Skip to content

CHANGELOG.md

Latest commit

 

History

History
652 lines (538 loc) · 40.8 KB

CHANGELOG.md

File metadata and controls

652 lines (538 loc) · 40.8 KB

Changelog

  • 2026-04-02: v0.2.37 - Rename deploy.py to robot.py, update GitHub username to x-hannibal (Hannibal)

    • chore(repo): rename scripts/deploy.py to scripts/robot.py and update all references
    • chore(repo): update remaining GitHub username references from x-monk to x-hannibal

  • 2026-04-01: v0.2.36 - Update GitHub username from annibale-x to x-hannibal (Hannibal)

    • chore(repo): update all GitHub URLs, manifest links, and documentation references to new username x-hannibal

  • 2026-03-24: v0.2.35 - Promote to stable, Python 3.10 support, remove MCP Registry CI workflow (Hannibal)

    • chore: promote classifier from Beta to Production/Stable in pyproject.toml
    • chore: lower minimum Python requirement from 3.11 to 3.10; add Python 3.10 classifier
    • docs: remove beta status badge from README
    • ci: remove publish-mcp-registry.yml workflow
    • chore(zed): update Windows stub binary (x86_64-pc-windows-msvc)

  • 2026-03-24: v0.2.34 - Promote to stable, Python 3.10 support, remove MCP Registry CI workflow (Hannibal)

