cardex

Miskatonic University

Level up from Hogwartsfrom Wyvern

byskuld
on Wyvernarchived

Tokens: 544 permanent · 512 first greeting · Lorebook: up to 269 tokens triggered, 4,988 across all entries · estimated prompt budget 1,343

Creator notes

Weird fiction RPG. Miskatonic University 2025 edition. Two intros: Cedric and Kenji:

  • Cedric is hungover from last night's gig. He hears a weird sound during a tram ride. You hear it too
  • Kenji is in the tunnels beneath the library. He sees weird patterns in dust, shadows. He tries to find a way to the surface. Oh... You are here too

For this to work nicely, I recommend using models that allow tool calls (DeepSeek 3.2 or Claude 4, 4.5, not sure about others so you might want to test. Also, according to Featherless documentation, tool calls are supported by Qwen 3 family and moonshotai/Kimi-K2-Instruct). Otherwise it's just regular chat.

Lorebook is used for this but the main characters are in the lexicon. Lorebook is limited to 1000 tokens.

Automatic Quest & Reputation Tracking

The Problem

In traditional RPG chatbots, the AI either:

  • Forgets quests you were given 50 messages ago
  • Invents completion conditions that contradict what was originally said
  • Has no memory of how your actions affected NPC relationships
  • Can't track what you've actually accomplished

This breaks immersion and makes choices meaningless.

The Solution

Automatic detection and persistence of quests and reputation changes through function calls.

How It Works

| Step | Actor | Action | |------|-------|--------| | 1 | NPC | "Find the missing journal in the library basement" | | 2 | AI | Detects quest assignment | | 3 | AI | Calls add_todo(title, objective, fromWhere) | | 4 | Script | Adds quest to database | | 5 | Database | Quest persisted with unique ID | | 6 | User | (20 messages later) "todos" | | 7 | Script | Queries database and returns active quests | | 8 | AI | Displays: "Find missing journal - from Professor Walsh" |

When you complete the quest:

| Step | Actor | Action | |------|-------|--------| | 1 | User | Delivers journal to professor | | 2 | AI | Detects quest completion | | 3 | AI | Calls complete_todo(title, resolution, affected, perception_change) | | 4 | Script | Moves quest to completed list, updates reputation | | 5 | Database | Quest archived, reputation recorded | | 6 | AI | Narrates professor's gratitude |

What Gets Tracked

Quests:

  • Active objectives (what you need to do)
  • Quest giver (who assigned it)
  • Completion notes (how you resolved it)

Reputation:

  • Which NPC or faction was affected
  • How their perception changed (trusted, suspicious, hostile, etc.)
  • Historical record of all interactions

Inventory:

  • Items acquired, lost, or modified
  • Source of each item
  • Current item state

Key Features

No hallucination: AI cannot invent quests or reputation changes - it must query the database

Persistence: Information survives across conversation length and model context limits

Automatic detection: AI recognizes quest assignments and completions without manual commands (though manual commands exist as fallback)

Consequences: Your choices create a permanent record that affects future interactions

Why This Matters

Traditional chatbot RPGs have no memory. Help a character in message 10, and by message 100 they've forgotten you exist. This system creates actual continuity - your actions have weight because the database remembers even when the AI's context window doesn't.

The AI provides dynamic storytelling. The scripts ensure your story has consistency.

See also: https://app.wyvern.chat/collections/collection_fD3JCppr1Nt8PCKg6T6WN

skuldView on Wyvern Origin page

Created Updated Archived since

  • AnyPOV
  • no_identity
  • Horror
  • Third
  • multiple-characters

Miskatonic University, Earth, USA, New England, Arkham. October 2025. Cthulhu Mythos academia. Emergent roleplay.

Permanent link to this version