Anthropic Developer Courses

repository·master·Indexed 28 days ago

https://github.com/anthropics/courses

Interactive, hands-on tutorials for developers to master Claude models. Includes courses on Anthropic API fundamentals, prompt engineering (including Amazon Bedrock integration), real-world prompting, prompt evaluations using promptfoo, and tool use workflows.

Tokens
101.7K
Snippets
182
Records
327
Agent score
94%

What's inside anthropics-courses

  1. Explore Anthropic educational courses

    master

    This repository provides a series of five educational courses designed to teach different aspects of working with Anthropic's Claude models. It is recommended to complete them in the following order:

    1. Anthropic API fundamentals: Covers Claude SDK essentials, including API key management, model parameters, multimodal prompts, and streaming responses.
    2. Prompt engineering interactive tutorial: A step-by-step guide to key prompting techniques.
    3. Real world prompting: Focuses on incorporating prompting techniques into complex, real-world scenarios.
    4. Prompt evaluations: Teaches how to write production-grade prompt evaluations to measure prompt quality.
    5. Tool use: Covers implementation of tool use workflows with Claude.

    Note on Model Usage: To minimize API costs, these courses often use Claude 3 Haiku. You may use other Claude models if preferred.

  2. Learn Anthropic API fundamentals via notebook tutorials

    master

    The anthropic_api_fundamentals course provides a series of Jupyter notebook tutorials designed to teach the essentials of working with Claude models and the Anthropic SDK. The curriculum covers the following topics:

    • Getting Started: Obtaining an API key and making simple requests.
    • Messages Format: Working with the standard messages API structure.
    • Model Comparison: Comparing the capabilities and performance of different models in the Claude family.
    • Model Parameters: Understanding and configuring model parameters.
    • Streaming: Implementing and working with streaming responses.
    • Vision Prompting: Using vision capabilities with Claude models.
  3. Explore the Prompt Engineering Course Curriculum

    master

    The tutorial is divided into three difficulty levels and an appendix covering advanced topics:

    Beginner

    • Chapter 1: Basic Prompt Structure
    • Chapter 2: Being Clear and Direct
    • Chapter 3: Assigning Roles

    Intermediate

    • Chapter 4: Separating Data from Instructions
    • Chapter 5: Formatting Output & Speaking for Claude
    • Chapter 6: Precognition (Thinking Step by Step)
    • Chapter 7: Using Examples

    Advanced

    • Chapter 8: Avoiding Hallucinations
    • Chapter 9: Building Complex Prompts (Chatbots, Legal Services, Financial Services, and Coding)

    Appendix (Advanced Methods)

    • Chaining Prompts
    • Tool Use
    • Empirical Performance Evaluations
    • Search & Retrieval
  4. Navigate the Prompt Engineering Interactive Tutorial curriculum

    master

    The Prompt Engineering Interactive Tutorial is a structured course designed to teach optimal prompt engineering for Claude. The course is divided into three difficulty levels and should be completed in chapter order. Each chapter includes a lesson and practical exercises.

    Course Structure:

    Beginner

    • Chapter 1: Basic Prompt Structure
    • Chapter 2: Being Clear and Direct
    • Chapter 3: Assigning Roles

    Intermediate

    • Chapter 4: Separating Data from Instructions
    • Chapter 5: Formatting Output & Speaking for Claude
    • Chapter 6: Precognition (Thinking Step by Step)
    • Chapter 7: Using Examples

    Advanced

    • Chapter 8: Avoiding Hallucinations
    • Chapter 9: Building Complex Prompts (Industry Use Cases: Chatbots, Legal, Financial, Coding)
    • Appendix: Beyond Standard Prompting (Chaining, Tool Use, Search & Retrieval)

    Learning Features:

    • Example Playground: Located at the bottom of each lesson for experimenting with prompt changes.
    • Answer Key: A Google Sheet is provided for verifying exercises.
    • Model Note: This tutorial uses Claude 3 Haiku. While Haiku is used for speed and cost-efficiency, users can apply these techniques to Claude 3 Sonnet and Claude 3 Opus.
  5. Learn Claude tool use via the Tool Use tutorial series

    master

    The Tool Use tutorial is a six-lesson course designed to teach developers how to implement tool use workflows with Claude. The course is sequential; it is highly recommended to follow the lessons in order as each builds upon the concepts from the previous one.

    Curriculum Path:

    1. Tool use overview: Foundational concepts.
    2. Your first simple tool: Implementing basic tool functionality.
    3. Forcing JSON with tool use: Techniques for structured outputs.
    4. The complete tool use workflow: End-to-end implementation.
    5. Tool choice: Controlling how Claude selects tools.
    6. Building a chatbot with multiple tools: Managing complex interactions with multiple tool definitions.
  6. Start the Prompt Engineering Interactive Tutorial (Bedrock Edition)

    master

    This interactive tutorial is designed to teach prompt engineering for Claude using Amazon Bedrock. The course is structured into 9 chapters with accompanying exercises and is intended to be completed in sequential order.

    Key Learning Objectives:

    • Master basic prompt structure.
    • Address common failure modes using '80/20' techniques.
    • Understand Claude's strengths and weaknesses.
    • Build prompts for common use cases from scratch.

    Implementation Details:

    • Model used: The tutorial uses Claude 3 Haiku (the smallest, fastest, and cheapest model).
    • Playgrounds: Each lesson includes an "Example Playground" for experimentation.
    • Answer Key: An answer key is provided. Note that while the key is structured for 1P (Anthropic) API requests, the solutions are identical for Bedrock usage.

    To begin, navigate to 01_Basic Prompt Structure.

  7. Complete the Prompt Evaluations course

    master

    The Prompt Evaluations course consists of nine sequential lessons designed to teach you how to implement evaluations in workflows using the Anthropic API. It is recommended to follow the lessons in order, starting with 'Evaluations 101', as each lesson builds on the concepts from the previous ones.

    Course Curriculum:

    1. Evaluations 101: Foundational concepts.
    2. Writing human-graded evals with Anthropic's Workbench: Using the Workbench for manual evaluation.
    3. Writing simple code-graded evals: Implementing programmatic checks.
    4. Writing a classification eval: Evaluating categorical outputs.
    5. Promptfoo for evals: an introduction: Introduction to using the promptfoo tool.
    6. Writing classification evals with promptfoo: Using promptfoo for classification tasks.
    7. Custom graders with promptfoo: Implementing specialized grading logic.
    8. Model-graded evals with promptfoo: Using LLMs to grade outputs.
    9. Custom model-graded evals with promptfoo: Advanced model-based evaluation patterns.