microsoft/poml

Features

• Prompt Templates - Provides a system for defining reusable prompt structures using a semantic markup language.
• Structured Prompting Tools - Provides a framework for designing complex LLM inputs using modular templates, variables, and conditional logic.
• Data-Driven Prompt Generation - Uses a templating engine with variables, loops, and conditionals to create complex, data-driven prompts.
• Prompt Templates - Implements reusable instruction blocks and dynamic variables to generate consistent and structured model inputs.
• Step-wise Task Specifications - Defines the primary action or directive the model must perform, including concrete completion steps.
• Prompt Persona Definitions - Assigns a specific role or perspective for the model to assume to influence tone and context.
• Persona and Behavioral Instructions - Defines the model's persona, specific tasks, and stepwise instructions to direct the response.
• LLM Tooling Integrations - Provides a configuration interface to define function schemas and manage requests and responses between models and external tools.
• LLM Tool Definitions - Provides tools for creating and exposing functional API schemas for discovery and execution by language models.
• Model Output Formatting - Defines schemas for generated output to ensure the model returns data in a predictable and parsable format.
• Schema Enforcement Layers - Provides mechanisms to force model outputs to adhere to strict JSON schemas for predictable, machine-parsable responses.
• Modular AI Components - Organizes prompts into reusable, swappable building blocks that can be composed from separate files.
• Conversational Message Schemas - Organizes exchanges between system, human, and AI participants using standardized message schemas.
• Prompt Templates - Creates structured, reusable prompt building blocks that can be referenced across different templates.
• Integration SDKs - Ships a set of libraries for integrating structured prompt logic and rendering workflows across multiple programming languages.
• Markup Parsing - Converts structured prompt markup into rich content or speaker messages for language models.
• Prompt Management Workflows - Implements a development workflow for authoring, testing, and managing structured prompt templates within a code editor.
• Output Formatting Constraints - Dictates the specific structural format for the response, such as JSON, XML, or CSV.
• Prompt Formatting - Provides standardized structures and templates to produce chat messages or plain text strings compatible with AI frameworks.
• Prompt Rendering Pipelines - Processes markup through a pipeline of parsing, variable substitution, and serialization to produce final strings or chat messages.
• Structural Prompt Design - Organizes prompt components and data into a structured markup format for consistent intermediate representation.
• Structured Output Enforcements - Enforces data models or schemas on task outputs to ensure reliable, machine-parsable processing.
• LLM Observability and Evaluation - Captures and logs final rendered prompts and context variables for debugging and performance auditing.
• LLM Schema Outputs - Transforms data models into strict JSON schemas to enforce structured and validated output from LLMs.
• Intermediate Representations - Transforms semantic XML-like syntax into a structured internal tree for consistent processing across different models.
• Structured JSON Generation - Enforces strict JSON schema adherence, including required properties, during the generative process of the model.
• Prompt Template Injection - Evaluates variables and expressions to insert dynamic content into a prompt template.
• Prompt Variable Definitions - Sets values and imports data from external sources to be reused across the prompt structure.
• Message Role Wrapping - Wraps content as system, human, or AI messages to maintain clear speaker boundaries in model inputs.
• Context-Aware Variable Resolution - Resolves dynamic variables and logical expressions within prompts by injecting runtime data objects.
• Intermediate Representations - Employs a hierarchical intermediate representation using semantic tags to organize prompt text and metadata.
• Logic and Presentation Separation - Decouples core prompt logic from visual presentation using an external stylesheet-based approach.
• Prompt Composition Patterns - Injects contents of external files into main templates to organize prompts into modular, reusable pieces.
• Prompt List Iteration - Repeats specific prompt elements for every item in a list using iteration attributes.
• Conditional Prompt Rendering - Shows or hides prompt elements based on the truth value of a variable or expression.
• Application Integration SDKs - Provides language-specific SDKs to embed structured prompt workflows and rendering capabilities into external applications.
• Conversation Simulators - Models conversational exchanges between humans and AI, including tool requests and responses.
• External Tool Integration - Provides capabilities for the model to interact with external APIs and data sources dynamically at runtime.
• Function Signature Definitions - Defines function signatures and parameters that allow an AI model to interact with external code.
• Model Parameters - Allows setting execution parameters to control the performance and reasoning characteristics of the language model.
• Community Prompt Galleries - Provides a gallery for storing, editing, and organizing built-in and custom prompt templates.
• Prompt Engineering Patterns - Includes a browser for exploring and implementing proven prompt engineering patterns and structures.
• Prompt Debugging Utilities - Captures the final rendered prompt structure and context variables for debugging and unit testing.
• Prompt Serialization Controls - Specifies serialization modes and markup languages to customize how nested content renders in the final prompt.
• Conditional Logic - Includes or excludes prompt sections based on logical conditions evaluated against the runtime context.
• Prompt Template Testing - Validates reusable prompt structures by sending rendered outputs to model providers without running application code.
• Reasoning Steering - Inserts helpful tips and guided questions to steer the model toward the correct output.
• Tool-Calling Schemas - Converts schema definitions into compatible tool declarations to enable automated function calling.
• Few-Shot Pattern Exemplification - Includes sample input-output pairs to demonstrate the desired structure, tone, and detail for the response.
• Structured Prompt Data Types - Stores arbitrary data values and JSON objects within prompt markup to provide programmatic context.
• Hierarchical Data Structures - Renders information using tables and directory trees to preserve nested layouts for better model comprehension.
• Schema-Adherent Generation - Includes a tool that converts data models into strict JSON schemas to enforce schema-adherent model generation.
• Code Completion - Offers intelligent suggestions for tags and attributes to ensure prompt files adhere to the required syntax.
• Prompt Rendering Engines - Renders structured prompts into final strings using software development kits for various programming languages.
• IDE Integrated Tooling - Integrates syntax highlighting, auto-completion, and real-time previews directly within the code editor.
• Tool Schema Management - Defines tool schemas and manages the exchange of requests and responses between the prompt and the model.
• Content Importance Prioritization - Assigns importance levels to text segments to determine which content is preserved when space is limited.
• Template Variable Replacement - Replaces placeholders within prompt templates with dynamic values from a global context object during rendering.
• Data Serialization Formats - Serializes files and data structures into model-optimized formats like CSV, XML, or plain text.
• Model Response Validation Wrappers - Verifies model responses by displaying them within the editor to validate output quality and format.
• Logic-Rendering Decouplers - Separates core prompt logic from presentation using a styling system to modify output formats independently.
• Observability Platform Exporters - Transmits captured prompt data to external monitoring frameworks for centralized analysis and auditing.
• Observability Platform Log Exporting - Records prompt calls and results and exports them to external observability platforms for monitoring.
• Prompt and Agent Versioning - Publishes prompt sources, stylesheets, and context as versioned entities to track changes over time.
• Real-time Syntax Checking - Detects syntax errors and validates prompt attributes in real-time within the editor.
• Token-Aware Truncation - Truncates text based on character or token counts and inserts markers to indicate where content was removed.
• Markup Element Styling - Applies consistent attributes to markup elements using a JSON stylesheet for uniform visual and structural style.
• Template Element Visibility - Toggles specific prompt components based on feature flags or conditional logic to customize output.
• Expression Evaluators - Evaluates dynamic logic and data transformations within prompt templates using provided context files.
• Side-by-Side Editor Previews - Provides a side-by-side live rendering of prompt structures and styles that updates automatically during editing.
• Scroll Synchronization - Synchronizes scroll position and selection highlighting between the prompt editor and its live preview pane.
• Component Styling - Defines global and local properties for prompt markup syntax and text formats using a centralized stylesheet.
• API Parameter Generation - Converts structured prompt files into dictionaries containing messages and runtime parameters compatible with chat completion APIs.

Star history