Skip to main content

AI Assistant

DM Loremind ships with a contextual AI assistant: on any entity (Lore page, scene, PC, NPC, arc, chapter…), a ✨ button opens a side panel where you can chat with a language model that knows the open entity.

Philosophy​

The AI does not replace the GM — it assists them. Typical use cases:

  • Filling the gaps: generate a description for a vague location or NPC
  • Narrative suggestions: "suggest 3 motivations for this NPC", "how could this fight turn around?"
  • Consistency: "is this NPC consistent with the Lore of the city of Vaeril?"
  • Speed: generate the skeleton of an arc in a few seconds, to refine afterwards

Supported providers​

Five LLM providers can be configured in the Settings → AI — one local and four cloud:

  • Ollama (local) — 100% private execution, your data never leaves your machine. Ideal with a GPU (works on CPU, but slowly).
  • 1min.ai (cloud) — aggregator (Anthropic, OpenAI, Google, Mistral…) through a single paid subscription.
  • Mistral (cloud) — Mistral AI's API (European models, good in French).
  • Gemini / Google AI Studio (cloud) — Google's Gemini models; free API key on aistudio.google.com.
  • OpenRouter (cloud) — multi-provider router, with free models available.

The choice is made globally (one active provider at a time), with a model selected by default. You can switch at any time. Details: Provider configuration.

The assistant proposes, you apply​

Beyond chat, the assistant can also write into your sheets — always on the same principle: it proposes, nothing is applied without your validation.

  • Field-by-field fleshing out — on arcs, chapters and scenes, a ✨ button next to the text fields proposes content that builds on the rest of the sheet and the campaign context. You accept, regenerate or ignore.
  • Scene drafts — on a chapter, the "Generate scenes" panel proposes complete scenes to insert into the chapter. It's also the shortcut opened by the "Fix" button of the guided prep when facing an empty chapter.
  • "Previously on…" recap — during a session, the assistant writes the recap of the previous session from your journal, logged in one click.
  • Lore page generation — field by field or in wizard mode at creation time (see Lore page generation).

Context injected automatically​

When you open the chat on an entity, the backend builds a prompt that includes:

  1. The entity itself — its significant fields (not images or extremely large notes)
  2. The linked Lore — related pages, for universe consistency
  3. The Game System — relevant rule sections (Combat, Classes…) depending on the detected intent
  4. The conversation history — the previous messages in the drawer

You don't need to manually re-paste this context — it is computed for you.

Detailed pages​