Skip to content

Latest commit

 

History

History
231 lines (182 loc) · 7.52 KB

CHANGELOG.md

File metadata and controls

231 lines (182 loc) · 7.52 KB

Changelog

All notable changes to StackMemory will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[1.0.0] - 2026-02-10

Highlights

  • First stable release of StackMemory
  • Licensed under BSL 1.1 (converts to MIT after 4 years)
  • 652 tests passing, TypeScript strict mode, Node 20+ required

Added

  • FTS5 full-text search with BM25 ranking and hybrid scoring
  • sqlite-vec embedding support with multi-provider architecture (Xenova, OpenAI)
  • Memory monitor daemon with auto-capture/clear cycle
  • Prompt Forge (GEPA) with AGENT.md watcher and extensions system
  • Skills system: spec generator, Linear task runner, agent prompt consolidation
  • DiffMem integration for long-term user memory
  • Retrieval signals, garbage collection, cold storage, multi-repo support
  • Node.js 20+ runtime version check at CLI startup
  • Claude Code hooks: auto-init on wrapper launch

Changed

  • Removed SMS/WhatsApp/Twilio integration (replaced with terminal bell notification)
  • Removed ngrok dependency
  • Reduced as any casts by 71% (153 to 44) with proper TypeScript interfaces
  • Consolidated test suite from 719 to 652 focused tests
  • CI/CD: Node 20+22 matrix, fixed deprecated actions, added concurrency groups
  • Build: clean dist before build, removed source maps from tarball (2.1MB to 735KB)
  • License changed from MIT to BSL 1.1

Fixed

  • Schema unification between FrameDatabase and SQLiteAdapter
  • FTS5 empty query fallback to LIKE search
  • Pre-publish script git dirty check race condition
  • Embedding provider initPromise reset on failure
  • SQLite pragma conflict in skills CLI
  • CLI test timeout flakes via module import caching

Security

  • Feature flags cleanup (removed stale whatsapp flag)
  • Removed dead feature stubs and scripts
  • Input validation improvements across CLI commands

[0.3.15] - 2025-01-14

Added

  • Phase 4: Two-Tier Storage System (STA-414)
    • Local storage tiers: Young (<24h), Mature (1-7d), Old (7-30d)
    • Remote storage: S3 + TimeSeries DB integration
    • Compression strategies: LZ4 for mature, ZSTD for old data
    • Background migration engine with configurable triggers
    • Offline queue with persistent retry logic
    • CLI commands: storage status, migrate, cleanup, config, test

Fixed

  • Frame manager test using non-existent getEvents method
  • Linear API authentication issues (env variable conflicts)
  • SQL syntax errors for SQLite INDEX creation
  • ESM import issues (missing .js extensions)

Changed

  • Repository structure reorganization
    • Documentation moved to /docs and /docs/archives
    • Test scripts moved to /scripts directory
    • Linear cleanup files archived to /archive/linear-cleanup-2026
    • Cleaner root directory structure

Performance

  • Storage compression ratios: LZ4 ~2.5x, ZSTD ~3.5x
  • Test suite: 467 passing tests, improved from 19 failures
  • Build time: ~90ms with esbuild
  • Session cleanup: Removed 1,032 old sessions (4.1MB saved)

[0.3.4] - 2024-01-06

Added

  • Advanced Frame Handoff Workflows (STA-100)
  • Claude Skills system for enhanced workflow automation
  • Recovery Checkpoint Manager for work state preservation
  • Context Archaeologist for deep historical search
  • Frame Handoff Orchestrator for streamlined team handoffs

Fixed

  • Context commands (push/pop/show) hanging indefinitely
  • Database path resolution in test scripts
  • Context bridge initialization blocking CLI operations

Changed

  • FrameManager now accepts options object with skipContextBridge flag
  • CLI commands skip async context bridge for synchronous operation
  • Database path standardized to project-local .stackmemory/context.db

