# Examples

This directory contains runnable examples demonstrating key features of Agency Swarm v1.x.

## Core Functionality
- **`multi_agent_workflow.py`** – Multi-agent collaboration with validation patterns
- **`agency_context.py`** – Sharing data between agents using agency context
- **`streaming.py`** – Real-time streaming responses
- **`guardrails.py`** – Input and output guardrails
- **`custom_persistence.py`** – Chat history persistence between sessions
- **`tools.py`** – Tool patterns: BaseTool and @function_tool with validation

## File Handling & Search
- **`agent_file_storage.py`** – Vector store creation and FileSearch tool usage
- **`message_attachments.py`** – File processing and message attachments
- **`web_search.py`** – Domain-filtered WebSearchTool example with source URL extraction

## Agent Communication
- **`custom_send_message.py`** – Custom SendMessage configurations and patterns
- **`interactive/hybrid_communication_flows.py`** – Combining SendMessage and handoffs in a software development workflow

## User Interfaces
- **`agency_visualization.py`** – Interactive HTML visualization
- **`interactive/tui.py`** – Terminal UI chat interface (sets up the matching terminal app automatically on first run and shows a short setup message)
- **`interactive/copilot_demo.py`** – Copilot UI chat interface

## Integration & External Services
- **`fastapi_integration/`** – FastAPI server and client examples
  - `server.py` – FastAPI server with streaming support
  - `client.py` – Client examples for testing endpoints
- **`mcp_servers.py`** – Using tools from MCP servers (local and hosted)
- **`connectors.py`** – Google Calendar integration using OpenAI hosted tools

## Model Providers
- **`third_party_models.py`** – Using third-party models (Claude, Gemini, Grok) via LiteLLM

## Observability
- **`observability.py`** – OpenAI, Langfuse and AgentOps tracing integration

Run any file with `python examples/<name>.py` after setting your `OPENAI_API_KEY`.