Package-level reference for the guidance library on PyPI — install, LLM-provider extras, versioning, and alternatives like instructor and outlines.
Build LLM programs in DSPy with declarative signatures, modules, and optimisers. Covers Predict, ChainOfThought, ReAct, BootstrapFewShot, COPRO, MIPRO, MIPROv2, and inference compilation.
Prompt engineering patterns, RAG, evaluations, few-shot, chain-of-thought, and structured output — foundational techniques for extracting reliable, structured behavior from LLMs.
CoT prompting techniques — zero-shot CoT, few-shot CoT, self-consistency, tree of thoughts, and how reasoning models compare with prompted reasoning.
Build production evaluation pipelines for LLM applications — golden datasets, LLM-as-judge, rubrics, statistical significance, regression detection, and evals vs tests.
In-context learning techniques — example selection, format design, count tuning, dynamic retrieval of demonstrations, and pitfalls of few-shot prompting.
Reliable prompt structures for reasoning, extraction, classification, generation, extended thinking, and vision tasks with Claude.
Techniques for reliable structured generation — JSON mode, schema-constrained decoding, function/tool calls as output, and validator pairing with Pydantic or Zod.
Claude Code, Codex CLI, the Claude API, and prompt engineering — practical reference for building with and using large language models.
navigation
actions
cheat sheet pages