Current Work

Active

Alpha 1: Missing Hammer Investigation (stages)
- Stage 1: Core Belief Structure ✅
- Stage 2: Descriptors & Identity ✅
- Stage 3: Multiple Similar Objects
- Stage 4+: Location, NPCs, Time, Events...
Combinatorial Explosion Components
- Phase 1: Foundation ✅
- Phase 2: Promotions ✅ (archived)
- Phase 3: Belief Resolution ✅
- Phase 4: Timeline Resolution ✅ (archived)
- Phase 5a: @tracks ← Next (plan)
- Phase 5b: alts (plan) - depends on Phase 5a
- Phase 6: Session.legacy (plan) - depends on Phase 4
- Phase 7: Observation System + recall() (plan) - independent
- Phase 8: Superposition API (plan) - depends on Phase 5b
- Deferred: Branch lifecycle, @path_certainty cache, decision time

Inspect handling server reload - Wait for nonexisting states to come back after restart
Resolve Remaining Circular Dependencies - Clean up safe but non-ideal import cycles (guide)
- Critical cycles (state ↔ mind) already resolved via registry pattern
- Remaining 22 static import cycles are safe but indicate tight coupling
- Priority targets: session ↔ channel, worker ↔ narrator, traittype ↔ archetype
- Core data model cycle (db ↔ belief ↔ state ↔ mind) is fundamental, low priority
- Use dpdm to track progress: npx dpdm public/worker/worker.mjs --circular
Mind Template Syntax: Support Bases - Enable specifying belief bases in declarative mind templates
- Current limitation: mind: {tavern: ['location']} only supports labeled subjects
- Need: Way to specify bases for beliefs created during learning
- Options: Use same format as beliefs (with bases/traits) OR add @bases meta-trait
- Goal: Avoid deep nesting in template syntax while supporting full belief construction

When working on features:

Start: Create plan in docs/plans/, update CURRENT.md to point to it
During: Track progress with checkboxes in the plan file
Complete: Mark plan complete, archive it, update CURRENT.md
Changelog: Add BRIEF summary to CHANGELOG.md when feature is fully complete
- Focus on capabilities and APIs: What can the system do now? What changed for users/implementers?
- Skip refactoring details: Don't mention code reorganization, delegation patterns, or internal structure changes
- Skip implementation details: No file names, test counts, or "how it works"
- Examples:
  - ✅ "Mind constructor requires parent_mind parameter"
  - ✅ "get_prototypes() returns full prototype chain"
  - ❌ "Moved resolution logic to owning classes using delegation pattern"
  - ❌ "Removed type-specific conditionals from Traittype infrastructure"

See existing plans in docs/plans/ for format examples.