    • chore: promote classifier from Beta to Production/Stable in pyproject.toml
    • chore: lower minimum Python requirement from 3.11 to 3.10; add Python 3.10 classifier
    • docs: remove beta status badge from README
    • ci: remove publish-mcp-registry.yml workflow
    • chore(zed): update Windows stub binary (x86_64-pc-windows-msvc)
  • 2026-03-24: v0.2.33 - Beta badge, server.json auto-bump, CI publisher fix (Hannibal)
    • docs: add beta status badge and "Beta Status" section in README with feedback CTA
    • ci: fix mcp-publisher install — dynamically resolve latest release tag via GitHub API instead of hardcoded /releases/download/latest/ (404 fix)
    • feat(deploy): add bump_server_json() — updates both version fields in server.json during cmd_bump
    • chore(deploy): integrate server.json into bump pipeline so version stays in sync across all manifests
  • 2026-03-21: v0.2.32 - Add MCP Registry publish workflow and server.json (Hannibal)
    • ci: add publish-mcp-registry.yml workflow — triggers on tag push, updates server.json version via jq, publishes via OIDC
    • chore: add server.json for MCP Registry (schema 2025-12-11, pypi package, stdio transport)
    • docs: add MCP Registry link in README Links section
  • 2026-03-21: v0.2.31 - Add MCP Registry ownership marker (Hannibal)
  • 2026-03-21: v0.2.30 - Fix IDE integration docs: Windsurf path/filename, VS Code servers key (Hannibal)
    • docs(ide): fix Windsurf config path — correct location is ~/.codeium/windsurf/mcp_config.json (not ~/Library/Application Support/Windsurf/mcp.json)
    • docs(ide): fix Windsurf config filename — mcp_config.json (not mcp.json)
    • docs(ide): remove incorrect Windsurf project-specific config section (not supported)
    • docs(ide): fix VS Code JSON key — "servers" (not "mcpServers"; that key is Claude Desktop-specific)
    • docs(ide): fix VS Code user-level config — use Command Palette "MCP: Open User Configuration" instead of hardcoded ~/.vscode/mcp.json
  • 2026-03-21: v0.2.29 - Documentation audit: fixes, consistency, completeness (Hannibal)
    • docs: expand README TOC from 8 to 14 entries — all sections now listed
    • docs: compile CHANGELOG placeholder entries for v0.2.26-0.2.28
    • docs(dev): document rebuild command, | sentinel separator, fingerprint venv marker, MCP server restart workflow
    • docs(zed): add missing Table of Contents to ZED.md
    • docs(tools): fix apply_memento_confidence_decay params — add memory_id optional; add async/await consistency note in Integration Patterns
    • docs(decay): clarify boost_memento_confidence boosts relationships not memory node; add no-op note
    • docs(integration): remove duplicate WAL Mode bullet point
    • docs(api): remove deprecated docker-compose version key; fix recall_mementos example (returns text not JSON)
    • docs(python): fix recall_mementos example (returns text not JSON)
    • docs(agent): fix recall_mementos example (returns text not JSON)
    • docker-compose.yml: remove deprecated version key
    • docs(contributing): update test count 169→206; fix API.md description
  • 2026-03-21: v0.2.28 - Windows path parsing fix + venv fingerprint + deploy rebuild command (Hannibal)
    • fix(stub): changed sentinel separator in local_wheel.txt from : to | — colon collided with Windows drive letters causing splitn(2, ':') to parse the drive letter as filename, crashing pip install
    • fix(stub): venv marker now stores full sentinel fingerprint (path|hash) instead of STUB_VERSION — wheel content change forces venv rebuild, eliminating stale-server bug
    • fix(deploy): dev-install now verifies pip exists before install, detects broken venvs, deletes them entirely so stub performs a clean rebuild on next boot
    • fix(deploy): success marker written only if direct pip install actually succeeds
    • feat(deploy): add rebuild command — fast dev cycle: build stub + wheel + invalidate Zed venv in one step, no git interaction required
    • docs(dev): document rebuild workflow, | separator, fingerprint sentinel logic in dev/README.md
  • 2026-03-21: v0.2.27 - Regression test suite: 25/25 tools verified on live MCP server (Hannibal)
    • test: executed 25-tool regression suite on live Zed MCP server
