StoryWeaver

This is the Story Weaver brain and control UI, for the Whispering Fable project. It was developed as a multi-agent architecture (before modern agent architectures were established) where a multitude of agent "subsystems" were orchestratd by an "overlord" agent, each using "thought processes" to track context for each concept or "task".

Screenshots

Local Development

1. Start local Supabase by running supabase start (must have Supabase CLI installed)
2. Run yarn dev

Thinking Architecture (note: this is outdated, from the old project)

The AI is composed of several GPT-based agent "subsystems", orchestrated using "signals". Each subsystem can influence the game world, and retrieve data using "actions".

Example subsystems:

• Storyteller: Gathers information about the world and available capabilities, and generates a storyline. Adapts a storyline with new data as needed.
• Mission Builder: Takes a storyline and generates one or more mission graphs, with objectives for players and NPCs.
• Action Analyzer: Analyzes complex player actions to interpret intent and potentially activate triggers set in place by Mission Builder.
• Character Builder: Generates characters with personalities and backstories dynamically, at the request of other subsystems.
• Playwright: Generates dialogue for characters dynamically, taking into account local environment.

When an inbound signal message is received, a "Thought Process" is created for that subsystem, which is a series of signals and actions, in a Q&A format, to respond to the initial signal, terminated by exausting necessary actions.

• Subsystems can communicate to each other using actions that create inbound signals.
• Subsystems can communicate with the game server using actions that create outbound signals.
• Subsystems can receive responses from the game server (confirmations of action, or some information retrieval, for example) with actions that watch for inbound response signals (where the response_to SQL column is set to the action's ID).

graph TD
  A[Inbound Signals] -->|direction: in| B(Thought Processes)
  A1[External Source] -->|Game Server or passage of time etc| A
  A2[Other Subsystems] --> A
  B -->|generates| C[Actions]
  C -->|responds| B
  C --> D[Outbound Signals]
  D -->|direction: out| E[Game Server]
  E --> F[Response Signals]
  F -->|direction: in, response_to: action ID| C

Loading