AI Features

Learn about the intelligent features built into this toolkit.

LLM configuration is required before using any AI feature. Check the guide at Configure LLMs.

You can suggest us more AI features at feature request. We would love to get feedbacks !

Prompt Lockbox isn’t just a management tool; it’s an intelligent toolkit designed to accelerate your workflow. By integrating powerful language models with the toolkit, it automates the most tedious parts of prompt engineering, allowing you to focus on creating brilliant prompts.

The toolkit supports LLMs from various providers like OpenAI, Anthropic, Mistral etc. It also support local models that runs completely offline or you can just use your own custom llm model.

Supported LLMs

The app uses LiteLLM for LLM integration. You can check the complete list of models at litellm models.

Let’s explore some intelligent features the toolkit provides !

Features

Prompt Lockbox is fully open-source — you’re welcome to suggest features or contribute directly at Prompt Lockbox Repo. Get started in just 5 minutes by checking out the Contribution guide!

Here are the key AI features that makes this toolkit awesome.

Automatic Documentation

Reads your prompt’s template, understands its purpose, and automatically writes a concise description and a list of relevant search tags directly into your .yml file—all while preserving your file’s original comments and layout.

See in action

Lets say you have a prompt ‘sql-generator’ with your prompt template written.

...
# Description and tags are empty
description: ""
tags: []
...

The feature automatically reads the prompt and writes the description and tags for you.

...
description: "Generates a SQL query from a database schema and a natural language question."
tags: [sql, database, query-generation, text-to-sql]
...

Usage

CLI

You can use plb prompt document command. Lets say we have a prompt template named ‘email-agent’:

# Apply to a single prompt 
plb prompt document email-agent 

# or document all prompts in the project at once
plb prompt document --all

Python

Python code to document the prompts:

from prompt_lockbox import Project
# Load the project
project = Project()

# Get the specific prompt you want to document
prompt = project.get_prompt("email-agent")

# Document Method will analyze the prompt and save the changes to its file.
prompt.document()

# You can also document multiple prompts programmatically
prompts_to_doc = [
    project.get_prompt("sql-generator"),
    project.get_prompt("email-summarizer")
]
project.document_all(prompts_to_document=prompts_to_doc)

Prompt Enhancer

Acts as an AI expert, providing a critique and a rewritten, more robust version of your prompt to enhance clarity, specificity, and security.

See in action

Lets say you have a prompt ‘sql-injection’. We’ll use plb prompt improve to strengthen the prompt. We’ll give it the note, “Make it more robust against SQL injection.”

...
template: |
    Given the following database schema, write a SQL query that answers the user's question.
    Schema: {{schema_definition}}
    Question: {{user_question}}
...

It will reads the prompt, enhances its clarity and specificity.

...
template: |
    You are a secure database assistant. Your task is to convert a natural language question into a safe, efficient SQL query for the given schema.
  
    IMPORTANT: Always use parameterized queries or escape inputs to prevent SQL injection.
  
    Schema: {{schema_definition}}
    Question: {{user_question}}
...

Usage

CLI

You can use plb prompt improve command on a prompt. Lets say we have a prompt template named ‘sql-injection’:

# This will show the diff and ask for confirmation
plb prompt improve sql-injection  

# You can pass a note as well 
plb prompt improve sql-injection --note "Make it more robust"

Python

Python code to improve the prompts:

from prompt_lockbox import Project

project = Project()
prompt = project.get_prompt("sql-generator")

# 1. Get the AI's suggestions (doesn't change the file)
critique = prompt.get_critique(note="Make it more robust.")

# 2. Programmatically review and then apply the changes
improved_template = critique["improved_template"]
prompt.improve(improved_template) # Saves changes

Next Steps

Now, you can start the basics guide!

Basics

Learn the essentials of this toolkit.