    • fix(tools): get_memento — missing "Relationships" section in output added
    • fix(tools): find_path_between_mementos — missing path array and incorrect hop count fixed
    • fix(tools): recall_mementos — duplicate results eliminated
    • fix(tools): search_mementos — match_mode=all and min_importance filters now work correctly
    • fix(tools): apply_memento_confidence_decay — output formatting corrected
    • fix(tools): get_memento_network — now returns nodes/edges topology instead of raw statistics
  • 2026-03-21: v0.2.26 - Zed extension: stale server version root cause fix (Hannibal)
    • fix(stub): resolved stale server version issue where Zed extension launched old Python server after rebuild
    • fix(stub): venv marker logic updated to detect PyPI vs local wheel reinstalls correctly
    • fix(deploy): dev-install enhanced with pip verification and broken venv detection
    • chore: updated release_workflow memory with new deploy workflow
  • 2026-03-20: v0.2.25 - Bug fixes: 8 critical tool regressions + 37 new regression tests (Hannibal)
    • fix(tools): find_memento_patterns — handler was using wrong arguments (from_memory_id); replaced with real SQL-based pattern detection
    • fix(tools): analyze_memento_graph — handler was using wrong argument (category); replaced with real graph analytics (density, degree, category coverage)
    • fix(tools): set_memento_decay_factor — decay_factor argument was silently ignored; added set_decay_factor() DB method that writes directly to relationships.decay_factor
    • fix(tools): get_recent_memento_activity — blocking subprocess git call caused MCP transport timeout; wrapped in ThreadPoolExecutor + asyncio.wait_for(timeout=3.0)
    • fix(tools): get_memento_clusters — returned raw nested stats dict; replaced with greedy shared-neighbour cluster detection
    • fix(tools): get_central_mementos — returned raw nested stats dict; replaced with degree-centrality bridge detection
    • fix(db): PRAGMA foreign_keys=ON added to SQLiteBackend.connect() — ON DELETE CASCADE was silently inactive, leaving orphan relationship rows on memory deletion
    • fix(db): FTS multi-word AND logic — _prepare_fts_query was wrapping multi-word queries in FTS5 phrase syntax; changed to AND of prefix tokens (t1* t2*). Same fix in _search_with_simple (per-word LIKE instead of whole-string LIKE)
    • test: add test_bug_regressions.py — 37 regression tests covering all 8 bugs; suite grows from 169 to 206 tests
  • 2026-03-20: v0.2.24 - README TOC PyPI fix + stub pip timeout + CI binary size fix (Hannibal)
    • README: fixed Table of Contents anchor links for PyPI compatibility using explicit HTML anchors ()
    • README: removed stale TOC entry pointing to non-existent section; simplified TOC in DECAY_SYSTEM, INTEGRATION, RELATIONSHIPS
    • Stub: added --timeout 120 to both pip install strategies to prevent false-failure on slow PyPI connections
    • Stub: removed Stdio::null() suppression during pip install so errors are visible in Zed logs
    • Stub: moved [profile.release] to workspace root Cargo.toml — CI binaries now correctly optimized (495 KB → 309 KB)
  • 2026-03-20: v0.2.23 - Stub: race condition fixes + slow-path tool discovery (Hannibal)
    • Stub: replace PID-based lockfile with flock advisory lock — kernel auto-releases on SIGKILL, eliminates zombie-PID stale lock bug
    • Stub: Windows lock via LockFileEx with LOCKFILE_FAIL_IMMEDIATELY (windows-sys 0.59)
    • Stub: inject notifications/tools/list_changed after proxy handoff so Zed discovers all 13 tools without restart
    • Stub: replay initialize to Python with synthetic id:-1 to avoid colliding with bootstrap-answered id:1
    • Stub: filter replay to initialize + notifications only — skip requests already answered by bootstrap
    • Stub: bootstrap initialize declares capabilities.tools.listChanged=true
  • 2026-03-20: v0.2.22 - MCP bootstrap proxy: fix "Context Server Stopped" on Linux (Hannibal)
    • Stub: MCP bootstrap proxy for slow-path (venv not ready on first install/upgrade)
    • Bootstrap serves a minimal JSON-RPC 2.0 MCP server on stdio while pip runs in background thread
    • Zed initialize answered within milliseconds — no more 60s timeout crash on first install
