Snowflake-Labs/sfguide-financial-advisor-for-asset-management

Snowflake-Labs/sfguide-financial-advisor-for-asset-management

License: MIT

Stars: 6

Forks: 7

Open issues: 0

Created: 2025-10-10T16:34:14Z

Pushed: 2026-02-06T19:46:48Z

Default branch: main

Fork: no

Archived: no

README:

Financial Advisor for Asset Management

This repo contains everything you need to get started with Snowflake Intelligence for wealth management and asset management. You will learn how Snowflake Intelligence leverages natural language to drive deep insights from both structured portfolio data and unstructured market research. This example demonstrates how financial advisors can use AI to analyze client portfolios, track performance against benchmarks, and access insights from equity research, market commentary, and regulatory filings.

Repository Structure

├── README.md # This file
├── LEGAL.md # Legal notice
├── LICENSE # License information
├── scripts/
│ ├── setup.sql # Database setup script
│ ├── configure_search_services.sql # Search services configuration
│ ├── semantic_models/ # Semantic model definitions
│ │ └── FINANCIAL_ADVISOR_MODEL.yaml
│ └── pdfs/ # Financial research documents
│ ├── AAPL_10K_excerpts_2024.pdf
│ ├── MSFT_10K_excerpts_2024.pdf
│ ├── equity_research_*.pdf
│ ├── Fed_FOMC_Minutes_November_2024.pdf
│ └── market_commentary_Q4_2024.pdf

📋 Getting Started

Required Setup

Building a Cortex Agent for Snowflake Intelligence involves four main steps: creating the necessary Snowflake objects, uploading your financial research documents, configuring search services, and finally, creating and configuring the agent itself in Snowsight.

1. Setting Up the Environment

First, you'll need to create the database, schema, stages, tables, and all necessary Snowflake objects. This ensures a clean and organized environment for your financial advisor demo.

Run the SQL commands in scripts/setup.sql in a Snowflake Workspaces SQL file.

2. Uploading Financial Research Documents via Snowsight

Once the stage is created, you can manually upload your PDF research documents directly through the Snowsight user interface.

1. Navigate to your Stage in Snowsight:

• In the Snowsight left-hand navigation pane, go to Catalog -> Database Explorer.
• Expand your database (FINANCIAL_ADVISOR_DEMO_DB).
• Expand the schema (ANALYTICS).
• Click on Stages.
• Click on the FINANCIAL_ADVISOR_STAGE stage.

2. Upload the Files:

• Download all the PDF files from scripts/pdfs/ directory
• Click the + Files button in the top-right corner.
• Click Browse
• Select all downloaded PDF files (equity research reports, 10-K excerpts, Fed minutes, and market commentary) then select Open.
• Click Upload to begin the upload. You will see a progress bar for each file, and they will appear in the stage's file list once complete.

3. Uploading the Semantic Model

The semantic model defines the structure of your data and enables natural language queries through Cortex Analyst. You need to upload the semantic model YAML file to make it available to the agent.

1. Navigate to the Semantic Models Stage in Snowsight:

• In the Snowsight left-hand navigation pane, go to Catalog -> Database Explorer.
• Expand your database (FINANCIAL_ADVISOR_DEMO_DB).
• Expand the schema (ANALYTICS).
• Click on Stages.
• Click on the SEMANTIC_MODELS_STAGE stage.

2. Upload the Semantic Model File:

• Download the FINANCIAL_ADVISOR_MODEL.yaml file from the scripts/semantic_models/ directory
• Click the + Files button in the top-right corner.
• Click Browse
• Select the FINANCIAL_ADVISOR_MODEL.yaml file then Open.
• Click Upload to begin the upload. Once complete, the file will appear in the stage's file list.

4. Configuring Search Services

With your research documents uploaded, you need to configure the Cortex Search Services that will enable the agent to search through unstructured content.

Run all SQL commands in scripts/configure_search_services.sql in a Snowflake Workspaces SQL file. This will create search services for:

• Client call transcripts
• Equity research reports
• Market commentary documents
• 10-K filing excerpts
• General financial documents

5. Creating the Cortex Agent

With your data loaded, semantic model uploaded, and search services configured, you can now create the agent in the Snowflake UI and connect it to your semantic models and search services.

1. Create the Agent:

• In Snowsight, go to AI & ML.
• Select Agents.
• Click the + Create Agent button.

2. Fill in the Agent Details:

• Display name: FINANCIAL_ADVISOR_AGENT
• Description:

This advanced AI assistant combines structured portfolio analytics with comprehensive market research to help you deliver exceptional client service. Instantly analyze client performance, track goal progress, and access insights from equity research, market commentary, and client interactions.

3. Add the Semantic Model as a Tool:

• In the agent creation wizard, go to the Tools tab.
• For the "Cortex Analyst" section, click the + Add button.
• Select the Semantic Model File option.
• Choose the following:
• Database: FINANCIAL_ADVISOR_DEMO_DB
• Schema: ANALYTICS
• Stage: SEMANTIC_MODELS_STAGE
• Select FINANCIAL_ADVISOR_MODEL.yaml from the list
• Name: FINANCIAL_ADVISOR_SEMANTIC_MODEL
• Description: Click "Generate with Cortex"
• Query Timeout: 60 seconds
• Click Add.

4. Add Cortex Search Services:

• Still in the Tools tab, under "Cortex Search Services", click + Add.
• For each search service below, configure:

CLIENT_TRANSCRIPTS_SEARCH_SERVICE:

• Database: FINANCIAL_ADVISOR_DEMO_DB
• Schema: ANALYTICS
• From the dropdown select: FINANCIAL_ADVISOR_DEMO_DB.ANALYTICS.CLIENT_TRANSCRIPTS_SEARCH_SERVICE
• Name: CLIENT_TRANSCRIPTS_SEARCH_SERVICE
• Description: Search Service for client transcripts
• Click Add

RESEARCH_SEARCH_SERVICE:

• Database: FINANCIAL_ADVISOR_DEMO_DB
• Schema: ANALYTICS
• From the dropdown select: FINANCIAL_ADVISOR_DEMO_DB.ANALYTICS.RESEARCH_SEARCH_SERVICE
• Name: RESEARCH_SEARCH_SERVICE
• Description: Search Service for equity research reports
• Click Add

DOCUMENTS_SEARCH_SERVICE:

• Database: FINANCIAL_ADVISOR_DEMO_DB
• Schema: ANALYTICS
• From the...