Unlock Insights from Unstructured Data with Snowflake Cortex AI
summary: Unlock Insights from Unstructured Data with Snowflake Cortex AI id: unlock-insights-from-unstructured-data-with-snowflake-cortex-ai categories: snowflake-site:taxonomy/solution-center/certification/quickstart, snowflake-site:taxonomy/product/data-engineering language: en environments: web status: Published feedback link: https://github.com/Snowflake-Labs/sfguides/issues author: Sean Morris, Stephen Dickson [environment_name]: ai209
Overview
This guide demonstrates how to create a Streamlit application running inside Snowflake that unlocks insights from unstructured data using Snowflake Cortex AI.
It shows how to translate, summarize, classify text, generate emails, and even analyze images — all without deploying external infrastructure.
Prerequisites
- A Snowflake account. If you don’t have one, you can sign up for a free trial here.
- The ACCOUNTADMIN role to ensure you have full access for setup and configuration.
What You’ll Learn
- How to create a database, schema, and stage in Snowflake
- How to store unstructured data in Snowflake stages
- How to use SQL to run LLM functions (
TRANSLATE,SENTIMENT,SUMMARIZE,COMPLETE) - How to build a frontend application with Streamlit inside Snowflake
What You'll Build
A fully functioning Streamlit app inside Snowflake that:
- Translates between multiple languages
- Scores sentiment of call transcripts
- Summarizes long-form text
- Classifies customer service inquiries
- Analyzes uploaded images using multimodal LLMs
Setup
Login to Snowsight
Log into Snowflake's web interface Snowsight using your Snowflake credentials.
Enable Cross-Region Inference
In the Snowsight UI on the left hand sidebar, select the Projects > Worksheets tab.
In the top right hand corner, click the + button to create a new SQL worksheet.
Run the following SQL commands to enable cortex cross-region inference.
-- set role context USE ROLE accountadmin; -- enable Cortex Cross Region Inference ALTER ACCOUNT SET CORTEX_ENABLED_CROSS_REGION = 'ANY_REGION';
Create Snowflake Objects
In the same SQL worksheet, run the following SQL commands to create the warehouse, database, schema, and stage
aside positive IMPORTANT:
If you use different names for objects created in this section, be sure to update scripts and code in the following sections accordingly.
-- Set role context USE ROLE sysadmin; -- Set Environment Name SET var_name = 'ai209'; -- Generate Environment Identifiers SET var_warehouse_name = $var_name||'_wh'; SET var_database_name = $var_name||'_db'; SET var_schema_name = $var_name||'_db.demo'; -- Set Warehouse Context CREATE WAREHOUSE IF NOT EXISTS IDENTIFIER($var_warehouse_name) WITH WAREHOUSE_SIZE = xsmall AUTO_SUSPEND = 60 INITIALLY_SUSPENDED = true; USE WAREHOUSE IDENTIFIER($var_warehouse_name); -- Set Database Schema Context CREATE DATABASE IF NOT EXISTS IDENTIFIER($var_database_name); CREATE SCHEMA IF NOT EXISTS IDENTIFIER($var_schema_name); USE SCHEMA IDENTIFIER($var_schema_name);
Build Streamlit Application
Let's create a Streamlit application for interactive image analysis:
Setting Up the Streamlit App
To create and configure your Streamlit application in Snowflake:
-
Navigate to Streamlit in Snowflake:
- Click on the Streamlit tab in the left navigation pane
- Click on + Streamlit App button in the top right
-
Configure App Settings:
- Enter a name for your app (e.g., "AI/ML Toolkit")
- Select the warehouse AI209_WH
- Choose AI209_DB.DEMO as your database and schema
- Click Create
-
Create the app:
- In the editor, paste the complete code provided in the streamlit_app.py file
- Click "Run" to launch your application
The application provides:
- Translation: Translate text between 14 different languages using Snowflake Cortex
- Sentiment Analysis: Analyze the sentiment of call transcripts
- Data Summarization: Generate concise summaries of large text datasets
- Next Best Action: Use foundation models to identify customer next actions
- Text Classification: Categorize text into predefined categories
- Email Generation: Create customer emails based on call transcripts
- Question Answering: Ask questions to foundation models
- Multi-Modal Image Analysis: Analyze and categorize images using multi-modal models
Conclusion And Resources
Congratulations! You’ve successfully built an end-to-end image analysis application using Snowflake Cortex AI models. This app showcases how to unlock value from unstructured data — including text and images - all within the Snowflake environment.
By combining Snowflake-native large language models (LLMs) and multimodal capabilities with Streamlit, you built an interactive solution that can:
- Translate and classify text across multiple languages
- Analyze sentiment and summarize long-form content
- Automatically generate emails and next-best-actions from transcripts
- Perform real-time image analysis using multimodal foundation models
This end-to-end pattern demonstrates how Snowflake Cortex can power intelligent applications directly on top of your data, reducing time-to-insight and simplifying operational complexity.
What You Learned
- How to provision warehouses, databases, schemas, and stages in Snowflake
- How to store and process unstructured text and image data
- How to run SQL-based AI functions (TRANSLATE, SUMMARIZE, SENTIMENT, etc.)
- How to analyze visual data using foundation models
- How to build a secure, fully integrated frontend using Streamlit in Snowflake
Related Resources
This content is provided as is, and is not maintained on an ongoing basis. It may be out of date with current Snowflake instances
