Working with Threads in Gen AI Builder

What are Threads

Threads in Gen AI Builder represent individual conversations or interaction histories, typically with an AI Assistant. They allow you to review past interactions, understand message flow, and analyze Assistant behavior.

A Thread captures:

• The entire sequence of messages between a user and an AI Assistant.
• Associated metadata (timestamps, IDs, context state).
• System-level details (Tools used, Rulesets applied, memory state).

In short: Threads provide the conversation memory and history that enables analysis, debugging, and state persistence in your AI Factory applications.

For a deep dive, see: Threads explained.

Why use Threads

• Review conversation history: Understand exactly what the user and Assistant said.
• Debug behavior: Analyze how Assistants behaved in specific interactions.
• Analyze user needs: See patterns of user queries and intents.
• Support compliance audits: Provide a record of AI responses.
• Manage context state: Track conversation state for Assistants with memory.

Threads are essential for production-grade AI applications where quality, compliance, and traceability matter.

When to use Threads

Use Threads when you need to:

• Review recent interactions for QA or user support.
• Audit Assistant responses for compliance.
• Debug unexpected Assistant behavior.
• Identify points of friction or failure in conversations.
• Analyze common user questions and improve knowledge coverage.
• Manage or reset conversation context for long-running Assistants.

How Threads fit into Gen AI Builder

User Input → Assistant → Retriever → Knowledge Bases → Retrieved Content
→ Rulesets → Behavioral Guidance
→ Tools → Action Execution
→ Memory → Context Maintenance
→ LLM → Response → User Output → Thread recorded

Every conversation with an Assistant:

• Creates or continues a Thread.
• Records:
• Messages
• Message metadata
• Tool calls and results
• Memory state (if applicable)

Key features of Threads

• Conversation history: Full message log with timestamps.
• Searchable: Find Threads by name, alias, or content.
• Thread metadata: ID, created/updated times, Assistant association.
• Message metadata: Message ID, content, Tool calls.
• Context state: Memory and conversation context tracked across turns.
• QA and debugging: Inspect message flow and Assistant behavior.

Typical patterns of use

Conversation QA

• Review recent Threads to verify tone, accuracy, and compliance.

Debugging

• Investigate specific user complaints or unexpected behaviors.

Compliance auditing

• Provide conversation transcripts for audit purposes.

Memory management

• Track and manage memory state across Threads.

Conversation analytics

• Identify trends and common user questions across Threads.

Getting started

See View and manage Threads for a full step-by-step guide.

Typical workflow:

1. Navigate to Threads.
2. Search or browse to find relevant Thread.
3. Open Thread and review:

• Conversation flow
• Assistant responses
• Tool invocations
• Context state

4. Take action:

• Edit or delete Thread (if needed).
• Export Thread for QA or audit.
• Use findings to improve Assistant configuration.

Related topics

• Threads explained
• View and manage Threads
• Assistants explained
• Structures explained
• Rulesets explained
• AI Factory Concepts
• Hybrid Manager: Using Gen AI Builder