This glossary explains the fundamental concepts and terminology used throughout Latitude to help you better understand the platform.

Prompt Engineering Basics

Prompt

A text-based instruction sent to a language model to guide its response. In Latitude, prompts can include variables, conditionals, loops, and other advanced features through PromptL syntax.

PromptL

Latitude’s prompt templating language that enables dynamic prompt construction with variables, conditions, loops, and other programming concepts.

Language Model (LLM)

The AI system that generates responses based on prompts. Examples include GPT-4, Claude, Gemini, and others.

Provider

The company or service that offers access to language models, such as OpenAI, Anthropic, Google, or Azure.

Latitude Platform Components

Prompt Manager

The core interface for creating, editing, and managing prompts in Latitude. Includes the Prompt Editor, configuration settings, and version control.

Playground

An interactive testing environment where you can run prompts with different inputs and configurations to see how they perform.

AI Gateway

The deployment layer that exposes prompts as API endpoints, making them available for integration with applications.

Evaluations

Tools for assessing prompt performance using different methodologies:

  • LLM-as-Judge: Using AI to evaluate outputs
  • Programmatic Rules: Using code-based criteria
  • Manual Evaluations: Human review of outputs

Logs

Records of all interactions between prompts and language models, including inputs, outputs, metadata, and performance metrics.

Datasets

Collections of input/output pairs used for testing, evaluations, and regression testing.

Advanced Features

Tools

Functions that prompts can call to access external capabilities, such as retrieving information, performing calculations, or taking actions.

Agents

Advanced prompts that can make decisions, use tools, and solve complex problems through multiple interaction steps.

JSON Mode

A configuration setting that enforces structured output formats through JSON schemas.

Cache

A mechanism that stores previously generated responses to improve performance and reduce costs.

Collaboration & Deployment

Version Control

Features for tracking changes to prompts, comparing versions, and managing the prompt lifecycle from draft to production.

Environments

Different contexts (e.g., development, staging, production) for deploying and testing prompts.

Telemetry

Automatic capture of metrics and performance data from prompt interactions.

Webhooks

Integration points that trigger actions in external systems when certain events occur in Latitude.

Development Concepts

SDK

Software Development Kits (available for TypeScript, Python, and other languages) that allow programmatic interaction with Latitude.

API

The Latitude HTTP API that provides access to platform features for custom integrations.

Self-Hosting

Running Latitude on your own infrastructure instead of using the managed cloud version.

Building Blocks

Prompt Template

The base structure of a prompt that includes placeholders for variables.

Snippets

Reusable prompt fragments that can be shared across multiple prompts.

System Message

Special instructions to the model that set context and expectations for behavior.

User Message

Content presented as coming from a user in a conversation with the model.

Assistant Message

Content presented as previously generated by the model in a conversation.

Next Steps

Now that you’re familiar with the core concepts, you can: