Which integrations are available for this server?

Used for environment variable management to configure the PromptLab server with API keys and other settings Enables version control and contribution workflows for the PromptLab project Hosts the repository and provides collaboration features for contributing to PromptLab Integrates with [LangChain](/mcp/servers/integrations/langchain) to provide structured prompt templates and processing workflows Leverages [LangGraph](/mcp/servers/integrations/langgraph) for building the workflow that transforms user queries into optimized prompts Provides LLM capabilities for the prompt enhancement engine, including content classification and parameter extraction Provides a template system for defining and managing prompt templates without coding knowledge

PromptLab: AI Query Enhancement with MLflow Integration

PromptLab is an intelligent system that transforms basic user queries into optimized prompts for AI systems using MLflow Prompt Registry. It dynamically matches user requests to the most appropriate prompt template and applies it with extracted parameters.

🔍 Overview

PromptLab combines MLflow Prompt Registry with dynamic prompt matching to create a powerful, flexible system for prompt engineering:

Centralized Prompt Management: Store, version, and manage prompts in MLflow
Dynamic Matching: Intelligently match user queries to the best prompt template
Version Control: Track prompt history with production and archive aliases
Extensible: Easily add new prompt types without code changes

Related MCP server: TypeScript Prompt MCP Server

🏗️ Architecture

The system consists of three main components:

Prompt Registry (register_prompts.py) - Tool for registering and managing prompts in MLflow
Server (promptlab_server.py) - Server with dynamic prompt matching and LangGraph workflow
Client (promptlab_client.py) - Lightweight client for processing user queries

Workflow Process

PromptLab Workflow

Prompt Registration: Register prompt templates in MLflow with versioning and aliasing
Prompt Loading: Server loads all available prompts from MLflow at startup
Query Submission: User submits a natural language query via the client
Intelligent Matching: LLM analyzes the query and selects the most appropriate prompt template
Parameter Extraction: System extracts required parameters from the query
Template Application: Selected template is applied with extracted parameters
Validation & Adjustment: Enhanced prompt is validated and adjusted if needed
Response Generation: Optimized prompt produces a high-quality response

📂 Code Structure

promptlab/ ├── promptlab_server.py # Main server with LangGraph workflow ├── promptlab_client.py # Client for processing queries ├── register_prompts.py # MLflow prompt management tool ├── requirements.txt # Project dependencies ├── advanced_prompts.json # Additional prompt templates └── README.md # Project documentation

Core Components:

`register_prompts.py`

Purpose: Manages prompts in MLflow Registry
Key Functions:
- register_prompt(): Register a new prompt or version
- update_prompt(): Update an existing prompt (archives previous production)
- list_prompts(): List all registered prompts
- register_from_file(): Register multiple prompts from JSON
- register_sample_prompts(): Initialize with standard prompts

`promptlab_server.py`

Purpose: Processes queries using LangGraph workflow
Key Components:
- load_all_prompts(): Loads prompts from MLflow
- match_prompt(): Matches queries to appropriate templates
- enhance_query(): Applies selected template
- validate_query(): Validates enhanced queries
- LangGraph workflow: Orchestrates the query enhancement process

`promptlab_client.py`

Purpose: Provides user interface to the service
Key Features:
- Process queries with enhanced prompts
- List available prompts
- Display detailed prompt matching information

🚀 Getting Started

Prerequisites

Python 3.12
Dependencies in requirements.txt
OpenAI API key for LLM capabilities

Installation

# Clone the repository git clone https://github.com/iRahulPandey/PromptLab.git cd PromptLab # Install dependencies pip install -r requirements.txt # Set environment variables export OPENAI_API_KEY="your-openai-api-key"

Registering Prompts

Before using PromptLab, you need to register prompts in MLflow:

# Register sample prompts (essay, email, technical, creative) python register_prompts.py register-samples # Register additional prompt types (recommended) python register_prompts.py register-file --file advanced_prompts.json # Verify registered prompts python register_prompts.py list

