🧠 LLM Workflow

A practical guide for building effective LLM-powered systems using pure Python—no heavy frameworks, just proven patterns and reusable code snippets.

📚 About This Repository

This repository provides real-world examples and design patterns for building reliable, composable LLM workflows. Whether you're working on agents, assistants, or task-specific pipelines, you'll find ready-to-use patterns like:

• 🔗 Prompt Chaining
• 🔀 Request Routing
• ⚡ Parallel Execution
• 🤖 Orchestrator–Worker Architecture

💡 Inspired by:

• Building AI Agents in Pure Python - by Dave Ebbelaar

🔗 Learn more about the theory and practice behind these patterns:

• Building Effective Agents - Anthropic's blog post

---

🧩 Contents

Part 1: Core LLM Techniques

• Basic LLM calls
• Structured outputs
• Tool usage
• Retrieval integration

Part 2: Workflow Patterns

• Prompt chaining
• Routing
• Parallel guardrails
• Orchestrator–Worker design

🔧 Requirements: Python basics, OpenAI SDK, and API key

---

🔁 Example Use Case: Calendar Assistant

Includes implementations using the above patterns to build a functional calendar assistant:

✅ Prompt Chaining

Break down complex tasks into structured steps for better reliability.

🔀 Routing

Direct user input to specialized logic based on intent.

⚡ Parallelization

Run multiple LLM checks (e.g., security + calendar validation) concurrently.

🤖 Orchestrator–Worker

Coordinate planning, writing, and review stages with specialized agents.

---

⚙️ Setup Guide

Follow these steps to set up the project locally:

1. Clone the repository:

   git clone <repository-url>
   cd llm-workflow-cookbook
   

2. Create and activate a virtual environment:

    python -m venv .venv
    source venv/bin/activate  # On Windows use `.venv\Scripts\activate`
   

3. Install dependencies:

   pip install -r requirements.txt
   

4. Set up environment variables: Create a .env file in the root directory and add your OpenAI API key:

    OPENAI_API_KEY=your-api-key-here
   

---