v1.2.0.0 — Causal Memory & Command Consolidation
This release major release implementing SpecKit Reimagined with 33 features, 107 tasks, and 5 new memory commands. Adds session deduplication, causal memory graph, intent-aware retrieval, BM25 (exact word matching) hybrid search, and comprehensive crash recovery. It explains the main changes in simple language so the release is easier to understand.
What Changed
Session Management (W-S)
This release updates Session Management (W-S). The release updates session management (w-s). This matters because the release is clearer and easier to work with.
Session deduplication
This release updates Session deduplication. Hash-based tracking prevents re-surfacing same memories (50% token savings). This matters because the release is clearer and easier to work with.
Crash recovery
There was a reliability problem around Crash recovery. Session_state table with automatic CONTINUE SESSION generation. This matters because it reduces confusing failures and makes the system more dependable.
SessionStateManager API
This release updates SessionStateManager API. ResetInterruptedSessions(), recoverState() with _recovered flag. This matters because the release is clearer and easier to work with.
Search and Retrieval (W-R)
This release updates Search and Retrieval (W-R). The release updates search and retrieval (w-r). This matters because the release is clearer and easier to work with.
BM25 (exact word matching) hybrid search
There was a reliability problem around BM25 (exact word matching) (exact word matching) hybrid search. Pure JavaScript BM25 (exact word matching) implementation for lexical search fallback. This matters because it reduces confusing failures and makes the system more dependable.
RRF fusion enhancement
This release updates RRF fusion enhancement. K=60 parameter with 10% convergence bonus for multi-source results. This matters because the release is clearer and easier to work with.
Intent-aware retrieval
There was a reliability problem around Intent-aware retrieval. 5 intent types (add_feature, fix_bug, refactor, security_audit, understand). This matters because it reduces confusing failures and makes the system more dependable.
Cross-encoder reranking
This release updates Cross-encoder reranking. Voyage/Cohere/local providers with length penalty (100 char threshold). This matters because the release is clearer and easier to work with.
Fuzzy matching
This release updates Fuzzy matching. Levenshtein distance (<=2) + ACRONYM_MAP (30+ entries). This matters because the release is clearer and easier to work with.
Decay and Scoring (W-D)
This release updates Decay and Scoring (W-D). The release updates decay and scoring (w-d). This matters because the release is clearer and easier to work with.
Type-specific half-lives
This release updates Type-specific half-lives. 9 memory types with distinct decay rates. This matters because the release is clearer and easier to work with.
Multi-factor decay
This release updates Multi-factor decay. Pattern alignment and citation recency factors. This matters because the release is clearer and easier to work with.
Usage boost scoring
This release updates Usage boost scoring. Enhanced 5-factor composite scoring. This matters because the release is clearer and easier to work with.
Consolidation pipeline
This release updates Consolidation pipeline. 5-phase duplicate detection and semantic merge. This matters because the release is clearer and easier to work with.
Causal Memory Graph (W-G)
This release updates Causal Memory Graph (W-G). The release updates causal memory graph (w-g). This matters because the release is clearer and easier to work with.
Causal edges
This release updates Causal edges. 6 relationship types (caused, enabled, supersedes, contradicts, derived_from, supports). This matters because the release is clearer and easier to work with.
Decision lineage
This release updates Decision run history. Depth-limited traversal (max 10 hops) with cycle detection. This matters because the release is clearer and easier to work with.
Learning from corrections
This release updates Learning from corrections. Stability penalties (0.5x) and boosts (1.2x) for replacements. This matters because the release is clearer and easier to work with.
Infrastructure (W-I)
The project needed a cleaner setup for Infrastructure (W-I). The release updates infrastructure (w-i). This matters because setup and maintenance are easier to manage.
Recovery hints catalog
There was a reliability problem around Recovery hints catalog. 49 error codes with actionable recovery guidance. This matters because it reduces confusing failures and makes the system more dependable.
Tool output caching
This release updates Tool output caching. 60s TTL with LRU eviction. This matters because the release is clearer and easier to work with.
Lazy model loading
This release updates Lazy model loading. Deferred embedding initialization. This matters because the release is clearer and easier to work with.
Standardized response envelope
This release updates Standardized response envelope. {summary, data, hints, meta} structure. This matters because the release is clearer and easier to work with.
Layered tool organization
This release updates Layered tool organization. 7-layer MCP architecture (L1-L7). This matters because the release is clearer and easier to work with.
Incremental indexing
This release updates Incremental indexing. Content hash + mtime-based diff updates. This matters because the release is clearer and easier to work with.
Pre-flight quality gates
The team needed stronger proof that Pre-flight quality gates worked as expected. ANCHOR validation, duplicate detection, token budget checks. This matters because the team has clearer proof that the change works.
