🛍️ AI-Shop Demo

Welcome to the AI-Shop demo! Dive into the synergy of Atlas Vector Search, Langchain JS, and OpenAI.

🌟 Features

• Product Catalog: Display of items.
• AI Assistant: Using LangChain and OpenAI, this assistant provides product suggestions based on varied inputs like recipes or ambiguous requests. With the help of MongoDB Atlas Vector search and dynamic aggregation pipeline we are able to format the results best for the user.

🚀 Workflow

1. Craft a prompt for LLM.
2. Format data and enhance with embeddings.
3. Produce MongoDB vector search aggregations to display relevant results.

🛠️ Requirements

• Atlas Project
• Atlas Cluster (free cluster works!)
• App Services Application (e.g., AI-Shop). Ensure linkage.
• OpenAI account with API access.

🔧 Project Setup

Step 1: Secrets and Values

1. Navigate to your App Services application.
2. Under "Values", create a secret named openAIKey with your OPEN AI API key.
3. Create a linked value named OpenAIKey and link to the secret.

Step 2: Database Setup

1. Create a Database trigger on insert for Open AI embeddings.
   • Database: ai_shop
   • Collection: products

Function: calcEmbeddings

exports = async function(changeEvent) {
    // Get the full document from the change event.
    const doc = changeEvent.fullDocument;

    // Define the OpenAI API url and key.
    const url = 'https://api.openai.com/v1/embeddings';
    // Use the name you gave the value of your API key in the “Values” utility inside of App Services
    const openai_key = context.values.get("openAIKey");

    try {
        console.log(`Processing document with id: ${JSON.stringify(doc)}`);
        

        // Call OpenAI API to get the embeddings.
        let response = await context.http.post({
            url: url,
             headers: {
                'Authorization': [`Bearer ${openai_key}`],
                'Content-Type': ['application/json']
            },
            body: JSON.stringify({
                input: `${doc.title} - ${doc.description}`,
                model: "text-embedding-ada-002"
            })
        });

        // Parse the JSON response
        let responseData = EJSON.parse(response.body.text());

        // Check the response status.
        if(response.statusCode === 200) {
            console.log("Successfully received embedding.");

            const embedding = responseData.data[0].embedding;

            // Get the cluster in MongoDB Atlas.
            const mongodb = context.services.get('mongodb-atlas');
            const db = mongodb.db('ai_shop'); // Replace with your database name.
            const collection = db.collection('products'); // Replace with your collection name.

            // Update the document in MongoDB.
            const result = await collection.updateOne(
                { _id: doc._id },
                // The name of the new field you’d like to contain your embeddings.
                { $set: { embeddings: embedding }}
            );

            if(result.modifiedCount === 1) {
                console.log("Successfully updated the document.");
                
                
            } else {
                console.log("Failed to update the document.");
            }
        } else {
            console.log(`Failed to receive embedding. Status code: ${response.body.text()}`);
        }

    } catch(err) {
        console.error(err);
    }
};

• Verify that the function runs under "SYSTEM" context after saving the trigger.

Step 4 - Ingest Data and Create Vector index

Ingest the data

We need to ingest the data under local directory data:

1. data/ai_shop.categories.json into collection named categories (hosting the possible categories in the shop).
2. data/ai_shop.products.json into collection named products (hosting the products in the shop).

We can use tools like mongoimport or compass to import those files. Or we can take the content and "copy" into "Data Explorer" > "INSERT DOCUMENT" tab: Once data is ingested the trigger should set "embeddings" field under each product.

Create an Atlas Search index by going to the Cluster "Search" tab and specify a new index on:

• Database : ai_shop
• Collection : products

Use "JSON Editor" and specify a "defualt" index:

{
  "mappings": {
    "dynamic": true,
    "fields": {
      "embeddings": {
        "dimensions": 1536,
        "similarity": "cosine",
        "type": "knnVector"
      }
    }
  }
}

Create the index and make sure it is "Active".

Step 5 - Setup application

Create .env file in the main directory with the following content:

OPEN_AI_KEY='<OPEN_AI_KEY>'
DB_URL='<ATLAS_CONNECTION_URI>'

• The database user must have read/write permissions on ai_shop database.

Install the application

npm install

Step 6 - Run application

Start the backend

npm run server

Start fronend

npm run dev

Go to the suggested vite url to browse application:

🛑 Disclaimer 🛑

Use at your own risk. Not an official MongoDB product.