Agentic Apps Monorepo

Collection of experimental and reference agent applications built on the A2A protocol and SDK. Each subfolder is a self-contained agent system or demo (some with autonomous LLM behavior, real-time dashboards, or network orchestration).

Contents

Project	Description	Key Tech
tourist_scheduling_system/	Multi-agent tourist scheduling with scheduler, UI dashboard, and autonomous guide/tourist agents	A2A SDK, FastAPI, WebSockets, OpenAI/Azure OpenAI
remote_agent_agp/	Remote agent gateway & client demos	A2A SDK, Docker Compose
mailcomposer/	AI-driven email composer workflow	A2A SDK, LangGraph
marketing-campaign/	Multi-agent marketing campaign planner	A2A SDK
email_reviewer/	Agent that reviews emails for quality & tone	A2A SDK
network_of_assistants/	Network-of-assistants stack (moderator, math, file, web, user proxy)	FastAPI, Docker, A2A SDK
weather_vibes_agp/	Weather plus mood/vibe agent demo	A2A SDK
api_bridge_agent_demos/	MCP / bridge agent examples	A2A SDK, MCP

Quick Start (Tourist Scheduling System)

The most feature-complete example lives in tourist_scheduling_system/. It demonstrates:

• Scheduler coordination of guide offers and tourist requests
• Real-time UI agent (web + WebSocket)
• Autonomous LLM-powered agents with heuristic fallback when cloud creds absent

Install

git clone https://github.com/agntcy/agentic-apps.git
cd agentic-apps/tourist_scheduling_system
python -m venv .venv
source .venv/bin/activate  # macOS/Linux
pip install -e .

Basic Run

In one terminal start the scheduler:

PYTHONPATH=src python src/agents/scheduler_agent.py --host localhost --port 10010

In a second terminal start the UI dashboard:

PYTHONPATH=src python src/agents/ui_agent.py --host localhost --port 10011 --a2a-port 10012

Send a guide offer and tourist request:

PYTHONPATH=src python src/agents/guide_agent.py --scheduler-url http://localhost:10010 --guide-id guide-1
PYTHONPATH=src python src/agents/tourist_agent.py --scheduler-url http://localhost:10010 --tourist-id tourist-1

Open http://localhost:10011 in your browser to view live activity.

Unified Script Demo

Use the helper script to orchestrate everything (auto detects running ports and can add autonomous agents):

cd tourist_scheduling_system
PYTHONPATH=src scripts/run_with_ui.sh --scheduler-port 10010 --ui-web-port 10011 --ui-a2a-port 10012

Add autonomous pair for a 1 minute smoke test:

PYTHONPATH=src scripts/run_with_ui.sh \
	--scheduler-port 10010 \
	--ui-web-port 10011 \
	--ui-a2a-port 10012 \
	--autonomous \
	--auto-duration 1 \
	--auto-guide-id ag-auto \
	--auto-tourist-id at-auto

Azure OpenAI (Optional)

Set these before running autonomous mode to enable LLM reasoning; otherwise heuristic fallback is used:

export AZURE_OPENAI_API_KEY=...
export AZURE_OPENAI_API_VERSION=2024-08-01-preview
export AZURE_OPENAI_ENDPOINT=https://your-endpoint.openai.azure.com/
export AZURE_OPENAI_DEPLOYMENT_NAME=gpt-4o

Troubleshooting

Symptom	Cause	Fix
Exit Code: 127 running agents	Virtual env not active / path mismatch	source .venv/bin/activate then rerun; ensure working directory is tourist_scheduling_system
Import errors after refactor	Missing PYTHONPATH=src	Prefix command with PYTHONPATH=src or install in editable mode (pip install -e .)
Port already in use	Previous run still active	Use lsof -i :10010 / kill PID or rely on script reuse logic
No LLM decisions, warning about Azure vars	Cloud creds not set	Export Azure env vars or ignore (heuristics operate)
UI blank page	Wrong port or UI agent not started	Confirm ui_agent.py running on specified web port

Development

Format & type check:

black tourist_scheduling_system/src tourist_scheduling_system/tests
isort tourist_scheduling_system/src tourist_scheduling_system/tests
mypy tourist_scheduling_system/src
pytest tourist_scheduling_system/tests

License

Apache License 2.0 (see LICENSE).

Contributing

See CONTRIBUTING.md for guidelines; feel free to open issues or PRs for new agent demos.