Snowflake-Labs/sfguide-intelligent-production-assistant-for-oil-gas

Snowflake-Labs/sfguide-intelligent-production-assistant-for-oil-gas

Language: Python

License: Apache-2.0

Stars: 0

Forks: 1

Open issues: 0

Created: 2026-02-16T18:56:13Z

Pushed: 2026-03-03T04:37:52Z

Default branch: main

Fork: no

Archived: no

README:

Intelligent Production Assistant (IPA) - Oil & Gas Operational Intelligence

> An AI-powered operational intelligence platform for oil & gas production, built entirely in Snowflake. This solution combines Cortex AI, Cortex Analyst, and Cortex Search to deliver autonomous monitoring of production assets, real-time safety compliance, and proactive cost optimization. From detecting rod pump failures to enforcing H2S permit limits, IPA transforms raw SCADA data into actionable insights through natural language conversations.

---

What You Build

This demo creates a complete operational intelligence platform with:

• 5-Page Streamlit Application: Mission Control dashboard, asset monitoring, safety inspection, and natural language Q&A
• 3 Detection Services: Sentinel (production), Guardian (safety), Fiscal (costs) - Python stored procedures using AI_COMPLETE
• 1 Cortex Agent (IPA_AGENT): True Cortex Agent combining text-to-SQL (Cortex Analyst) + document search (Cortex Search)
• Document Search Service: Cortex Search over technical documents for RAG capabilities
• Complete Data Pipeline: SCADA sensor data, work permits, financial records, and alert management

---

What You Will Learn

By implementing this demo, you'll gain hands-on experience with:

1. AI Complete: Generate AI insights using AI_COMPLETE() in Python stored procedures 2. Cortex Analyst: Build semantic models for natural language to SQL conversion 3. Cortex Search: Create vector search services for document retrieval (RAG) 4. Cortex Agents: Combine multiple AI tools (Analyst + Search) into unified agents 6. Streamlit in Snowflake: Deploy multi-page applications with Git integration 7. Python Stored Procedures: Automate workflows using Snowpark Python 8. Time-Series Analysis: Process and analyze industrial sensor data 9. Alert Management: Implement deduplication, severity classification, and workflow automation 10. Multimodal AI: Process both text and images using AI_COMPLETE() with vision models

---

Prerequisites

Snowflake Account Requirements

• Snowflake Account: Free trial account or existing account
• Role: ACCOUNTADMIN (for initial setup only - creates IPA_DEMO_ROLE for ongoing usage)

---

Quick Start

Step 1: Run Setup SQL

1. Open Snowsight SQL Worksheet 2. Copy and paste entire contents of `scripts/setup.sql` 3. Run the script as ACCOUNTADMIN

• Creates database with sample SCADA data
• Creates IPA_DEMO_ROLE with proper access controls
• Deploys Streamlit app from Git
• Creates and registers Cortex Agent

Step 2: Access the Application

1. In Snowsight, navigate to Projects → Streamlit 2. Click on Intelligent Production Assistant to open the application 3. Start exploring the 5-page dashboard!

*Note: After setup, users with IPA_DEMO_ROLE can access all demo features without requiring ACCOUNTADMIN privileges.*

Step 3: Access Snowflake Intelligence

Once deployed, you can also interact with the IPA_AGENT through Snowflake Intelligence:

1. Navigate to Snowsight → AI & ML → Snowflake Intelligence 2. Look for IPA_AGENT in your available agents 3. Ask questions like:

• "What are the current critical alerts?"
• "Show me H2S readings in Zone-B"
• "Find documents about Well-RP-05 failures"

4. The agent combines SQL queries (via Cortex Analyst) with document search (via Cortex Search)

---

Technical Implementation

The setup script automatically creates:

• Database: IPA with 3 schemas (SCADA_CORE, KNOWLEDGE_BASE, APP)
• Tables: Asset master, tags, time-series data, documents, alerts, permits, financials
• Detection Services: Sentinel, Guardian, Fiscal (Python stored procedures)
• Cortex Search: Document retrieval service for RAG
• Cortex Agent: IPA_AGENT (true Cortex Agent with text-to-SQL + document search)
• Streamlit App: 5-page dashboard deployed from Git
• Snowflake Intelligence: Agent registered for natural language queries

---

📱 Application Pages

1. 🖥️ Mission Control

Purpose: Unified command center showing alerts from all 3 detection services

Features:

• Real-time alert dashboard with critical/warning/info counts
• Scenario generators for demo storytelling (inject synthetic failures)
• "Run All Services" button to trigger Sentinel, Guardian, and Fiscal scans
• Live alert feed with acknowledge workflow

Demo Flow: 1. Click "🤖 Run All Services" to scan current conditions 2. Or click scenario buttons to inject test events:

• "🔴 Sentinel – Rod Pump Failure"
• "⚠️ Guardian – H₂S Safety Event"
• "💰 Fiscal – Cost Variance"

3. Review alerts in live feed, acknowledge critical items

---

2. 🛡️ Production Sentinel

Purpose: Asset-specific diagnostics for rod pumps and drilling rigs

Features:

• Asset selector with 4 key assets (RIG-9, WELL-A10, WELL-B03, WELL-RP-05)
• Rod Pump: Dynamometer chart shows failure patterns (fluid pound → rod part)
• Drilling Rig: ROP (Rate of Penetration) analysis detects NPT (Non-Productive Time)
• Time range selector (24h, 48h, 7 days)
• Active alert count per asset

Demo Flow: 1. Select "WELL-RP-05" → See rod pump failure pattern (load drops to zero) 2. Select "RIG-9" → See 6-hour NPT event (zero ROP) with business impact calculation 3. Click "Run All Services" on Mission Control → Sentinel generates alerts for these anomalies

---

3. 🦺 HSE Guardian

Purpose: Safety monitoring with permit-gas correlation

Features:

• Real-time H2S gas monitoring
• Active work permit tracking with safety constraints
• Permit violation detection (e.g., Hot Work Permit allows max 10 ppm H2S)
• Zone status map showing geographic zones

Demo Flow: 1. Observe H2S spike > 15 ppm in Zone-B (violates Hot Work Permit HWP-2026-001) 2. See "STOP WORK REQUIRED" alert 3. Review permit details and corrective actions 4. Check Guardian alert history

Value Prop: Prevents fatal H2S incidents through real-time safety barrier monitoring

---

4. 👁️ Visual Inspector

Purpose: AI-powered…