Document Embeddings with InterSystems IRIS Cloud SQL

This project demonstrates how to create embeddings from documents using Nomic’s multimodal embedding models and store them in InterSystems IRIS Cloud SQL. It also provides a Streamlit UI to search across your embedded data.

---

0. Requirements

This project requires:

• Python 3.11+

• pip or uv for environment setup

• GPU (for local inference with nomic-ai/colnomic-embed-multimodal-3b or 7b)

  • Recommended at least 12gb VRAM: NVIDIA T4 or RTX 3090

---

1. Cloud SQL Trial

To get started, you’ll need access to the InterSystems Cloud SQL Trial:

1. Sign up at InterSystems Cloud SQL Trial.
2. Create a new SQL database instance.
3. Copy your connection details (host, port, user, password, namespace).
4. Download the connection certificate from cloud SQL and rename connection.pem (and put in the connection folder.)

Here’s a step-by-step guide to getting set up with InterSystems IRIS Cloud SQL, whether you're using the free trial or a paid subscription:

Step 1: Access the Cloud Services Portal

Head over to the InterSystems Cloud Services Portal to begin — here you can sign up for either a Cloud SQL trial or a paid subscription, and create a deployment.(InterSystems Documentation)

Step 2: Navigate to the Services Page

Once inside the portal, go to the Services section (usually under the main menu). There, you'll find a card for InterSystems IRIS Cloud SQL where you can click Subscribe to view pricing and subscription options.(InterSystems Documentation)

Step 3: Choose Trial or Subscription

• Trial access: Often accompanied by introductory credits (e.g., the site mentions “$300 in FREE credits”).(InterSystems Corporation)
• Paid Subscription: If you're ready for production-grade use, the portal lets you subscribe directly or via your cloud provider (e.g., AWS Marketplace). Billing is handled through InterSystems or your cloud account, depending on your selection.(InterSystems Documentation)

Step 4: Launch a New Deployment

After subscribing, head to the Deployments page. Click Create New Deployment and follow the prompts to configure details such as size, region, service level (e.g., Development, Test, Live), and any additional options like external connectivity.(InterSystems Documentation)

Step 5: Manage and Configure Your Deployment

Once created, you can access your deployment’s Overview page to:

• View connection details (host, port, namespace, etc.)
• Enable or disable external access as needed (0.0.0.0/0 or other subnet)
• Adjust security options like IP restrictions or TLS settings(InterSystems Documentation, InterSystems Documentation)
• Download the SQL pem certificate and rename it connection.pem

---

Optional: Video Walkthrough

If you'd prefer a quick visual tour, there's a helpful video demonstration titled “Working with InterSystems IRIS Cloud SQL” that walks through deployment and connection processes.(InterSystems Developer Community)

---

Summary Table

Step	Action
1️⃣	Log in to InterSystems Cloud Services Portal
2️⃣	Navigate to Services → choose Cloud SQL → Subscribe
3️⃣	Select Trial (if available) or Paid Subscription
4️⃣	Under Deployments, click Create New Deployment
5️⃣	Configure your deployment and then access its Overview
6️⃣	Download driver → gather connection info → connect securely (e.g., via Python DB-API)

---

2. Setting up Project

Clone the repo and install dependencies:

git clone https://github.com/isc-nmitchko/iris-document-search.git
cd iris-document-search
pip -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Run setup to configure your environment:

python main.py --setup

This will create a .env file containing your IRIS connection string and model settings.

---

4. Embedding Data

You can embed documents (PDFs, text files, etc.) using the Nomic multimodal embedding models:

• "nomic-ai/colnomic-embed-multimodal-3b" (lighter, faster)
• "nomic-ai/colnomic-embed-multimodal-7b" (larger, more accurate)

Example usage:

python main.py --embed ./data/my_docs/

This will:

1. Extract text from your documents.
2. Generate embeddings using the configured model.
3. Insert the embeddings into your InterSystems IRIS Cloud SQL instance.

---

5. Searching using Streamlit UI

To search across embedded data, launch the Streamlit UI:

streamlit run ui2.py

Features include:

• Free-text semantic search
• Similarity ranking of documents
• Viewing results directly in a web interface