Running the Server

# Start the server python promptlab_server.py

Using the Client

# Process a query python promptlab_client.py "Write a blog post about machine learning" # List available prompts python promptlab_client.py --list # Enable verbose output python promptlab_client.py --verbose "Create a presentation on climate change"

📋 Prompt Management

Available Prompt Types

PromptLab supports a wide range of prompt types:

Prompt Type	Description	Example Use Case
essay_prompt	Academic writing	Research papers, analyses
email_prompt	Email composition	Professional communications
technical_prompt	Technical explanations	Concepts, technologies
creative_prompt	Creative writing	Stories, poems, fiction
code_prompt	Code generation	Functions, algorithms
summary_prompt	Content summarization	Articles, documents
analysis_prompt	Critical analysis	Data, texts, concepts
qa_prompt	Question answering	Context-based answers
social_media_prompt	Social media content	Platform-specific posts
blog_prompt	Blog article writing	Online articles
report_prompt	Formal reports	Business, technical reports
letter_prompt	Formal letters	Cover, recommendation letters
presentation_prompt	Presentation outlines	Slides, talks
review_prompt	Reviews	Products, media, services
comparison_prompt	Comparisons	Products, concepts, options
instruction_prompt	How-to guides	Step-by-step instructions
custom_prompt	Customizable template	Specialized use cases

Registering New Prompts

You can register new prompts in several ways:

1. From Command Line

python register_prompts.py register \ --name "new_prompt" \ --template "Your template with {{ variables }}" \ --message "Initial version" \ --tags '{"type": "custom", "task": "specialized"}'

2. From a Template File

# Create a text file with your template echo "Template content with {{ variables }}" > template.txt # Register using the file python register_prompts.py register \ --name "long_prompt" \ --template template.txt \ --message "Complex template"

3. From a JSON File

Create a JSON file with multiple prompts:

{ "prompts": [ { "name": "prompt_name", "template": "Template with {{ variables }}", "commit_message": "Description", "tags": {"type": "category", "task": "purpose"} } ] }

Then register them:

python register_prompts.py register-file --file your_prompts.json

Updating Existing Prompts

When you update an existing prompt, the system automatically:

Archives the previous production version
Sets the new version as production

python register_prompts.py update \ --name "essay_prompt" \ --template "New improved template with {{ variables }}" \ --message "Enhanced clarity and structure"

Viewing Prompt Details

# List all prompts python register_prompts.py list # View detailed information about a specific prompt python register_prompts.py details --name "essay_prompt"

🛠️ Advanced Usage

Template Variables

Templates use variables in {{ variable }} format:

Write a {{ formality }} email to my {{ recipient_type }} about {{ topic }} that includes: - A clear subject line - Appropriate greeting ...

When matching a query, the system automatically extracts values for these variables.

Production and Archive Aliases

Each prompt can have different versions with aliases:

production: The current active version (used by default)
archived: Previous production versions

This allows for:

Rolling back to previous versions if needed
Tracking the history of prompt changes

Custom Prompt Registration

For specialized use cases, you can create highly customized prompts:

python register_prompts.py register \ --name "specialized_prompt" \ --template "You are a {{ role }} with expertise in {{ domain }}. Create a {{ document_type }} about {{ topic }} that demonstrates {{ quality }}." \ --message "Specialized template" \ --tags '{"type": "custom", "task": "specialized", "domain": "finance"}'

🔧 Troubleshooting

No Matching Prompt Found

If the system can't match a query to any prompt template, it will:

Log a message that no match was found
Use the original query without enhancement
Still generate a response

You can add more diverse prompt templates to improve matching.

LLM Connection Issues

If the LLM service is unavailable, the system falls back to:

Keyword-based matching for prompt selection
Simple parameter extraction
Basic prompt enhancement

This ensures the system remains functional even without LLM access.

PromptLab MCP Server