How ai-team-team/ai-novel works
AI Novel Writer turns a user’s novel blueprint into chapter prose while maintaining a continuity model: initialize the story world, plan and write chapters, review and scan the results, resolve fact conflicts, then commit approved state for later retrieval. The README also documents ATT agent committees, interruption recovery, and a live dashboard. The graph reflects those documented capabilities; unsampled implementation details are represented at their documented subsystem paths, and uncertain internal wiring is omitted.
Main entry points: workflow.py Workflow manager · memory.py Memory manager · state_manager.py Context retrieval · llm_client.py LLM client
Components 21
Novel Workflow
- CLI
src/main.py - Workflow manager
src/workflow.py - Project bootstrap
src/workflow_components/project_mixin.py - Chapter planning
src/workflow_components/planning_mixin.py - Chapter writing
src/workflow_components/writing_mixin.py - Review and revision
src/workflow_components/discussion.py - Fact scanning
src/workflow_components/scanning_mixin.py
Agent Orchestration
- ATT committees
src/workflow_components/autonomy_mixin.py - Database governance
src/att/db_committee.py - Conflict resolution
src/workflow_components/conflict_resolver.py - LLM client
src/llm_client.py
Continuity Memory
- Context retrieval
src/state_manager.py - Memory manager
src/memory.py - Story database
src/memory_components/schema_mixin.py - Vector index
src/memory.py - Conflict and commits
src/memory_components/conflict_commit_mixin.py
Runtime Operations
- Recovery checks
src/workflow_components/resume_mixin.py - Live dashboard
src/utils/dashboard.py
Other components
- Novel writer
- Model providers
- Embedding service
Connections 32
- Novel writer → CLI: invokes
- CLI → Workflow manager: dispatches
- Workflow manager → Project bootstrap: initializes
- Workflow manager → Chapter planning: plans chapters
- Workflow manager → Chapter writing: generates prose
- Workflow manager → Review and revision: reviews prose
- Workflow manager → Fact scanning: scans facts
- Workflow manager → Conflict resolution: resolves conflicts
- Workflow manager → Context retrieval: retrieves context
- Workflow manager → Memory manager: commits facts
- Chapter planning → Context retrieval: requests context
- Chapter writing → Context retrieval: requests context
- Fact scanning → LLM client: extracts facts
- Chapter planning → LLM client: generates guide
- Chapter writing → LLM client: generates prose
- Review and revision → LLM client: reviews and revises
- Conflict resolution → ATT committees: convenes debate
- Database governance → ATT committees: uses ATT manager
- Memory manager → Database governance: audits writes
- Memory manager → Story database: reads and writes
- Memory manager → Vector index: indexes facts
- Context retrieval → Memory manager: queries memory
- Context retrieval → Story database: reads structured facts
- Context retrieval → Vector index: searches details
- Memory manager → Conflict and commits: records commits
- Conflict and commits → Story database: updates commit state
- LLM client → Model providers: sends prompts
- LLM client → Embedding service: requests embeddings
- Workflow manager → Recovery checks: validates on recovery
- Recovery checks → Memory manager: checks memory state
- Workflow manager → Live dashboard: updates status
- ATT committees → Live dashboard: reports activity