Agentic Pipeline

Agentic pipelines enable dynamic, AI-driven decision-making within your pipeline workflows. Unlike standard pipelines that follow a fixed sequence of steps, agentic pipelines use AI agents to reason and decide actions at runtime.

Overview

GLChat provides a default agentic pipeline that you can use out of the box. However, if you need custom agentic behavior tailored to your specific use case, you can create your own agentic pipeline.

What is an Agentic Pipeline?

An Agentic Pipeline combines the reliability of pipelines with the flexibility of AI agents:

• Pipelines: Provide predictable, auditable, and repeatable data processing (e.g., "Always search the database, then summarize")

• Agents: Use LLMs to reason and decide actions dynamically at runtime (e.g., "Figure out if the user needs a search or a calculation, and do it")

Agentic pipelines allow you to:

• Build workflows that adapt to different scenarios

• Make intelligent decisions based on user input

• Combine multiple tools and capabilities dynamically

• Handle complex, multi-step reasoning tasks

Default Agentic Pipeline

GLChat includes a default agentic pipeline feature called Agentic RAG that you can enable in your Standard RAG pipeline.

Availability

• Pipeline Type: Only available in Standard RAG pipeline

• Current Capability: Currently works only for web search (uses AI Agent to search the web)

  • Note: This feature will be improved in the future to support additional search capabilities

How to Enable

To enable the default agentic pipeline:

1. Open your chatbot's Preset Config in the Admin Dashboard

2. Find the enable_agentic_smart_search configuration option

3. Set it to true (enabled)

Once enabled, your Standard RAG pipeline will use an AI agent to intelligently search the web when processing queries, providing more dynamic and context-aware search results.

Creating Custom Agentic Pipelines

If you need to create your own agentic pipeline with custom behavior, tools, or reasoning logic, refer to the following documentation:

Related Documentation

For detailed information on creating custom agentic pipelines, please refer to:

• GL AIP (GL AI Platform) Quick Start Guide - Get started creating agents with GL AIP. This is the primary resource for building custom agents that can be integrated into your pipelines.

• GL SDK Pipelines and Agents - Learn how to integrate pipelines and agents, including patterns like Pipeline-as-a-Tool and Agent-as-a-Step.

Key Concepts

When building custom agentic pipelines, you'll work with:

1. Agent Orchestrator: Manages agent execution and strategy resolution

2. Agent Strategies: Different execution strategies (Native, AIP, Remote A2A)

3. Tools: Functions that agents can call to perform actions

4. Agent Configuration: Settings that define agent behavior and capabilities

Integration with Pipelines

Agentic pipelines can be integrated with standard pipelines using two main patterns:

1. Pipeline-as-a-Tool: Expose a robust pipeline as a tool for an agent to call

2. Agent-as-a-Step: Embed an agent as a standardized step within a pipeline

For more details on these patterns, see the GL SDK Pipelines and Agents documentation.

When to Use Agentic Pipelines

Use agentic pipelines when:

1. You need dynamic decision-making based on user input

2. The workflow depends on complex reasoning

3. You want to combine multiple tools or capabilities intelligently

4. The sequence of actions should be determined at runtime

Use standard pipelines when:

1. You need predictable, repeatable workflows

2. The sequence of steps is fixed and well-defined

3. You want full control over execution flow

4. Auditability and reproducibility are critical