Protocol abstractions
This release updates Protocol abstractions. Provider-agnostic embedding interfaces. This matters because the release is clearer and easier to work with.
API key validation
The team needed stronger proof that API key validation worked as expected. Secure key handling with actionable guidance. This matters because the team has clearer proof that the change works.
Fallback chain
There was a reliability problem around Fallback chain. Primary API to Local to BM25 (exact word matching)-only with logging. This matters because it reduces confusing failures and makes the system more dependable.
Deferred indexing
This release updates Deferred indexing. Graceful degradation when embeddings fail. This matters because the release is clearer and easier to work with.
Retry logic
This release updates Retry logic. Exponential backoff (1s, 2s, 4s) with jitter. This matters because the release is clearer and easier to work with.
Transaction atomicity
This release updates Transaction atomicity. SAVEPOINT/ROLLBACK patterns. This matters because the release is clearer and easier to work with.
Files
This release updates Files. The release updates files. This matters because the release is clearer and easier to work with.
Session manager - Session deduplication and
There was a reliability problem around Session manager - Session deduplication and. Session manager - Session deduplication and crash recovery (~1050 lines). This matters because it reduces confusing failures and makes the system more dependable.
Recovery hints - 49 error codes
There was a reliability problem around Recovery hints - 49 error codes. Recovery hints - 49 error codes with hints (~520 lines). This matters because it reduces confusing failures and makes the system more dependable.
Bm25 index - Pure JS BM25 (exact word matching)
This release updates Bm25 index - Pure JS BM25 (exact word matching) (exact word matching). Bm25 index - Pure JS BM25 (exact word matching) implementation (~380 lines). This matters because the release is clearer and easier to work with.
Cross encoder - Reranking with Voyage/Cohere/local
This release updates Cross encoder - Reranking with Voyage/Cohere/local. Cross encoder - Reranking with Voyage/Cohere/local (~490 lines). This matters because the release is clearer and easier to work with.
Intent classifier - Intent detection for
This release updates Intent classifier - Intent detection for. Intent classifier - Intent detection for 5 types. This matters because the release is clearer and easier to work with.
Fuzzy match - Levenshtein + acronym
This release updates Fuzzy match - Levenshtein + acronym. Fuzzy match - Levenshtein + acronym expansion [NEVER IMPLEMENTED]. This matters because the release is clearer and easier to work with.
Corrections - Memory corrections tracking (~560
This release updates Corrections - Memory corrections tracking (~560. Corrections - Memory corrections tracking (~560 lines) [NEVER IMPLEMENTED]. This matters because the release is clearer and easier to work with.
Archival manager - Automatic archival background
This release updates Archival manager - Automatic archival background. Archival manager - Automatic archival background job (~450 lines). This matters because the release is clearer and easier to work with.
Consolidation - 5-consolidation pipeline [NEVER IMPLEMENTED]
This release updates Consolidation - 5-consolidation pipeline [NEVER IMPLEMENTED]. Consolidation - 5-phase consolidation pipeline [NEVER IMPLEMENTED]. This matters because the release is clearer and easier to work with.
Layer definitions - 7-layer MCP architecture
This release updates Layer definitions - 7-layer MCP architecture. Layer definitions - 7-layer MCP architecture (~320 lines). This matters because the release is clearer and easier to work with.
Provider chain - Embedding fallback chain
There was a reliability problem around Provider chain - Embedding fallback chain. Provider chain - Embedding fallback chain (~340 lines) [NEVER IMPLEMENTED]. This matters because it reduces confusing failures and makes the system more dependable.
Envelope - Standardized response structure (~280 lines)
This release updates Envelope - Standardized response structure (~280 lines). The release updates envelope - standardized response structure (~280 lines). This matters because the release is clearer and easier to work with.
Causal edges - Causal edge management
This release updates Causal edges - Causal edge management. Causal edges - Causal edge management (~529 lines). This matters because the release is clearer and easier to work with.
Memory context - L1 Orchestration unified
This release updates Memory context - L1 Orchestration unified. Memory context - L1 Orchestration unified entry (~420 lines). This matters because the release is clearer and easier to work with.
Causal graph - MCP handlers for
This release updates Causal graph - MCP handlers for. Causal graph - MCP handlers for causal graph (~319 lines). This matters because the release is clearer and easier to work with.
Commands
This release updates Commands. The release updates commands. This matters because the release is clearer and easier to work with.
/memory
There was a reliability problem around /memory. /memory:continue - Session recovery from crash/compaction. This matters because it reduces confusing failures and makes the system more dependable.
/memory
This release updates /memory. /memory:context - Unified entry with intent awareness. This matters because the release is clearer and easier to work with.