    • Exposes memento_status tool returning human-readable setup progress
    • On Unix: re-exec after setup for zero-overhead hand-off with fully inherited stdio
    • On Windows: bidirectional pipe proxy after setup completes
    • Pip install timeout raised from 60s to 120s
  • 2026-03-20: v0.2.21 - README TOC PyPI fix + stub pip timeout (Hannibal)
    • Fixed README Table of Contents anchor links for PyPI compatibility (HTML explicit anchors, removed emoji-derived slugs)
    • Removed stale TOC entry pointing to non-existent Auto-Onboarding Protocol section
    • Simplified plain-text TOC in DECAY_SYSTEM.md, INTEGRATION.md, RELATIONSHIPS.md (anchor links broken on PyPI)
    • Stub: added --timeout 120 to pip install to prevent false-failure on slow PyPI connections
    • Stub: removed Stdio::null() suppression during pip install so errors are visible in Zed logs
  • 2026-03-20: v0.2.19 - Documentation audit and cleanup (Hannibal)
    • Rewrote API.md: removed non-existent HTTP REST/FastAPI/Node.js SDK/Kubernetes content; now documents only MCP client, Docker, and CLI export/import
    • Fixed DECAY_SYSTEM.md: removed phantom bi-temporal SQL fields, non-existent YAML config keys, missing CI workflow references, and obsolete sections
    • Fixed CONTRIBUTING.md: corrected changelog format, removed obsolete manual release checklist, updated project structure, removed mcp-memento entry point reference
    • Rewrote tests/README.md: accurate test file list, correct pytest/pytest-asyncio template, removed legacy run_tests.py references
    • Removed mcp-memento CLI alias entry point (was a no-op gui-script wrapper)
    • Removed memento-gui gui-script entry point (crashed silently due to pythonw.exe suppressing stdio)
    • Added TOC to 8 documents that lacked one (TOOLS, RELATIONSHIPS, AGENT_CONFIGURATION, INTEGRATION, SCHEMA, DECAY_SYSTEM, IDE, AGENT)
    • Eliminated duplicate Claude Desktop section between IDE.md and AGENT.md
    • Eliminated 5+ redundant copies of Configuration Methods across integration guides
    • Aligned docs/dev/README.md and scripts/README.md release workflow documentation
    • Fixed Python 3.8+ → 3.10+ in ZED.md and integrations/zed/README.md
    • Fixed ~/.zed-memento path → ~/.mcp-memento in IDE.md
    • Fixed RELATIONSHIPS.md: replaced non-existent EXEMPLIFIES type with CONFIRMS
    • Removed /etc/memento/memento.yaml (never implemented) from INTEGRATION.md
    • Compacted README.md: added TOC, reduced Natural Language section, compacted Background section
  • 2026-03-19: v0.2.18 - Debug logging opt-in and robot.py CLI redesign (Hannibal)
    • Zed extension and stub logging disabled by default; both enabled via debug.enable marker file in Zed work directory
    • Stub debug_enabled() checks MEMENTO_WORK_DIR first, then falls back to stub binary directory
    • Fixed WASM log path: replaced hardcoded /tmp/memento-zed.log with cross-platform std::env::temp_dir()
    • robot.py: bump X.Y.Z always dev (non-interactive); promote replaces prod bump; publish defaults to PyPI; publish -t for TestPyPI
    • Updated integrations/zed/README.md: unified debug logging section, updated release workflow
    • Updated scripts/robot.py docstring and help to reflect new CLI
  • 2026-03-19: v0.2.17 - Isolated venv for Zed extension stub (Hannibal)
    • Zed stub now creates and manages an isolated Python venv inside the Zed extension work directory (~/.local/share/zed/extensions/work/mcp-memento/venv/)
    • mcp-memento is installed exclusively inside the venv — no longer touches system Python or user site-packages
    • Venv is validated at every startup via a version marker file; rebuilt automatically on extension update
    • WASM extension passes MEMENTO_WORK_DIR env var to stub so venv location is always inside Zed's sandbox
    • Removed pipx fallback strategy (superseded by venv isolation)
    • robot.py bump now keeps STUB_VERSION in main.rs in sync with STUB_EXT_RELEASE in lib.rs
  • 2026-03-19: v0.2.16 - Tilde expansion fix for MEMENTO_DB_PATH on all platforms (Hannibal)
    • Fixed tilde (~) not expanding to home directory on Linux/Darwin when using MEMENTO_DB_PATH env var
    • Added _PathEnvVar descriptor for automatic tilde expansion in path-based config options
