literallyeverything22-ux/creativeos_layered_memory

Purpose and Runtime Entry Points

CreativeOS is a local-first AI writing studio for developing shared franchise/IP knowledge. A React/Vite desktop-browser UI communicates with a Python backend, which uses local SQLite persistence and a locally running Ollama model. Startup is intended through run.bat; backend application entry is backend/main.py, and frontend bootstrap is frontend/src/main.tsx.

Client Application Shell

The frontend organizes the studio around a shared application state, sidebar navigation, and feature views for projects, documents, chat, and agent settings. frontend/src/App.tsx composes the application; frontend/src/modules/core/context/AppContext.tsx is the cross-view state boundary; frontend/src/modules/core/components/Sidebar.tsx selects workspace areas.

Project and Document Workspace

Project management and authoring are distinct UI workflows: dashboard/project creation establish the active creative workspace, while document views support browsing and editing its knowledge base. Core components are frontend/src/modules/project/components/DashboardView.tsx, frontend/src/modules/project/components/NewProjectModal.tsx, and frontend/src/modules/document/components/DocumentEditor.tsx. Backend resources are separated by backend/app/modules/project/router.py and backend/app/modules/document/router.py.

Canon and Knowledge Lifecycle

The central control boundary is document status: drafts may be edited and reviewed, but agents consume canon context unless a user explicitly pins other material. The backend’s data model and persistence layer are backend/app/core/models.py and backend/app/core/database.py. Canon source templates define the initial shared knowledge domains in Master_file templets/# PROJECT BIBLE.txt, Master_file templets/# WORLD CANON.txt, and related canon files.

AI Department Orchestration

Chat requests enter the agent layer, where routing selects one of six specialist departments—World, Character, Story, Screenplay, Producer, or Consistency—and assembles appropriate project context. This orchestration boundary is implemented in backend/app/modules/agent/router.py and backend/app/modules/agent/departments.py; user interaction is in frontend/src/modules/chat/components/ChatView.tsx and frontend/src/modules/chat/components/ChatInput.tsx.

Skills, Memory, and Model Boundary

Departments are constrained by versioned Markdown skill instructions rather than opaque private agent state. The active skill catalog is under backend/app/modules/agent/skills/, including story/story-structure-craft.SKILL.md and shared/master-file-protocol.SKILL.md. Shared memory and atom-level knowledge handling are represented by backend/app/modules/agent/memory.py and backend/app/modules/agent/memory_atoms.py. backend/app/core/llm.py is the adapter boundary to Ollama, expected to serve llama3:8b locally.

Approval, Inspection, and Export-Oriented Use

Agents propose changes; the user remains responsible for approval and canonization. The UI exposes agent/entity inspection through frontend/src/modules/agent/components/AgentSettingsModal.tsx and frontend/src/modules/agent/components/EntityBrowser.tsx. System-level API behavior is isolated in backend/app/modules/system/router.py; the architecture is designed to retain local, inspectable project data and enable document/project exports rather than cloud synchronization.