🧠 Cortex Retrieval Agent Lab for Slack

Enable Slack-native analytics from structured and unstructured data with Cortex Agents

Overview

This hands-on lab shows you how to build a natural-language agent in Slack using Snowflake Cortex. You'll create a solution that analyzes structured CRM-style data and unstructured meeting transcripts—delivering fast, auditable insights directly in Slack, without writing SQL.

You’ll build a Python Slack bot that dynamically decides whether to run SQL, search a transcript, or summarize meeting outcomes—all using Snowflake-native capabilities.

By the end of this hands-on lab, you'll build a Cortex Agent that:

• Understands natural language questions
• Retrieves the right insights using Cortex Search and Cortex Analyst
• Delivers clear, auditable answers from Snowflake

Lab Workflow

This lab walks through a 3-phase agent orchestration flow:

Phase 1 – Structured analytics with Cortex Analyst

• Load mock CRM-style data into Snowflake
• Define a semantic model with synonyms, filters, metrics, and joins
• Set up Slack integration for natural-language querying
• Run verified queries and see SQL + answers in Slack

Phase 2 – Add Unstructured Retrieval with Cortex Search

• Upload mock meeting transcripts (PDFs) to a Snowflake stage
• Parse and chunk content using PARSE_DOCUMENT + SPLIT_TEXT_RECURSIVE_CHARACTER
• Analyze with Cortex Complete: extract sentiment, risk, next steps, key phrases
• Add transcript facts to the semantic model + verify queries across both data types
• Enable Dynamic Literal Retrieval for smarter question matching

Phase 3 – Full Cortex Agent Orchestration

• Deploy a Python-based Slack agent that:
  • Selects the right tool (Search, Analyst, or both)
  • Generates SQL, runs it, and summarizes results
  • Handles ambiguity and open-ended questions
  • Run the agent locally

Repository Contents

Path	Purpose
notebooks/	Phase 1 & 2 setup notebooks
data/	Mock CRM SQL DDL + six PDF transcripts
semantic-model/	data_agent_semantic_model.yaml (P1) • data_agent_semantic_model2.yaml (P2)
scripts/	slackapp.py Lightweight bot for Phases 1-2 • slackapp_agent.py Final Cortex Agent (Phase 3)
config/	check_deps.py to verify Python packages • app_manifest.json Slack-app manifest • dataagent.env.sample template for Snowflake / Slack creds

ℹ️ This repo ships mock data & code only. Internal Snowflake grading (DORA-xx) and Seismic links remain on the private enablement portal.

Prerequisites

• Snowflake account with Cortex features enabled and warehouse with USAGE & OPERATE on Cortex
• Python 3.10+ environment
• Slack workspace with admin privileges
• Basic understanding of SQL and Python

Getting Started

1. Clone
2. Create & activate venv
3. pip install -r requirements.txt
4. Configure .env
5. Load data & create model via notebook
6. Run agent.py

Resources

• Cortex Analyst Documentation
• Cortex Analyst REST API Documentation
• Create Search Services in Snowflake
• Complete Structured Output Docs
• Cortex Agents Documentation
• Snowflake Cortex Agents Tutorials