    • Added fallback for HOME/USERPROFILE unavailability in sandboxed environments
    • Added test coverage for tilde expansion in config and database engine
  • 2026-03-19: v0.2.15 - Tilde expansion fix for Linux/Darwin (Hannibal)
    • Fixed tilde (~) variable expansion on Linux/Darwin systems
  • 2026-03-19: v0.2.14 - PyPI repository URL fix and Zed Marketplace submission (Hannibal)
    • Fixed incorrect repository URL in pyproject.toml (memento-mcp-server → mcp-memento)
    • Submitted mcp-memento extension to Zed Extension Marketplace (PR zed-industries/extensions#5301)
  • 2026-03-19: v0.2.13 - robot.py release workflow hardening (Hannibal)
    • Fixed bump prod after --dev on same version: tag local-only no longer causes fatal error, retag + push offered automatically
  • 2026-03-19: v0.2.12 - robot.py release workflow hardening (Hannibal)
    • Fixed "release not found": upload_stub_binaries_to_release now creates the GitHub Release object before uploading assets
    • Fixed bump resume: tag already on remote treated as mid-flight resume, tagging step skipped gracefully
    • Added git_tag_exists_remote() using git ls-remote for reliable local-vs-remote tag distinction
    • Added upload-stubs subcommand for manual recovery (create GH Release + upload local stub binaries)
    • Fixed PermissionError when copying stub into Zed work dir while Zed is running: atomic tmp→rename strategy
    • Fixed "nothing to commit" treated as fatal error on re-run: git_commit now handles empty commits gracefully
    • Fixed changelog duplication: --dev bumps no longer write to CHANGELOG.md
    • Fixed changelog deduplication: prepend_changelog skips silently if entry already exists
    • Changed changelog strategy: --dev scaffolds a placeholder entry; prod bump verifies entry is filled and blocks on placeholder text
  • 2026-03-19: v0.2.11 - robot.py robustness fixes for release workflow (Hannibal)
    • Fixed bump command crashing when prod release attempted after a --dev bump on the same version (tag exists only locally)
    • Fixed "release not found" error: upload_stub_binaries_to_release now creates the GitHub Release before uploading assets
    • Fixed bump resume: tag already on remote is now treated as a mid-flight resume, not a fatal error
    • Added git_tag_exists_remote() helper using git ls-remote for reliable remote tag detection
    • Added upload-stubs subcommand for manual recovery (create GH Release + upload local binaries)
  • 2026-03-18: v0.2.10 - Zed extension dual-channel stub workflow (Hannibal)
    • Added dev-latest pre-release channel for iterative Zed extension testing without full releases
    • Added zed-stub-dev.yml CI workflow: triggers on dev branch push with stub/Cargo.toml changes
    • Updated robot.py: bump --dev builds stub locally and uploads to dev-latest pre-release
    • Updated lib.rs: STUB_CHANNEL constant switches download URL between dev and prod channels
    • Fixed configuration defaults from "auto" to "default" placeholder for WASM sandbox compatibility
    • Removed failed %USERPROFILE% resolution attempts in WASM; replaced with static placeholders
    • Reorganized Zed documentation: user guide to docs/extensions/ZED.md, dev guide to integrations/zed/README.md
  • 2026-03-18: v0.2.9 - Zed extension native stub binary (Hannibal)
    • Added native stub binary for Zed WASM sandbox: extension downloads pre-compiled binary at runtime
    • Added zed-stub-release.yml CI workflow: cross-compiles 5 targets and uploads to GitHub Release
    • Added stub/main.rs: lightweight native binary bridging WASM sandbox to Python MCP server
    • Fixed 404 error in development mode caused by WASM sandbox isolation from source tree
    • Added ext-binaries robot.py subcommand to download CI-built stubs into stub/bin/
  • 2026-03-18: v0.2.8 - Zed MCP extension scaffold (Hannibal)
    • Added initial Zed editor extension integrating mcp-memento as an MCP server
    • Added extension.toml, lib.rs and Cargo.toml for the Zed WASM extension
    • Added robot.py support for bumping Zed extension version alongside Python package
  • 2026-03-18: v0.2.7 - Complete documentation rewrite and quality fixes (Hannibal)
