⚠ DESIGN PREVIEW · NOT YET BUILT · NO PUBLIC RELEASE
// READ(ccrecall.memories + knowledge_map) → REASON → ADR.md
An independent npm package that reads ccRecall's shared schema to derive Architecture Decision Records. Turns the decision context you've discussed with Claude into committable ADR markdown. Design is locked in; implementation is still pending.
// No release date — design first; implementation kicks off after Phase 02 ccRewind converges.
ccReason's core action is reading ccRecall's memories and knowledge_map tables to derive ADRs — its read/write path depends on ccRecall's SQLite schema. Three architectural options sat on the table:
Making it a third peer would mean reinventing the storage layer — its own SQLite schema, its own indexer, its own lifecycle. Wasteful, plus there's a double-write risk (the memory ADR reasoning needs may not have synced yet).
// duplicate storage layer
Making it ccrecall reason adr would pollute ccRecall's codebase — ADR reasoning is a separate concern; folding it into the memory layer makes both more complicated. Lifecycles would also tightly couple.
// codebase pollution
An independent @tznthou/ccreason npm package that reads ccRecall's SQLite schema. Independent lifecycle, independent codebase, shared data layer. Best of both worlds.
// independent npm + shared schema
// Still a CLI, still ADR markdown, still Apache-friendly. Complements the V2 ADR CLI direction; doesn't overturn it.
The commands and output below are a design preview, not yet runnable. The actual interface may shift during implementation.
$ ccreason adr "why did we switch to NATS" # Reading ccRecall: 12 relevant memories # knowledge_map → topic="messaging" → 8 related entries → Synthesizing... ✓ ADR-042 written: docs/adr/0042-switch-to-nats.md Status: Accepted Context: 5 weeks of latency issues with RabbitMQ persistent queues... Decision: Migrate to NATS JetStream for at-least-once delivery... Consequences: +3 services impacted, -40% p99 latency observed... # Sources: 12 memories from ccRecall (2026-03-15 to 2026-04-22) # Confidence: 0.78 (high · 8/12 sources direct, 4/12 inferred)
// ccReason doesn't make decisions for you — it only synthesizes what you and Claude have already discussed. Below-threshold confidence aborts the run and lists the missing context.