/memory
This release updates /memory. /memory:why - Decision run history tracing via causal graph. This matters because the release is clearer and easier to work with.
/memory
This release updates /memory. /memory:correct - Learning from mistakes with stability penalties. This matters because the release is clearer and easier to work with.
/memory
This release updates /memory. /memory:learn - Explicit learning capture (patterns, pitfalls, insights). This matters because the release is clearer and easier to work with.
MCP Tools
This release updates MCP Tools. The release updates mcp tools. This matters because the release is clearer and easier to work with.
Memory_drift_why - Trace causal chain for
This release updates Memory_drift_why - Trace causal chain for. Memory_drift_why - Trace causal chain for "why" queries. This matters because the release is clearer and easier to work with.
Memory_causal_link - Create causal relationships between memories
This release updates Memory_causal_link - Create causal relationships between memories. The release updates memory_causal_link - create causal relationships between memories. This matters because the release is clearer and easier to work with.
Memory_causal_stats - Graph statistics and coverage metrics
This release updates Memory_causal_stats - Graph statistics and coverage metrics. The release updates memory_causal_stats - graph statistics and coverage metrics. This matters because the release is clearer and easier to work with.
Memory_causal_unlink - Remove causal relationships
Memory_causal_unlink - Remove causal relationships was no longer needed in its old form. The release removes or retires memory_causal_unlink - remove causal relationships. This matters because the project becomes smaller and easier to maintain.
Memory_drift_context - Unified context with intent
Memory_drift_context - Unified context with intent was no longer needed in its old form. Memory_drift_context - Unified context with intent awareness [NEVER IMPLEMENTED -- removed from roadmap]. This matters because the project becomes smaller and easier to maintain.
Memory_drift_learn - Capture explicit learning [NEVER IMPLEMENTED
Memory_drift_learn - Capture explicit learning [NEVER IMPLEMENTED was no longer needed in its old form. Memory_drift_learn - Capture explicit learning [NEVER IMPLEMENTED -- removed from roadmap]. This matters because the project becomes smaller and easier to maintain.
Database Schema
The project needed a cleaner setup for Database Schema. The release updates database schema. This matters because setup and maintenance are easier to manage.
V8
This release updates V8. Causal_edges table (6 relation types, strength, evidence). This matters because the release is clearer and easier to work with.
V9
This release updates V9. Memory_corrections table (stability tracking, pattern extraction). This matters because the release is clearer and easier to work with.
V10
There was a reliability problem around V10. Session_state table (crash recovery, dirty flag). This matters because it reduces confusing failures and makes the system more dependable.
V11
This release updates V11. Added is_archived, archived_at columns. This matters because the release is clearer and easier to work with.
Templates
People needed clearer guidance about Templates. The release updates templates. This matters because the release notes and instructions are easier to read and trust.
Context template v2.2
People needed clearer guidance about Context template v2.2. CONTINUE_SESSION section, session_dedup metadata, causal_links. This matters because the release notes and instructions are easier to read and trust.
Save
The team needed stronger proof that Save worked as expected. Phase 0 pre-flight validation, section 16 session deduplication, deferred indexing. This matters because the team has clearer proof that the change works.
Search
This release updates Search. Hybrid search architecture, intent-aware retrieval, compression tiers. This matters because the release is clearer and easier to work with.
Memory Commands
This release updates Memory Commands. The release updates memory commands. This matters because the release is clearer and easier to work with.
All 9 memory commands
This release updates All 9 memory commands. All 9 memory commands updated with spec 082 cross-references. This matters because the release is clearer and easier to work with.
All 7 spec_kit commands
This release updates All 7 spec_kit commands. All 7 spec_kit commands updated with Five Checks Framework. This matters because the release is clearer and easier to work with.
12 YAML workflow assets
This release updates 12 YAML workflow assets. 12 YAML workflow assets updated with context loading integration. This matters because the release is clearer and easier to work with.
Folder detector
There was a reliability problem around Folder detector. CLI arguments now take priority over alignment validation (never override explicit user intent). This matters because it reduces confusing failures and makes the system more dependable.
Files Changed
| File | What Changed |
|---|---|
Spec Folder: specs/003-memory-and-spec-kit/082-speckit-reimagined/ |
Updated in this release. |
Implementation Tasks: 107 tasks across 5 workstreams (W-S, W-R, W-D, W-G, W-I) |
Updated in this release. |
Test Coverage: cross-encoder (22), intent-classifier (25), provider-chain (28), causal-edges (29), archival-manager (32), fuzzy-match (40), five-factor-scoring (65), tier-classifier (78); overall 83.2% (89/107 tasks) |
Updated in this release. |