    • Full audit and rewrite of all 14 documentation files (README, CONTRIBUTING, TOOLS, RULES, RELATIONSHIPS, DECAY_SYSTEM, INTEGRATION, AGENT_CONFIGURATION, IDE, PYTHON, AGENT, API, DEV, SCHEMA)
    • Fixed all broken GitHub links (memento-mcp-server → mcp-memento)
    • Corrected Python API references: all docs now use the correct MCP client pattern (mcp library) instead of non-existent direct method calls
    • Fixed tool profile tables: Core/Extended/Advanced tool lists now match the actual source code
    • Fixed deprecated YAML config keys (sqlite_path → db_path, tool_profile → profile)
    • Fixed non-existent CLI commands (memento --maintenance, memento --export without --format)
    • Added MCP tool call disambiguation notes throughout (pseudocode vs importable library)
    • Fixed PyPI build: relative links now correctly converted to absolute GitHub URLs; changelog section injected dynamically into PyPI long description
    • Fixed robot.py: git_is_clean now ignores untracked files (Windows NUL phantom entry)
  • 2026-03-17: v0.2.6 - Environment variable standardization and CLI option cleanup (Hannibal)
    • Renamed environment variables for consistency:
      • MEMENTO_SQLITE_PATH → MEMENTO_DB_PATH
      • MEMENTO_TOOL_PROFILE → MEMENTO_PROFILE
    • Updated configuration file keys:
      • sqlite_path → db_path
      • tool_profile → profile
    • Enhanced CLI options:
      • Added --db option as primary database path argument
      • Removed deprecated profile aliases: advanced, lite, standard, full
      • Kept --profile option with values: core, extended, advanced
    • Documentation updates: Updated all README, integration guides, and examples
    • Configuration files: Updated memento.yaml and ~/.mcp-memento/config.yaml
    • Backward compatibility: Maintained support for old variable names with deprecation warnings
    • Test suite validation: All 167 tests passing with updated configuration
  • 2026-03-17: v0.2.5 - Test suite stabilization and documentation enhancement (Hannibal)
    • Test warnings resolution: Eliminated RuntimeWarning for unawaited coroutines in test suite (MagicMock → proper async mock handling)
    • CLI test reliability: Wrapped main() calls in try/except to handle SystemExit gracefully during testing
    • Enhanced test coverage validation: Verified 168 tests passing with 0 warnings
    • Documentation overhaul: Added "🌱 A Gentle Introduction" section clarifying non-agentic nature and usage patterns
    • Developer guidance enhancement: Added Memory Types Guide table defining solution, problem, code_pattern, decision types
    • Workflow documentation: Added step-by-step examples for Debugging, Feature Development, and Optimization scenarios
    • Agentic mindset clarification: Explained AI needs explicit instruction for tool usage, bridging traditional developer workflows
    • User feedback integration: Addressed confusion about autonomous agent behavior and memory storage mechanisms
  • 2026-03-17: v0.2.2 - TestPyPI release validation and build process verification (Hannibal)
    • TestPyPI validation: Successfully published and tested v0.2.1 on TestPyPI
    • Build process verification: Confirmed automated README link conversion works correctly
    • Package integrity: Verified wheel and sdist packages pass twine checks
    • Installation testing: Confirmed package installs and runs correctly from TestPyPI
    • CLI functionality: Verified memento CLI commands work after installation
  • 2026-03-16: v0.2.1 - Build infrastructure fixes and PyPI README compatibility (Hannibal)
    • PyPI README compatibility: Added dynamic link conversion for PyPI releases
    • Build script optimization: Consolidated build logic into single Python script
    • License configuration fix: Corrected pyproject.toml license field from deprecated dictionary to SPDX string
    • Automated README patching: Build-time conversion of relative links to absolute GitHub URLs
    • Cross-platform consistency: Simplified build.bat and build.sh to call central Python script
    • Build reliability: Fixed configuration errors preventing successful package builds
  • 2026-03-16: v0.2.0 - Comprehensive audit and refactoring for public release (Hannibal)
    • Breaking changes: Removed deprecated Bi-temporal tracking system (replaced with confidence-based decay)
