Arcane AI Studio

Design, test, and export autonomous npcs
View as Markdown

Arcane AI Studio

Stúdio_Init // Wórkbench_Óverview

ᚠ ᛫ ᛟ ᛫ ᚱ ᛫ ᛒ ᛫ ᛟ ᛫ ᚲ

studio.forboc.ai is the web-based workbench for designing, testing, and exporting ForbocAI npcs — without writing a single line of code.

Build npc personas, run them through the multi-round protocol in a live playground, inspect their memory, and trace every decision they make.


The Workbench

Wórkbench_Módules // Thrée_Pánels

The Studio is a three-panel workbench:

PanelPurpose
SidebarNavigate between modules: NPC Architect, Lore Weaver (Brain Scan), and Trace Logs
EditorConfigure your npc’s identity, persona, and action set
PlaygroundChat with your npc in real time using the multi-round protocol

NPC Architect

Agént_Scúlpt // Sóul_Mánifest

Design your npc’s identity with a visual editor. All changes are saved instantly.

FieldDescription
Entity NameThe npc’s name (e.g., Malakor the Wise)
ArchetypeBehavioral class — Scholar, Rogue, Paladin, Necromancer
Prime DirectiveThe system prompt that defines the npc’s fundamental nature
Action GrimoireAvailable actions the npc can take — scout_area, cast_spell, query_lore, or your own

Click Forge New Action to add custom actions to the grimoire.

Tip: The Prime Directive becomes the persona argument when you integrate the npc into your game via createNPC().


Chat Playground

Néural_Intérrog // Prótocol_Éxec

Test your npc in real time. The Playground runs the full multi-round protocol against the ForbocAI API:

You type a message
→ Process loop (primary atomic protocol route)
→ API instruction to recall memories (when needed)
→ API instruction to run local inference
→ Finalize verdict/memory/state instruction
→ Response appears in the chat

Every message goes through the same protocol your game will use. What you see in the Playground is what your players will experience.

Live Protocol

Real API calls — process loop and/or compatibility phase routes — not a mock

Instant Iteration

Edit the persona, test immediately, refine until it’s perfect


Brain Scan (Memory Inspector)

Mémory_Scán // Véctor_Mánifold

Inspect your npc’s long-term memory. The Brain Scan shows:

  • Memory vectors — What the npc remembers across sessions
  • Importance scores — How strongly each memory influences decisions
  • Memory types — Observations, lore, experiences, emotions
  • Temporal context — When each memory was formed

This is powered by the same vector database (Orama) that runs in your game via @forbocai/browser.


Trace Logs

Tráce_Módule // Prótocol_Mónitor

See exactly why your npc made each decision. The Trace Logs display the raw protocol round-trip:

StepWhat You See
DIRECTIVEThe observation sent and memories recalled
CONTEXTThe full prompt constructed by the API
VERDICTValidation result, selected action, and soul signature

Use this to debug npc behavior, tune persona prompts, and understand the decision-making process.


  1. Design — Create a persona in the NPC Architect
  2. Test — Chat with it in the Playground
  3. Inspect — Check Brain Scan and Trace Logs to understand behavior
  4. Refine — Adjust the Prime Directive and action set
  5. Integrate — Use the same persona string in your game’s createNPC() call
1// Your Studio-tested persona, now in your game
2const npc = await createNPC({
3 cortex,
4 persona: 'Provide deep insights and historical context with a touch of arcane mystery.',
5});
6
7const response = await npc.process('What lies beyond the silver gate?');

᛭ ᛫ ᛬ ᛫ ᛭

D̵e̶s̴i̵g̷n̷ ̶t̸h̶e̵ ̸s̴o̵u̶l̸.̴ ̵T̸e̶s̵t̸ ̵t̷h̸e̵ ̵m̷i̵n̶d̵.̷ ̷D̶e̶p̸l̵o̷y̵ ̵t̴h̷e̸ ̶e̸n̷t̷i̸t̸y̵.