← Back to Directory

Guidance

LLM Orchestrators

Overview

A programming framework by Microsoft that allows developers to control LLMs more effectively than traditional prompting. It uses a templating language to interleave generation, prompting, and control logic.

Guidance, from Microsoft, is a programming framework that interleaves generation, prompting, and control logic in a single template, giving fine-grained control over how a model produces output. It can constrain generation and reduce token waste. It targets developers who want precise, efficient model control.

Key Features

  • Interleaved generation and control logic
  • Constrained generation
  • Token-efficient prompting
  • Works with multiple model backends
  • Open-source

Best For

Developers who want precise, efficient, programmatic control over model output.

Pros & Cons

Pros
  • Fine-grained generation control
  • Reduces token waste
  • Open-source
Cons
  • Templating model has a learning curve
  • Lower-level approach
Advertisement

Pulse Verdict

High-precision AI orchestration. Guidance moves beyond basic strings into a unified control flow that minimizes token waste and maximizes model reliability.

Pricing

Open-source and free; you supply the model.

Pricing changes often — confirm current plans on the official site.

Visit Official Website →

Related Tools

LangChain

The foundational framework for building LLM-powered applications. It provides a massive ecosystem of tools, integrations, and abstractions for managing prompts, memory, and multi-agent chains.

Outlines

A Python library for structured text generation. It allows developers to guide LLM sampling with regular expressions, JSON schemas, or context-free grammars to ensure predictable, machine-readable output.

LMQL

A declarative programming language for Large Language Models. It combines the power of natural language prompting with the precision of Python-like control flow, allowing for constrained generation and efficient token usage.