    • Breaking changes: Eliminated fake Cypher query support, keeping only native SQLite backend
    • Breaking changes: Updated tool names to follow MCP conventions (search_mementos, recall_mementos)
    • Expanded test suite: 157 → 168 tests, coverage ~46% → 51%
    • Fixed critical documentation inconsistencies (invalid "decision" type → "general")
    • Corrected agent configuration protocols to use exact MCP tool names
    • Updated database initialization with WAL mode (PRAGMA journal_mode=WAL, busy_timeout=5000) for multi-user concurrency
    • Resolved naming conflicts (removed legacy "Context Keeper", "MemoryGraph" references)
    • Fixed CLI command examples (export_mementos() → CLI memento --export)
    • Comprehensive cleanup of ghost implementations and logical contradictions
  • 2026-03-16: v0.1.20 - Codebase consistency restoration and documentation enhancement (Hannibal)
    • Legacy code cleanup: Removed mcp_context_keeper module, example.py, setup.py, and test_basic.py to restore consistent codebase
    • Version bump: Updated to v0.1.20 in init.py and pyproject.toml
    • Project memory updates: Corrected test count (157 tests) and removed references to legacy code
    • Complete README restructuring: Linear, modular structure without redundancies
    • Removed "Universal" terminology: Replaced with "Cross-Platform", "Shared", "Consistent Behavior"
    • Added natural language interaction examples: Show chat-based memento usage with "Remember that..." patterns
    • Relationship Types cross-linking: Added prominent references to RELATIONSHIPS.md documentation
    • Eliminated duplicate content: Each piece of information now in one place only
    • Improved documentation flow: Clear path from installation to advanced usage
    • Multi-language trigger phrases: English, Italian, Spanish support for natural interaction
    • All examples in English: Removed Italian code examples from README
    • Enhanced integration table: Clear overview of all supported platforms
    • Configuration hierarchy clarification: Environment variables → YAML files → CLI args → defaults
    • Tool profile explanation: Core (13), Extended (17), Advanced (25) tools
    • Architecture overview: Simplified explanation of consistent behavior across platforms
    • Documentation bug fixes: Addressed nomenclatura inconsistencies, clarified decay triggers, added parameter requirements, resolved Node.js antipattern references
  • 2026-03-16: v0.1.19 - Test suite fixes and warning resolution (Hannibal)
    • Fixed 2 failing tests in server startup and initialization flow
    • Corrected server cleanup to call disconnect() instead of close() on database backend
    • Updated all test mocks to expect disconnect.assert_called_once() instead of close.assert_called_once()
    • Added proper warning filters for AsyncMock coroutine warnings in test suite
    • Fixed async mock configuration in CLI tests to avoid RuntimeWarning for unawaited coroutines
    • Enhanced test reliability with proper mock specifications for all async methods
    • All 157 tests now passing with 0 warnings
    • Improved test suite stability and maintainability
  • 2026-03-16: v0.1.18 - Memento onboarding protocol enhancement and tool renaming (Hannibal)
    • Renamed help_memento_tools_usage tool to memento_onboarding for clearer onboarding purpose
    • Enhanced tool with comprehensive onboarding protocol including initialization, retrieval flow, and storage triggers
    • Added optimized retrieval flow guide to prevent inefficient tool usage (6+ tool calls)
    • Integrated external Memento protocol with automatic storage triggers and on-demand triggers
    • Added memory schema requirements with mandatory tags (project, tech, category) and importance scoring
    • Added tool selection decision tree for efficient retrieval (1-3 tools for simple info, max 5 for complex tasks)
    • Added practical examples showing inefficient vs optimized tool usage patterns
    • Enhanced topic parameter support: "protocol", "retrieval_flow", "distinction", "examples", "best_practices"
    • Updated all documentation references to reflect new tool name and enhanced functionality
    • Maintained backward compatibility through topic-based access to previous guidance content
