Use Ploofy with Mantella

Mantella can send NPC dialogue requests to an OpenAI-compatible language-model service. Ploofy supplies that language-model endpoint; your Mantella installation still handles the game connection, voices, and speech input.

Before you connect

Create your Ploofy account, confirm your email or sign in with a supported provider, and complete the account’s required verification and policy steps. Subscribe from your hub when a spot is available. The subscription is $19.99 USD per month, billed monthly, with five-hour and weekly usage limits. Your hub shows the remaining allowance and reset times.

In your hub, create and name an API key for this client. Copy it when it is first displayed; it cannot be revealed again. Keep it in the client’s private settings. Ploofy does not email your API key. If it is lost or shared, revoke it and create a replacement.

Set the language-model service

1. Install and confirm that Mantella and its required Skyrim components work using the official Mantella installation guide.
2. In Mantella’s UI, open the Large Language Model tab. Enter https://ploofy.ai/llm/v1 directly in the LLM Service textbox rather than choosing the OpenAI or OpenRouter preset.
3. Enter an allowed Ploofy model ID, for example swift, in the Model textbox. Use a lowercase public ID from the list below.
4. Place your private Ploofy key in Mantella’s GPT_SECRET_KEY.txt file. Use the file location for your installed Mantella version as described in its official guide; do not create or upload a publicly accessible copy.
5. Save the settings, restart Mantella if your version requires it to reload the key file, and try one short NPC exchange. Check Mantella’s local status or error message before increasing conversation length.

Use the public model IDs

The current Ploofy lineup is sage, novelist, bard, and swift. Enter the lowercase ID, not a model’s underlying download name. Your account and the authenticated model list show which models your key can currently access. Availability can change while a model is offline.

Separate dialogue from voice setup

A language-model reply does not configure text-to-speech or speech recognition. If text dialogue works but there is no spoken reply, check your selected voice service and Mantella’s audio setup separately. Ploofy’s subscription is for the roleplay models; do not paste the Ploofy key into another company’s voice-service settings.

Long conversations and memory summaries also use the language model and count toward the account’s allowance. Begin with a small conversation and avoid running multiple test conversations at once. For a base-URL field, keep the URL ending at /v1; Mantella’s compatible client adds the chat path.

When a connection fails

• 401 or 403: check that you used a Ploofy API key, that it has not been revoked, and that your subscription and model access are active. An OpenRouter or OpenAI key is not interchangeable with a Ploofy key.
• 404: check the endpoint format. A base URL ends in /llm/v1; a complete chat endpoint ends in /llm/v1/chat/completions. Do not append either suffix twice.
• 400 or 422: check the model ID, request format, and conversation length. Reduce saved history and output length if the context is too large.
• 429: check your five-hour and weekly usage in the hub. Respect any retry delay and reduce simultaneous requests. Repeated retries do not increase your allowance.
• Timeout or 5xx: check the model status in your account, wait, and retry once rather than repeatedly submitting the same request.

For help, email support@ploofy.ai with your client version, model ID, and error code. Never include your API key, payment details, or private conversations. These instructions describe verified configuration interfaces; a successful connection on your installed client still needs a short in-client test.