Performance

  • Context operations: ~400ms (previously hung indefinitely)
  • Task operations: 89-98% faster than manual alternatives
  • Database operations: Consistent ~500ms response time

[0.3.3] - 2024-01-05

Added

  • Dual Stack Manager for seamless context transitions
  • Context Bridge for real-time synchronization
  • Clear survival system for preserving context across resets
  • Quality gates and post-task validation hooks

[0.3.2] - 2024-01-04

Added

  • Context overflow handling with graceful management
  • Ledger system providing 10-15x context compression
  • Automatic thresholds (60% warning, 70% critical, 85% force-save)
  • Perfect restoration after /clear operations

[0.3.1] - 2024-01-03

Added

  • Linear OAuth authentication implementation
  • Bidirectional sync between Linear and StackMemory
  • Auto-sync background synchronization
  • Webhook support for real-time Linear updates

[0.3.0] - 2023-12-30

Added

  • Phase 3 Collaboration features complete
  • Shared team stacks with dual architecture
  • Frame handoff mechanism
  • Team collaboration features
  • Shared context system

[0.2.5] - 2023-12-28

Added

  • Hybrid digest generation (60% deterministic, 40% AI)
  • Smart trace detection and bundling
  • Configurable scoring with weight profiles
  • Railway storage optimization (3-tier: Redis/Buckets/GCS)

[0.2.0] - 2023-12-26

Added

  • Phase 2 Intelligence Layer complete
  • LLM-driven context retrieval
  • Summary generator with compression
  • Hierarchical retrieval system
  • Context Retriever orchestration

[0.1.5] - 2023-12-20

Added

  • Pebbles Task Store with Git-tracked JSONL
  • Task-aware context system
  • Task priorities (low, medium, high, urgent)
  • Task states (pending, in_progress, completed, blocked, cancelled)

[0.1.0] - 2023-12-15

Added

  • Initial release with Phase 1 Core Runtime
  • Frame-based call stack architecture
  • SQLite database storage
  • MCP server with 20+ tools
  • Basic Claude Code integration
  • Project initialization and management

Key Metrics

Performance Improvements (v0.3.4)

  • Context reestablishment: 95% faster (5min → 15s)
  • Task completion: 40% faster
  • Rework rate: 80% reduction (25% → 5%)
  • Context accuracy: 58% improvement (60% → 95%)

Storage Statistics

  • Database size: ~588KB for typical project
  • Task storage: ~214KB for 250 tasks
  • Compression ratio: 10-15x for context
  • Frame depth support: 10,000+ levels

Integration Status

  • MCP Tools: 20+ available
  • Linear Integration: Fully bidirectional
  • Git Integration: Repository-aware
  • Claude Code: Full auto-trigger support

Migration Guide

From v0.2.x to v0.3.x

  1. Run npm install -g @stackmemoryai/stackmemory@latest
  2. Execute stackmemory init in your project
  3. Configure Linear integration with stackmemory linear:auth
  4. Enable Claude Skills with stackmemory skills:enable

From v0.1.x to v0.2.x

  1. Backup .stackmemory/context.db
  2. Update to latest version
  3. Run database migration: stackmemory migrate
  4. Reconfigure MCP server settings

Breaking Changes

v0.3.0

  • FrameManager constructor signature changed (now accepts options object)
  • Context bridge initialization is now optional
  • CLI commands require STACKMEMORY_CLI environment variable

v0.2.0

  • Database schema updated for LLM retrieval
  • Frame digest format changed to hybrid structure
  • Scoring system now uses configurable weights

Deprecations

Deprecated in v0.3.0

  • Direct contextBridge initialization in FrameManager (use options instead)
  • Synchronous context operations without skipContextBridge flag

Deprecated in v0.2.0

  • Simple text digests (replaced with hybrid digests)
  • Fixed scoring weights (replaced with profiles)

Support

For issues, questions, or contributions: