Snowflake-Labs/sfguide-Build-an-AI-Assistant-for-FSI-with-AISQL-and-Snowflake-Intelligence

Snowflake-Labs/sfguide-Build-an-AI-Assistant-for-FSI-with-AISQL-and-Snowflake-Intelligence

Language: Jupyter Notebook

License: Apache-2.0

Stars: 0

Forks: 0

Open issues: 0

Created: 2025-11-27T13:45:44Z

Pushed: 2025-12-13T00:05:36Z

Default branch: main

Fork: no

Archived: no

README:

Building an AI Assistant for FSI using AISQL and Snowflake Intelligence

Overview

Build a production-ready AI assistant for financial services using Snowflake Cortex AI, Snowflake Intelligence, and Document AI.

No downloads required! Deploy directly from GitHub using Snowflake's Git integration.

What You'll Build

• Unstructured Data Processing: Extract insights from PDFs, images, and audio using Document AI
• Structured Data Processing: Build ML models for stock prediction using Snowflake ML
• Agent Tools: Create semantic search services and natural language query interfaces
• Applications: Deploy intelligent agents with Snowflake Intelligence and Streamlit

Use Case

Build a Stock Selection Agent that processes and analyzes data from diverse sources:

• 30 Analyst Reports from 6 research firms with ratings and price targets
• 92 Earnings Call Transcripts from 11 companies with sentiment analysis
• 11 Financial Infographics with extracted KPIs
• 950 Analyst Emails with AI-extracted tickers and ratings
• 6,420 Stock Price Data Points for predictive modeling
• 22 Annual Reports with comprehensive financial statements
• 4 Audio Files transcribed with speaker identification

Quick Start (15-20 minutes)

Prerequisites

• Snowflake account (free trial works for most features)
• ACCOUNTADMIN role access
• No downloads or CLI tools needed

> ⚠️ Note for Trial Accounts: The Web Search feature requires External Access Integration, which is not enabled by default on trial accounts. If you need this feature, contact your Snowflake representative to have it enabled. All other features work without this.

Step 1: Setup Git Integration

Open a SQL Worksheet in Snowflake and run:

USE ROLE ACCOUNTADMIN;

-- Create separate database for Git repos (persists across deployments)
CREATE DATABASE IF NOT EXISTS SNOWFLAKE_QUICKSTART_REPOS;
CREATE SCHEMA IF NOT EXISTS SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS;

-- Create API integration for GitHub
CREATE OR REPLACE API INTEGRATION git_api_integration
API_PROVIDER = git_https_api
API_ALLOWED_PREFIXES = ('https://github.com/Snowflake-Labs/')
ENABLED = TRUE;

-- Create Git repository
CREATE OR REPLACE GIT REPOSITORY SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.ACCELERATE_AI_IN_FSI_REPO
API_INTEGRATION = git_api_integration
ORIGIN = 'https://github.com/Snowflake-Labs/sfguide-Build-an-AI-Assistant-for-FSI-with-AISQL-and-Snowflake-Intelligence.git';

-- Fetch latest code
ALTER GIT REPOSITORY SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.ACCELERATE_AI_IN_FSI_REPO FETCH;

Step 2: Deploy (Run Scripts 01-08)

-- Run deployment scripts directly from GitHub
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.ACCELERATE_AI_IN_FSI_REPO/branches/main/assets/sql/01_configure_account.sql;
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.ACCELERATE_AI_IN_FSI_REPO/branches/main/assets/sql/02_data_foundation.sql;
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.ACCELERATE_AI_IN_FSI_REPO/branches/main/assets/sql/03_deploy_cortex_analyst.sql;
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.ACCELERATE_AI_IN_FSI_REPO/branches/main/assets/sql/04_deploy_streamlit.sql;
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.ACCELERATE_AI_IN_FSI_REPO/branches/main/assets/sql/05_deploy_notebooks.sql;
-- Optional GPU notebook (skip if GPU unavailable):
-- EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.ACCELERATE_AI_IN_FSI_REPO/branches/main/assets/sql/05b_deploy_gpu_notebook.sql;
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.ACCELERATE_AI_IN_FSI_REPO/branches/main/assets/sql/06_deploy_documentai.sql;
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.ACCELERATE_AI_IN_FSI_REPO/branches/main/assets/sql/07_deploy_snowmail.sql;
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.ACCELERATE_AI_IN_FSI_REPO/branches/main/assets/sql/08_custom_agent_tools.sql;

What Gets Deployed

✅ Database: ACCELERATE_AI_IN_FSI
✅ 20+ tables (~10,000 rows)
✅ 5 Cortex Search Services
✅ 2 Cortex Analyst Semantic Views
✅ 1 Snowflake Intelligence Agent (One Ticker)
✅ 1 Streamlit Application (StockOne)
✅ 4 Snowflake Notebooks
✅ 132 Document files (PDFs, images, audio)

---

Try It Out

1. StockOne Streamlit App

Navigate to: AI & ML Studio → Streamlit → STOCKONE_AGENT

Try:

• "What is the latest Snowflake stock price?"
• "Show me revenue trend over the last 4 quarters"
• "What did analysts say in the latest earnings call?"

2. One Ticker Agent

Navigate to: AI & ML Studio → Snowflake Intelligence → One Ticker

Try:

• "Give me top 3 vs bottom 3 trade predictions"
• "What do analysts say about Snowflake?"
• "Show me SNOW stock performance over time"
• "Search the web for latest Snowflake news"
• "Send me an email summary"

3. SnowMail

Navigate to: Data Products → Apps → SNOWMAIL

View AI-generated email reports from the agent.

4. Snowflake Notebooks

Navigate to: AI & ML Studio → Notebooks

Run:

• 1_EXTRACT_DATA_FROM_DOCUMENTS - Document AI processing
• 2_ANALYSE_SOUND - Audio transcription and analysis
• 3_BUILD_A_QUANTITIVE_MODEL - ML model training (GPU)
• 4_CREATE_SEARCH_SERVICE - Build Cortex Search services

---

Folder Structure

quickstart/
├── README.md → This file
├── quickstart.md → Complete step-by-step guide
├── LICENSE → Apache 2.0 license
│
└── assets/
├── sql/ → Deployment scripts (01-08)
├── data/ → CSV/Parquet files
├── documents/ → PDFs, images, audio
├── Notebooks/ → Snowflake notebooks
├── Streamlit/ → StockOne application
├── native_app_snowmail/ → SnowMail Native App
└── semantic_models/ → YAML definitions

---

Re-deploying / Reset

Since the Git repo is in a separate database, you can easily reset:

-- Drop the main database (Git repo stays safe!)
DROP DATABASE IF EXISTS ACCELERATE_AI_IN_FSI;

-- Fetch latest code
ALTER GIT REPOSITORY SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.ACCELERATE_AI_IN_FSI_REPO FETCH;

-- Re-run deployment scripts (01-08)
EXECUTE IMMEDIATE FROM…