Building Effective Dapr Agents

Building Effective Dapr Agents

These examples showcase how to implement Anthropic's research paper [Building Effective Agents](https://www.anthropic.com/research/building-effective-agents) using the [Dapr Agents framework](https://dapr.github.io/dapr-agents/). For more details, see this complementary [blog post](https://www.diagrid.io/blog/building-effective-dapr-agents).

Why Dapr for Agents?

Dapr brings unique advantages to agent implementation:

• Workflow Orchestration: Durable, reliable execution of complex agent tasks
• Pub/Sub Messaging: Event-driven communication between agent components
• State Management: Built-in persistence with pluggable state stores
• Platform Agnostic: Runs locally or on any cloud platform
• LLM Abstraction: Enables reliable, observable interactions with LLMs and allows swapping LLM providers without code changes

Prerequisites

• Python 3.10+
• Docker
• Dapr CLI
• OpenAI API key

Setup

# Create a virtual environment
python3.10 -m venv .venv

# Activate the virtual environment 
# On Windows:
.venv\Scripts\activate
# On macOS/Linux:
source .venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Set up your API key
echo "OPENAI_API_KEY=your_api_key_here" > .env

Agent Patterns Overview

This repository includes implementations of the following patterns:

1. **Augmented LLM**: The basic building block enhanced with memory and tools
2. **Stateful LLM**: Adding persistence and reliability to LLM agents
3. **Prompt Chaining**: Breaking complex tasks into sequential steps
4. **Routing**: Directing inputs to specialized handlers
5. **Parallelization**: Processing multiple aspects of a task simultaneously
6. **Orchestrator-Workers**: Dynamically p