Skip to content

Free Dev Day — June 25 — Virtual

Don’t just hear about AI — build it. Luminary talks and hands-on labs.

Claim my spot
Snowflake for Developers/Guides/Cortex AI Cost Dashboard
Community Solution

Cortex AI Cost Dashboard

Colin Mattimore, Ashwin Proddutoor
Fork Repo

Overview

Duration: 2

A unified cost reporting dashboard for all Snowflake Cortex AI services, with dynamic tiered pricing based on account configuration and a Cortex Analyst chatbot.

This dashboard aggregates usage data from Cortex AI service views in SNOWFLAKE.ACCOUNT_USAGE into local tables for fast querying, computes dollar costs using a service-tier and region-based pricing model, then visualizes the data in a Streamlit app with an integrated Cortex Analyst chatbot.

Prerequisites

  • Snowflake account with ACCOUNTADMIN role (or equivalent with ACCOUNT_USAGE access)
  • A warehouse for query execution
  • Streamlit in Snowflake (SiS) enabled

What You'll Learn

  • How to aggregate Cortex AI usage data across services
  • How to model tiered, region-based pricing for Cortex services
  • How to build a Streamlit-in-Snowflake dashboard with an integrated Cortex Analyst chatbot

What You'll Build

  • 16 Snowflake tables (1 unified summary + 1 user costs + 14 detail tables)
  • Reference tables for tiered credit rates and per-model token rates
  • 2 stored procedures with dynamic pricing logic
  • A daily refresh task
  • A Streamlit dashboard with a Cortex Analyst chatbot

Pricing Model

Duration: 3

Service Tiers

Services are classified into two tiers:

TierServices
AI_SERVICESCortex Code CLI, Cortex Code Snowsight, Cortex Agent, Snowflake Intelligence
STANDARDAll other services (AISQL, Analyst, Search, Document Processing, etc.)

Credit Rates (CREDIT_BASE_RATES table) - Customize with your account pricing

Service TierRegion TypeBase RateTier DiscountEffective RatePeriod
AI_SERVICESGLOBAL$2.00{{AI_TIER_DISCOUNT_PCT}}%varies2026-04-01 onwards
AI_SERVICESREGIONAL$2.20{{AI_TIER_DISCOUNT_PCT}}%varies2026-04-01 onwards
TRADITIONALANY{{CREDIT_RATE}}0%{{CREDIT_RATE}}/credit2026-04-01 onwards
TRADITIONALANY{{HISTORICAL_CREDIT_RATE}}0%{{HISTORICAL_CREDIT_RATE}}/credit2025-01-01 to 2026-03-31

Region Detection

At refresh time, the stored procedures run SHOW PARAMETERS LIKE 'CORTEX_ENABLED_CROSS_REGION' IN ACCOUNT. This applies to AI_SERVICES tier only — STANDARD services use a flat rate regardless of cross-region setting:

Parameter ValueRegion Type
ANY_REGION, AWS_GLOBAL, GCP_GLOBAL, AZURE_GLOBALGLOBAL
Any other valueREGIONAL

Cost Formula

  • AI_SERVICES tier: COST_USD = CREDITS × BASE_RATE × (1 - TIER_DISCOUNT_PCT / 100) (rate depends on GLOBAL/REGIONAL)
  • STANDARD tier: COST_USD = CREDITS × BASE_RATE (flat rate, no region classification)
  • REST API: COST_USD = (INPUT_TOKENS × INPUT_RATE + OUTPUT_TOKENS × OUTPUT_RATE) / 1,000,000 × (1 - DISCOUNT_PCT/100) from REST_API_TOKEN_RATES table (per Snowflake Credit Consumption Table 6(c))

Cortex AI Services Covered

Duration: 1

ServiceDescriptionCost Basis
CORTEX_AISQLAI SQL functions (COMPLETE, EXTRACT, SUMMARIZE, etc.)Credits (STANDARD)
CORTEX_ANALYSTText-to-SQL natural language queriesCredits (STANDARD)
CORTEX_DOCUMENT_PROCESSINGDocument parsing and extractionCredits (STANDARD)
CORTEX_FINE_TUNINGModel fine-tuningCredits (STANDARD)
CORTEX_FUNCTIONS_QUERYPer-query LLM function detailsCredits (STANDARD)
CORTEX_PROVISIONED_THROUGHPUTPTU consumptionCredits (STANDARD)
CORTEX_REST_APIREST API calls$ per 1M tokens
CORTEX_SEARCHCortex Search daily usageCredits (STANDARD)
CORTEX_SEARCH_SERVINGSearch serving costsCredits (STANDARD)
DOCUMENT_AIDocument AI extractionCredits (STANDARD)
CORTEX_CODE_CLICortex Code CLI usageCredits (AI_SERVICES)
CORTEX_CODE_SNOWSIGHTCortex Code Snowsight usageCredits (AI_SERVICES)
CORTEX_AGENTCortex Agent invocationsCredits (AI_SERVICES)
SNOWFLAKE_INTELLIGENCESnowflake Intelligence queriesCredits (AI_SERVICES)

Setup option A (via Cortex Code skill):

Duration: 15

  1. Download the repo from the FORK REPO link above. Open the repo in Cortex Code CLI or Cortex Code Desktop and give the following prompt, "Use the skill to deploy the ai cost dashboard." Follow the prompts to input your account specific parameters. Deploy.

