Dynamic Prompts

Common errors

• ModuleNotFoundError: No module named 'dynamicprompts.samplers'

  cause This error typically occurs when attempting to import or use an optional feature (like specific samplers, Magic Prompt, or Attention Grabber) that was not included during the initial `pip install dynamicprompts` command.
  fix Install `dynamicprompts` with its optional dependencies using `pip install "dynamicprompts[magicprompt, attentiongrabber]"` or specifically for the feature you need (e.g., `pip install "dynamicprompts[magicprompt]"`).

• Prompting eventually crashes / Prompts stop working after some iterations / Error while generating prompt

  cause Often caused by malformed or syntactically incorrect prompt templates. For example, a missing closing brace `}` or a typo in a wildcard name can lead to the parser failing after some initial successful attempts, especially if the problematic section is not hit immediately.
  fix Inspect your prompt templates for any syntax errors, such as mismatched braces (`{}`), incorrect wildcard delimiters (`__wildcard__`), or invalid variant syntax (`{opt1|opt2}`). Test with a very simple template first to isolate the issue. Check GitHub issues for similar parsing problems.

Warnings

• gotcha The `num_prompts` argument in generator methods (e.g., `generate`) acts as an upper bound, not an exact count. If the number of unique combinations possible in your template is less than `num_prompts`, the generator will yield fewer prompts than requested. This is particularly relevant for combinatorial generation where unique combinations can be limited.
  Fix: Be aware that the actual number of generated prompts might be less than `num_prompts` if the template's possible variations are exhausted. For precise control over combinatorial generation, consider using `CombinatorialPromptGenerator` and inspecting its `get_template_variants()` method or explicitly setting `max_prompts`.
• gotcha Using optional features like 'Magic Prompt' can incur a significant initial overhead. The first time a Magic Prompt model is used, it will download a large (approx. 500MB) model file, which can consume time and require substantial VRAM, potentially leading to CUDA errors on systems with limited GPU memory.
  Fix: Ensure you have sufficient disk space and a stable internet connection for the initial download. If encountering CUDA errors, verify your system meets the VRAM requirements or consider running on a CPU if performance is not critical.
• gotcha Subtle syntax errors in prompt templates (e.g., using `]` instead of `}` for variants, or incorrect wildcard formatting) can lead to parsing failures or unexpected prompt generation behavior. These errors might not be immediately obvious and can manifest as seemingly random crashes or incorrect output after some successful generations.
  Fix: Carefully review template syntax. The library expects specific delimiters for variants (`{item1|item2}`) and wildcards (`__wildcard__`). Test your templates thoroughly with smaller `num_prompts` counts to catch issues early. Refer to the official syntax guide for details.

Install

• pip install dynamicprompts Basic Installation
• pip install "dynamicprompts[magicprompt, attentiongrabber]" With Optional Features (Magic Prompt, Attention Grabber)

Imports

• RandomPromptGenerator

  from dynamicprompts.generators import RandomPromptGenerator

• WildcardManager

  from dynamicprompts.wildcards import WildcardManager

Quickstart

This quickstart demonstrates how to set up `RandomPromptGenerator` and `WildcardManager` to create dynamic prompts. It uses both inline variants (`{option1|option2}`) and a simple wildcard file (`__colours__`) to show how diverse prompts can be generated from a single template. The `WILDCARD_DIR` is created locally, and a sample wildcard file is generated for immediate execution.

from pathlib import Path
from dynamicprompts.wildcards import WildcardManager
from dynamicprompts.generators import RandomPromptGenerator

# Define a directory for wildcards (optional, create if needed)
WILDCARD_DIR = Path("./wildcards")
WILDCARD_DIR.mkdir(exist_ok=True)

# Create a dummy wildcard file for demonstration
(WILDCARD_DIR / "colours.txt").write_text("red\ngreen\nblue")

# Initialize WildcardManager and RandomPromptGenerator
wm = WildcardManager(WILDCARD_DIR)
generator = RandomPromptGenerator(wildcard_manager=wm)

# Generate 5 random prompts using a template with a variant and a wildcard
template = "A {flower|tree} in __colours__ hue."
num_prompts = 5

print(f"Generating {num_prompts} prompts from template: '{template}'")
for i, prompt in enumerate(generator.generate(template, num_prompts=num_prompts)):
    print(f"Prompt {i+1}: {prompt}")