setup-deploy

name	setup-deploy
description	Deploy AI Cost Dashboard stored procedures and Streamlit app to Snowflake.
parent_skill	ai-cost-dashboard

Setup / Deploy AI Cost Dashboard

When to Load

From ai-cost-dashboard SKILL.md when user selects Setup / Deploy intent.

Prerequisites

• Snowflake connection is active
• ACCOUNTADMIN role or role with ACCOUNT_USAGE access
• Source files available at cortex_ai_cost_dashboard/

Configurable Parameters

All object names are configurable. The source files use template placeholders ({{DATABASE}}, {{SCHEMA}}, {{WAREHOUSE}}). The deploy process resolves these before execution.

Workflow

Step 1: Configure Deployment Parameters

Goal: Collect or confirm all deployment targets.

Actions:

1. Ask user using ask_user_question with all parameters in a single call:

   • Connection name (text, default: current connection)
   • Database (text, default: "ADMIN_DB")
   • Schema (text, default: "AI_COSTS")
   • Warehouse (text, default: "COMPUTE_WH")
   • Role (text, default: "ACCOUNTADMIN")
   • Credit rate — the customer's current standard $/credit (text, default: "2.88")
   • Historical credit rate — the customer's previous $/credit before current rate (text, default: "3.01")

2. Store as CONNECTION, DATABASE, SCHEMA, WAREHOUSE, ROLE, CREDIT_RATE, HISTORICAL_CREDIT_RATE.

Output: Confirmed deployment parameters

Step 2: Resolve Templates and Run Setup SQL

Goal: Create database, schema, tables, stored procedures, rate tables, and daily task.

Actions:

1. Read cortex_ai_cost_dashboard/setup.sql

2. Resolve the {{WAREHOUSE}} placeholder with user's warehouse value

3. Update {{CREDIT_RATE}} with user's current credit rate and {{HISTORICAL_CREDIT_RATE}} with their historical rate

4. Write resolved SQL to /tmp/ai_cost_dashboard_deploy/setup.sql

5. Present deployment plan:

   Will create in <DATABASE>.<SCHEMA>:
   - 15+ tables (CORTEX_AI_UNIFIED_COSTS, USER_AI_COSTS, + detail tables)
   - 2 stored procedures (REFRESH_CORTEX_AI_COSTS, REFRESH_USER_AI_COSTS)
   - 1 scheduled task (REFRESH_CORTEX_AI_COSTS_DAILY - 6AM UTC daily)
   - Credit rate tables (current: $<CREDIT_RATE>/credit, historical: $<HISTORICAL_CREDIT_RATE>/credit)
   - Initial data load: 365 days
   

6. STOP: Get user approval before executing.

7. Execute via bash:

   snow sql -f /tmp/ai_cost_dashboard_deploy/setup.sql \
     --connection <CONNECTION> --role <ROLE> \
     --warehouse <WAREHOUSE> --database <DATABASE> --schema <SCHEMA>
   

8. Verify:

   SELECT TABLE_NAME, ROW_COUNT
   FROM <DATABASE>.INFORMATION_SCHEMA.TABLES
   WHERE TABLE_SCHEMA = '<SCHEMA>'
   ORDER BY TABLE_NAME;
   

If errors occur:

• ACCOUNT_USAGE access denied -> User needs ACCOUNTADMIN or IMPORTED PRIVILEGES on SNOWFLAKE db
• Table/procedure creation fails -> Check role permissions on target database

Output: Created database objects

Step 3: Deploy Streamlit App

Goal: Deploy the Streamlit app to Snowflake.

Actions:

1. Resolve cortex_ai_cost_dashboard/snowflake.yml — replace {{DATABASE}} → <DATABASE>, {{SCHEMA}} → <SCHEMA>, {{WAREHOUSE}} → <WAREHOUSE>. Write back in-place (will restore after deploy).

2. Create stage:

   USE SCHEMA <DATABASE>.<SCHEMA>;
   CREATE STAGE IF NOT EXISTS STREAMLIT_STAGE;
   

3. Deploy via bash from the project directory:

   snow streamlit deploy --connection <CONNECTION> --replace
   

4. Restore snowflake.yml to its template version (with {{}} placeholders).

5. Verify:

   SHOW STREAMLITS LIKE 'CORTEX_AI_COST_DASHBOARD' IN SCHEMA <DATABASE>.<SCHEMA>;
   

Output: Deployed Streamlit app

Step 4: Upload Semantic Model

Goal: Upload resolved semantic model YAML to stage.

Actions:

1. Resolve cortex_ai_cost_dashboard/cortex_ai_costs_model.yaml — replace {{DATABASE}} → <DATABASE>, {{SCHEMA}} → <SCHEMA>. Write to /tmp/ai_cost_dashboard_deploy/cortex_ai_costs_model.yaml.

2. Upload via bash:

   snow stage copy /tmp/ai_cost_dashboard_deploy/cortex_ai_costs_model.yaml \
     @<DATABASE>.<SCHEMA>.STREAMLIT_STAGE --connection <CONNECTION> --overwrite
   

Output: Semantic model on stage

Step 5: End-to-End Validation

Goal: Confirm everything works together.

Actions:

1. Test stored procedures:

   CALL <DATABASE>.<SCHEMA>.REFRESH_CORTEX_AI_COSTS(7);
   CALL <DATABASE>.<SCHEMA>.REFRESH_USER_AI_COSTS(7);
   

2. Check task status:

   SHOW TASKS LIKE 'REFRESH_CORTEX_AI_COSTS_DAILY' IN SCHEMA <DATABASE>.<SCHEMA>;
   

   Ask user if they want to resume the task.

3. Cleanup: rm -rf /tmp/ai_cost_dashboard_deploy

4. Present summary:

   Deployment complete:
   - Database/Schema: <DATABASE>.<SCHEMA>
   - Warehouse: <WAREHOUSE>
   - Credit Rate: $<CREDIT_RATE>/credit (current), $<HISTORICAL_CREDIT_RATE>/credit (historical)
   - Tables: created and populated
   - Procedures: REFRESH_CORTEX_AI_COSTS, REFRESH_USER_AI_COSTS
   - Task: REFRESH_CORTEX_AI_COSTS_DAILY (suspended/resumed)
   - Streamlit: <DATABASE>.<SCHEMA>.CORTEX_AI_COST_DASHBOARD
   - Semantic Model: @<DATABASE>.<SCHEMA>.STREAMLIT_STAGE/cortex_ai_costs_model.yaml
   

Output: Validated deployment

Stopping Points

• After Step 2 plan: Before executing setup SQL
• After Step 3 plan: Before deploying Streamlit app

Output

Fully deployed AI Cost Dashboard with all objects in user-specified <DATABASE>.<SCHEMA>.