Setup option B (manual): Run the SQL Setup Script

Duration: 10

  1. Download setup.sql from this guide's assets folder.
  2. Connect to your Snowflake account with SYSADMIN role (or ACCOUNTADMIN).
  3. Open setup.sql in a Snowflake worksheet.
  4. (Optional) Modify the configuration at the top: 
    USE ROLE ACCOUNTADMIN; -- Use role with appropriate privileges (Create DB, Create Task, IMPORTED PRIVILEGES on SNOWFLAKE DB, USAGE ON WAREHOUSE)
USE WAREHOUSE {{WAREHOUSE}};
USE DATABASE {{DATABASE}};
  5. Run the entire script.

This will create:

  • 16 tables (1 unified summary + 1 user costs + 14 detail tables)
  • 1 reference table (CREDIT_BASE_RATES) with tiered/regional rates
  • 1 reference table (REST_API_TOKEN_RATES) with per-model token rates
  • 2 stored procedures (REFRESH_CORTEX_AI_COSTS, REFRESH_USER_AI_COSTS) with dynamic pricing logic
  • 1 daily refresh task (runs at 6 AM UTC on {{WAREHOUSE}}, auto-resumed)
  • Initial data load for the last 365 days

Setup: Deploy the Streamlit App

Duration: 5

Upload streamlit_app.py and cortex_ai_costs_model.yaml and create the Streamlit object:

CREATE STAGE IF NOT EXISTS {{DATABASE}}.{{SCHEMA}}.STREAMLIT_STAGE;

PUT file:///path/to/streamlit_app.py @{{DATABASE}}.{{SCHEMA}}.STREAMLIT_STAGE OVERWRITE=TRUE AUTO_COMPRESS=FALSE;
PUT file:///path/to/cortex_ai_costs_model.yaml @{{DATABASE}}.{{SCHEMA}}.STREAMLIT_STAGE OVERWRITE=TRUE AUTO_COMPRESS=FALSE;

CREATE OR REPLACE STREAMLIT {{DATABASE}}.{{SCHEMA}}.CORTEX_AI_COST_DASHBOARD
  ROOT_LOCATION='@{{DATABASE}}.{{SCHEMA}}.STREAMLIT_STAGE'
  MAIN_FILE='/streamlit_app.py'
  QUERY_WAREHOUSE='{{WAREHOUSE}}'
  TITLE='Cortex AI Cost Dashboard';

Using the Dashboard

Duration: 5

Dashboard Features

  • KPI Metrics: Total credits, total cost ($), tokens, requests, services used, active days, active users
  • Credit Distribution by Category: Metric cards showing cost breakdown per service category
  • Month over Month Costs: Bar chart with credit and dollar tooltips
  • Daily Credit Trend: Area chart showing usage over time
  • Credits by Service: Bar chart per service
  • Tokens by Model: Pie chart showing top 10 models by token consumption

Detailed Data Tabs

TabDescription
Unified SummaryAggregated costs by date, service, and model
By ServiceCosts grouped by service category
By ModelCosts grouped by AI model
By UserPer-user cost breakdown with drill-down to individual queries
Cortex AgentCosts by agent name (nulls shown as 'API CALL'), with daily detail
Snowflake IntelligenceCosts by SI name (nulls shown as 'API CALL'), with daily detail
Cortex CodeCredits by user for the last 30 days — Snowsight grid on top, CLI grid below

Cortex Analyst Chatbot

The dashboard includes an integrated Cortex Analyst chatbot at the bottom, powered by a semantic model (cortex_ai_costs_model.yaml). Ask natural language questions about your AI costs, e.g.:

  • "What is the total cost in dollars over the last 30 days?"
  • "Which service has the highest credit usage this month?"
  • "Show me daily costs for Cortex Agent"

Sidebar Controls

  • Time Range: Filter by last 7/14/30/60/90/365 days or custom range
  • Service Categories: Filter by specific service types
  • Refresh Data: Click to call both stored procedures and pull latest data

Refreshing Data

Duration: 5

CALL {{DATABASE}}.{{SCHEMA}}.REFRESH_CORTEX_AI_COSTS(N); -- # of days
CALL {{DATABASE}}.{{SCHEMA}}.REFRESH_USER_AI_COSTS(N); -- # of days

Scheduled Refresh

The daily task runs automatically at 6 AM UTC on {{WAREHOUSE}}. To check status:

SHOW TASKS LIKE 'REFRESH_CORTEX_AI_COSTS_DAILY' IN SCHEMA {{DATABASE}}.{{SCHEMA}};

Conclusion And Resources

Duration: 1

You've built an end-to-end Cortex AI cost dashboard with tiered pricing and a Cortex Analyst chatbot.

What You Learned

  • Aggregating Cortex AI usage from SNOWFLAKE.ACCOUNT_USAGE
  • Modeling tiered, region-based pricing for Cortex services
  • Building a Streamlit-in-Snowflake dashboard with Cortex Analyst

Related Resources

Updated 2026-06-08

This content is provided as is, and is not maintained on an ongoing basis. It may be out of date with current Snowflake instances