Promptlandia

Stop guessing and start engineering your prompts! Promptlandia is a powerful web application that helps you analyze, refine, and improve your prompts using AI-powered feedback. Whether you’re a seasoned prompt engineer or just getting started, this tool will help you get better results from your generative AI models.

Key Features

• Prompt Improvement: Submit an existing prompt and get an improved version generated by a powerful AI model.
• Prompt Health Checklist: Evaluate your text prompts against a comprehensive checklist of best practices and receive detailed feedback on potential issues.
• Video Prompt Health Checklist: Evaluate your video prompts against a checklist of best practices for video generation.
• Prompt Trimmer: Reduces prompt size by removing generic LLM best practices while preserving core task-specific instructions.
• Playground: Experiment with different prompts and models in a simple and intuitive interface.
• Settings: View the underlying prompts used by the application for full transparency.

How to Use

1. Improve a Prompt: Go to the ‘Promptlandia’ page. Enter your original prompt in the text boxes and describe the improvements you’d like to see. Click ‘Improve Prompt’ to get a revised, optimized version.
2. Check Prompt Health: Go to the ‘Checklist’ page. Paste your prompt and click the ‘Evaluate’ button to get a detailed analysis of its quality against proven best practices.
3. Check Video Prompt Health: Go to the ‘Video Checklist’ page. Paste your prompt and click the ‘Evaluate’ button to get a detailed analysis of its quality against best practices for video generation.
4. Trim a Prompt: Go to the ‘Trimmer’ page. Enter a verbose prompt and click ‘Trim Prompt’ to remove generic instructions while keeping the core task requirements.
5. Experiment: Use the ‘Playground’ to freely experiment with different prompts and model settings.

Running Locally

1. Create a .env file: Use the .env.dotenv file as a template to create a .env file in the root of the project. This file should contain your GCP project information.

   cp .env.dotenv .env

2. Install dependencies and run the application:

   uv sync
   source .venv/bin/activate
   mesop app.py

Configuration

The application is configured using environment variables, which can be set in a .env file in the project root.

Variable	Description	Default
PROJECT_ID	Your Google Cloud Project ID.	(Required)
LOCATION	The Google Cloud region for Vertex AI.	us-central1
MODEL_ID	The primary Gemini model ID used for generation and improvement.	gemini-3.1-pro-preview
PLANNING_MODEL_ID	The model ID used specifically for the “thinking thoughts” (planning) step of prompt improvement.	Defaults to MODEL_ID

CLI Tools

Promptlandia also provides command-line tools for its core functionality. These can be run using uv run.

For detailed usage instructions and examples for all available tools, please refer to the CLI README.

Quick Start Examples

Trim a prompt:

echo "Your verbose prompt here..." | uv run cli/trim.py

Improve a prompt:

echo "Write a story about a dog." | uv run cli/improve.py -i "Make it sci-fi."

Check prompt health:

echo "Write a poem about a cat." | uv run cli/checklist.py

Deployment

Create a service account (once)

export PROJECT_ID=$(gcloud info --format='value(config.project)')

export DESC="promptlandia"
export SA_NAME="sa-promptlandia"
export SA_ID=${SA_NAME}@${PROJECT_ID}.iam.gserviceaccount.com

# create a service account
gcloud iam service-accounts create $SA_NAME --description $DESC --display-name $SA_NAME

# assign vertex and cloud run roles
gcloud projects add-iam-policy-binding $PROJECT_ID --member="serviceAccount:${SA_ID}" --role "roles/aiplatform.user"
gcloud projects add-iam-policy-binding $PROJECT_ID --member="serviceAccount:${SA_ID}" --role "roles/run.invoker"

Deploy / redeploy

You may choose to deploy this either unauthenticated or authenticated with Cloud Run IAP.

Setup

export PROJECT_ID=$(gcloud info --format='value(config.project)')
# The SA_ID is the email address of the service account created in the previous step.
export SA_ID=sa-promptlandia@${PROJECT_ID}.iam.gserviceaccount.com

Using Cloud Run, unauthenticated

gcloud run deploy promptlandia --source . --service-account=$SA_ID --region us-central1 --set-env-vars PROJECT_ID=$(gcloud config get project),MODEL_ID=gemini-3.1-pro-preview,LOCATION=us-central1 --allow-unauthenticated

Using Cloud Run IAP

gcloud alpha run deploy promptlandia --source . --iap --service-account=$SA_ID --region us-central1 --set-env-vars PROJECT_ID=$(gcloud config get project),MODEL_ID=gemini-3.1-pro-preview,LOCATION=us-central1

Add IAP users/groups

Utilize the gcloud alpha iap web add-iam-policy-binding to manage user/group access for IAP.

# group: Google Cloud Applied AI Engineering (mock email)
gcloud alpha iap web add-iam-policy-binding --member=group:cloud-aaie@google.com --role=roles/iap.httpsResourceAccessor --resource-type=cloud-run --region=us-central1 --service=promptlandia

For Developers

If you’re interested in contributing to the project, understanding the architecture, or learning how to add new tests, please see our Developer’s Guide.

Disclaimer

This is not an officially supported Google product.