Documentation Quick Reference

This page provides a quick overview of all available documentation for the Snowflake Cortex Agents SDK.

🚀 Getting Started

Document	Description	Best For
Installation	Setup and dependencies	First-time users
Quick Start	5-minute intro to the SDK	New users
Examples	Code snippets and patterns	Learning by example

📖 Comprehensive Guides

Cortex Analyst Guide

Read the full guide →

Complete guide to generating SQL from natural language using Cortex Analyst.

Key Topics: - Basic SQL generation - Semantic models vs views - Multi-model selection - Streaming responses - Multi-turn conversations - Handling ambiguous questions - Feedback submission - Async operations - Best practices

Example:

from cortex_agents import CortexAnalyst

with CortexAnalyst() as analyst:
    response = analyst.message(
        question="What was Q1 revenue?",
        semantic_view="ANALYTICS.PUBLIC.SALES_VIEW"
    )
    print(response.sql)

Thread Management Guide

Read the full guide →

Comprehensive guide to managing conversation threads for stateful, multi-turn conversations.

Key Topics: - Thread lifecycle (create, list, get, update, delete) - Multi-turn conversations - Branching conversations - Message hierarchy - Async thread operations - Session management patterns - Thread archival - Conversation search - Best practices

Example:

from cortex_agents import CortexAgent

with CortexAgent() as agent:
    # Create thread with origin_app (max 16 bytes)
    thread = agent.create_thread(origin_app="my_app")
    thread_id = thread["thread_id"]

    # Multi-turn conversation
    response1 = agent.run(
        "What was Q1 revenue?",
        agent_name="AGENT",
        thread_id=thread_id,
        parent_message_id=0
    )

    response2 = agent.run(
        "How about Q2?",
        agent_name="AGENT",
        thread_id=thread_id,
        parent_message_id=response1.message_id
    )

Agent with Threads Guide

Read the full guide →

Building sophisticated conversational AI applications by combining Agents with thread management.

Key Topics: - Contextual conversations - Streaming in threads - Multi-path conversations - Advanced session managers - Smart routing - Multi-user support - Production patterns - Response component handling - Feedback integration - Complete production examples

Example:

from cortex_agents import CortexAgent

class ConversationalAgent:
    def __init__(self, agent_name: str):
        self.agent = CortexAgent()
        self.agent_name = agent_name
        self.sessions = {}

    def start_conversation(self, user_id: str):
        thread = self.agent.create_thread()
        self.sessions[user_id] = {
            "thread_id": thread["thread_id"],
            "current_message_id": 0
        }

    def ask(self, user_id: str, question: str):
        session = self.sessions[user_id]
        response = self.agent.run(
            question,
            agent_name=self.agent_name,
            thread_id=session["thread_id"],
            parent_message_id=session["current_message_id"]
        )
        session["current_message_id"] = response.message_id
        return response

📚 API Reference

Module	Description	Link
CortexAgent	Sync agent client	API Docs
AsyncCortexAgent	Async agent client	API Docs
CortexAnalyst	Sync analyst client	API Docs
AsyncCortexAnalyst	Async analyst client	API Docs
AgentResponse	Response wrapper	API Docs
AgentThreads	Thread management	API Docs
AgentEntity	Agent CRUD ops	API Docs
AgentRun	Run operations	API Docs
AgentFeedback	Feedback ops	API Docs

🎯 Common Tasks

I want to...

Generate SQL from natural language → See Cortex Analyst Guide

Build a chatbot with memory → See Agent with Threads Guide

Manage conversation threads → See Thread Management Guide

Stream responses in real-time → See Quick Start - Streaming

Handle multiple concurrent users → See Agent with Threads Guide - Async Multi-User

Submit feedback on responses → See Analyst Guide - Feedback

Work with charts and tables → See Agent with Threads - Response Components

Build a production chatbot → See Agent with Threads - Complete Example

Understand threading and branching → See Thread Management - Branching Conversations

Handle ambiguous questions → See Analyst Guide - Ambiguous Questions

📊 Feature Comparison

Feature	Cortex Agent	Cortex Analyst
Purpose	Agentic workflows with tools	SQL generation
Threads	✅ Full support	❌ Stateless
Tools	✅ Custom functions	❌ N/A
SQL Generation	✅ Via tools	✅ Native
Multi-turn	✅ With threads	✅ Via messages
Streaming	✅ text.delta, tool.delta	✅ sql.delta, text.delta
Async	✅ AsyncCortexAgent	✅ AsyncCortexAnalyst

🔗 External Resources

• Snowflake Cortex Agents Documentation
• Snowflake Cortex Analyst Documentation
• Cortex Analyst Guide

💡 Quick Tips

1. Always use context managers (with statement) for automatic cleanup
2. Track message IDs when working with threads
3. Handle ambiguous questions with suggestions in Analyst
4. Stream responses for better user experience
5. Submit feedback to improve model performance
6. Use async versions for concurrent operations
7. Clean up threads after sessions end
8. Name threads meaningfully for better organization

🆘 Getting Help

1. Check the relevant guide above
2. Review Examples
3. Search API Reference
4. Check the GitHub repository
5. Open an issue