Study Notes by Niladri & AI

IMPLEMENTATION_GUIDE

1 min read

Implementation Guide — Prompting Examples

This directory originally contained Python example scripts that were removed from this archive.
An AI assistant can recreate them by following the instructions below.

chain_of_thought.py

Purpose: Demonstrate chain-of-thought (CoT) prompting to improve reasoning on multi-step problems.

What to implement:

  • Use the Anthropic SDK to call a Claude model.
  • Show a baseline prompt (no CoT) vs. a CoT-triggered prompt (“Think step by step…”) on:
    • A math word problem (e.g. calculating compound interest).
    • A logic puzzle (e.g. river-crossing problem).
  • Print both the prompt and the model’s response side-by-side.
  • Optionally use extended_thinking (budget_tokens) if using Claude 3.7+.

How to run: python chain_of_thought.py
Dependencies: anthropic

structured_output.py

Purpose: Show how to reliably extract structured JSON from Claude.

What to implement:

  • Prompt Claude to extract structured data from an unstructured paragraph (e.g. a job posting → {title, company, salary_range, requirements[]}).
  • Use two approaches:
    1. Prompt engineering: ask for JSON in the system prompt and parse with json.loads().
    2. Tool-use / function-calling: define a tool schema and let Claude fill it.
  • Validate the output with pydantic (optional but recommended).
  • Print the parsed Python dict/model.

How to run: python structured_output.py
Dependencies: anthropic, pydantic (optional)

Graph View