Parallel

A concurrent workflow where multiple agents process the same input simultaneously, ideal for comparative analysis, multi-model testing, or distributed checks.

Multiple agents tackle independent subtasks simultaneously to reduce overall latency.

Overview

Use this pattern when subtasks do not depend on each other and you want faster responses by running them side by side. The outputs can be displayed separately or merged downstream.

Demo Scenario: Trip Planning with Specialized Agents

Two travel specialists work in parallel on the same user request:

• Logistics agent – focuses on flights, hotels, and transportation

• Activities agent – curates attractions, food, and experiences

The demo runs both agents at the same time and then prints a combined plan.

Diagram

Implementation Steps

1. Create two focused agents

   from glaip_sdk import Agent
   
   logistics_agent = Agent(
       name="logistics_agent",
       instruction="You are a logistics specialist. Focus ONLY on flights, hotels, and transport. Be concise.",
   )
   
   activities_agent = Agent(
       name="activities_agent",
       instruction="You are an activities specialist. Focus ONLY on attractions, food, and local experiences. Be concise.",
   )

2. Run both agents simultaneously and collect their outputs

   import concurrent.futures
   from glaip_sdk.utils.rendering.renderer import print_panel
   
   query = "Plan a 5-day trip to Tokyo"
   
   print_panel(
       "Running both agents simultaneously (silent renderers)",
       title="🤖 Logistics & Activities Agents 🤖",
       border_style="green",
   )
   
   with concurrent.futures.ThreadPoolExecutor(max_workers=2) as executor:
       logistics_future = executor.submit(logistics_agent.run, query, renderer="silent")
       activities_future = executor.submit(activities_agent.run, query, renderer="silent")
   
       logistics_result = logistics_future.result()
       activities_result = activities_future.result()
   
   combined = "Logistics:\n" + logistics_result + "\n\nActivities:\n" + activities_result
   print_panel(combined, title="Combined Trip Plan", border_style="green")

3. Clean up demo agents

   # Only needed if agents were deployed
   # logistics_agent.delete()
   # activities_agent.delete()

How to Run

From the repository root glaip-sdk/python/glaip-sdk:

poetry run python examples/multi-agent-system-patterns/parallel/parallel.py

Ensure your .env contains:

OPENAI_API_KEY=your-openai-key-here

Output

────────────────────────────────── Combined Trip Plan ──────────────────────────────────

Logistics:
• Flights: book round-trip flights to HND or NRT.
• Airport transfer: Monorail or taxi from HND; Narita Express from NRT.
• Hotel: stay near Shinjuku or Shibuya for convenient transit.
• Local transport: grab a Suica/Pasmo card for subways and JR lines.

Activities:
• Day 1: Shibuya Crossing, Takeshita Street, Meiji Shrine.
• Day 2: Senso-ji Temple, Sumida River, Ueno Park.
• Day 3: Observation deck in Shinjuku, Omoide Yokocho, Shinjuku Gyoen.
• Day 4: Akihabara electronics crawl, Ginza shopping, depachika food halls.
• Day 5: TeamLab Planets, Odaiba Seaside Park, Tsukiji Outer Market breakfast.

Notes

• Use additional workers in the thread pool (or asyncio) to scale outward for more specialists.

• Combine this pattern with an aggregator if you want a single unified artifact.

• For long-running tasks, monitor usage stats returned by each agent run to detect bottlenecks.

Related Documentation

• Agents guide — Configure instructions and streaming renderers.

• Automation & scripting — Capture transcripts or usage metrics in CI workflows.

• Security & privacy — Apply tool-output and memory policies when sharing results downstream.