, and it offers a variety of components that you can use right away. In this guide, we will guide you through an example of how to use the SemanticSearch component to build an intelligent note taking application for the terminal that learns from your notes and is able to find relevant notes based on a query.

Step 0: Create a new Kotlin application project

This can be done using your preferred way of Kotlin project initialization. If you're using IntelliJ IDEA, you can follow the steps below:

  1. Open IntelliJ IDEA and click on Create New Project.
  2. Select New Project in the left panel and then select Kotlin and Gradle in the right panel. Choose Kotlin as the Build DSL.
  3. Click Create and then enter the project name and location.

Step 1: Add the eLLMental dependencies using JitPack

Import the eLLMental dependencies in your build.gradle.kts file.

You can use this one as a starting point:

import org.jetbrains.kotlin.gradle.tasks.KotlinCompile

plugins {
kotlin("jvm") version "1.9.0"

group = "com.theagilemonkeys"

version = "1.0-SNAPSHOT"

allprojects { repositories { maven { url = uri("") } } }

repositories { mavenCentral() }

dependencies {

tasks.test { useJUnitPlatform() }

tasks.withType<KotlinCompile> { kotlinOptions.jvmTarget = "17" }

tasks.withType<KotlinCompile>().configureEach {
kotlinOptions { freeCompilerArgs = freeCompilerArgs + "-Xcontext-receivers" }

tasks.named<JavaExec>("run") { standardInput = System.`in` }

application { mainClass.set("MainKt") }

Step 2: Creating the Note class

Let's create a Note class. This class will have two properties: id type Int for the note identifier and content type String for the note content.

data class Note(
val id: Int,
val content: String

Step 3: Initializing the SemanticSearch component

Before initializing the SemanticSearch component, we need to set up the OpenAIEmbeddingsModel and PineconeVectorStore, which are two key components required by SemanticSearch.

You can initialize these components in their own function, so you can do stuff like loading the tokens from your environment variables.

You can follow both the PineCone quickstart guide as well as the OpenAI API keys guide to get the required tokens.

import com.aallam.openai.client.OpenAI
import com.theagilemonkeys.ellmental.embeddingsmodel.openai.OpenAIEmbeddingsModel
import com.theagilemonkeys.ellmental.vectorstore.pinecone.PineconeVectorStore

fun embeddingsModel(): OpenAIEmbeddingsModel {
val openaiToken = System.getenv("OPEN_AI_API_KEY")
check(openaiToken != null) { "OPEN_AI_API_KEY environment variable is not set" }
with(OpenAI(token = openaiToken)) {
return OpenAIEmbeddingsModel()

fun vectorStore(): PineconeVectorStore {
val pineconeToken = System.getenv("PINECONE_API_KEY")
check(pineconeToken != null) { "PINECONE_API_KEY environment variable is not set" }
val pineconeUrl = System.getenv("PINECONE_URL")
check(pineconeUrl != null) { "PINECONE_URL environment variable is not set" }
// This is an optional parameter. It allows you to specify a namespace for your vector store.
val pineconeNamespace = System.getenv("PINECONE_NAMESPACE")
return PineconeVectorStore(apiKey = pineconeToken, url = pineconeUrl, namespace = pineconeNamespace)

Next, we will do the same for initializing the SemanticSearch instance. Import the SemanticSearch component from eLLMental and create an instance using the functions you created above.

import com.theagilemonkeys.ellmental.embeddingsmodel.openai.OpenAIEmbeddingsModel
import com.theagilemonkeys.ellmental.semanticsearch.SemanticSearch
import com.theagilemonkeys.ellmental.vectorstore.pinecone.PineconeVectorStore

fun semanticSearch(): SemanticSearch {
val embeddingsModel = embeddingsModel()
val vectorStore = vectorStore()
with(embeddingsModel) {
with(vectorStore) {
return SemanticSearch()

Step 4: Implementing the learn and search features

Let's implement the learn and search features to add new notes into the SemanticSearch index and find them later.

import com.theagilemonkeys.ellmental.semanticsearch.SemanticSearch
import com.theagilemonkeys.ellmental.semanticsearch.SearchInput

suspend fun learn(semanticSearch: SemanticSearch, note: Note) {

suspend fun search(semanticSearch: SemanticSearch, query: String): List<String> {
// We will put a limit of 10 items for now, itemsLimit = 10) {
// Here entries can be mapped to your corresponding data models based on the returned `SemanticEntry` object
// we're just gonna return the id

Step 5: Creating a main function

The only thing that's missing now is an application loop. Let's implement so in the main function.

fun main() = runBlocking {
val semanticSearch = semanticSearch()
println("Available commands: learn, search, quit\n")
var shouldQuit = false
while (!shouldQuit) {
print("> ")
when (readln()) {
"help" -> {
println("Available commands: learn, search, quit")
"learn" -> {
println("Enter note content:")
val txt = readln()
val note = Note(
id = 1,
content = txt
learn(semanticSearch, note)
"search" -> {
println("Enter query:")
val query = readln()
val results = search(semanticSearch, query)
"quit" -> {
shouldQuit = true
else -> {
println("Invalid command")

Step 6: Running the application

Now you can run your application either from the IDE run button or from the command line using ./gradlew run.

Available commands: learn, search, quit
> learn

Enter note content:
This is a note about eLLMental, an open source library for building intelligent applications.

> search

Enter query:
[This is a note about eLLMental, an open source library for building intelligent applications.]

> quit

This is how you can use the SemanticSearch component from eLLMental to build an intelligent note taking application backend that learns from your notes and is able to find relevant notes based on a query.

Here is a link to the repository that contains this entire example.

Wanna learn more? Read the Components page to discover the different components available in eLLMental, together with their thorough usage, and if you have any question, you are very welcome at the community Discord server!

Additional sample code

In addition to this guide, you can find a simpler sample application that exposes the SemanticSearch component from eLLMental as a REST API in the eLLMental Examples repository.