    • Updated tool definitions, registry, and imports to reflect new naming convention
    • Enhanced configuration handling for improved onboarding experience
    • Fixed test suite to work with renamed tool and updated functionality
  • 2026-03-15: v0.1.17 - Complete environment variable renaming and configuration file standardization (Hannibal)
    • Renamed all remaining CONTEXT_ environment variables to MEMENTO_ prefix:
      • CONTEXT_SQLITE_PATH → MEMENTO_SQLITE_PATH
      • CONTEXT_TOOL_PROFILE → MEMENTO_TOOL_PROFILE
      • CONTEXT_LOG_LEVEL → MEMENTO_LOG_LEVEL
      • CONTEXT_ENABLE_ADVANCED_TOOLS → MEMENTO_ENABLE_ADVANCED_TOOLS
      • CONTEXT_ALLOW_CYCLES → MEMENTO_ALLOW_CYCLES
      • CONTEXT_BACKEND → MEMENTO_BACKEND
      • CONTEXT_AUTO_EXTRACT_ENTITIES → MEMENTO_AUTO_EXTRACT_ENTITIES
      • CONTEXT_SESSION_BRIEFING → MEMENTO_SESSION_BRIEFING
      • CONTEXT_BRIEFING_VERBOSITY → MEMENTO_BRIEFING_VERBOSITY
      • CONTEXT_BRIEFING_RECENCY_DAYS → MEMENTO_BRIEFING_RECENCY_DAYS
    • Standardized configuration file name from context-keeper.yaml to memento.yaml
    • Updated all documentation references to use new configuration file name
    • Fixed remaining references in integration guides and examples
    • Updated configuration loading logic to prioritize memento.yaml over legacy names
    • Enhanced backward compatibility with deprecation warnings for old variable names
    • All 159 tests passing with new environment variable naming
    • Complete project renaming from Context Keeper to Memento finalized
  • 2026-03-15: v0.1.16 - Documentation reorganization and README restructuring (Hannibal)
    • Completely reorganized README.md to focus on user needs (636 → 441 lines)
    • Created clear separation between MCP server usage and Python library usage
    • Added comprehensive documentation structure in docs/ directory:
      • integrations/PYTHON.md - Complete guide to using Memento as a Python library
      • integrations/IDE.md - Detailed IDE integration guides
      • integrations/AGENT.md - Agent and CLI integration guides
      • RULES.md - Usage rules and best practices
      • AGENT_CONFIGURATION.md - Agent configuration templates
      • CONTRIBUTING.md - Contribution guidelines for developers
    • Improved README structure to answer three key user questions:
      1. How to install? - Clear installation instructions
      2. How to run? - IDE configuration and execution modes
      3. What can I do with it? - Use cases for developers, teams, and AI assistants
    • Enhanced documentation navigation with clear links between files
    • Maintained focus on MCP server usage for IDEs (Zed, Cursor, Windsurf, VSCode) and CLI agents
    • Updated GitHub repository URLs to use placeholders for consistency
    • All documentation cross-references verified and functional
  • 2026-03-15: v0.1.15 - Confidence system reorganization and tool distribution optimization (Hannibal)
    • Reorganized confidence system tools for better accessibility across profiles
    • Moved 3 essential confidence tools to Core profile (13 tools total):
      • get_low_confidence_mementos - Identify obsolete knowledge
      • boost_memento_confidence - Reinforce valid knowledge
      • adjust_memento_confidence - Manual confidence correction
    • Moved apply_memento_confidence_decay to Extended profile (17 tools total)
    • Kept set_memento_decay_factor in Advanced profile (25 tools total)
    • Updated tool documentation with new distribution strategy
    • Enhanced TOOLS.md with comprehensive confidence system guidance
    • All confidence system tests passing (13/13)
    • Demo script fully functional with new tool distribution
    • Improved user experience: all users now have basic confidence management
    • Better search result ordering with confidence-based prioritization
  • 2026-03-15: v0.1.14 - FTS schema fix and database stability improvements (Hannibal)
    • Fixed critical 'no such column: T.title' error in SQLite FTS table
    • Updated database schema to create FTS tables without problematic content='nodes' option
    • Added refresh_fts_support() method to properly detect FTS availability after schema creation
    • Fixed get_memento tool by adding missing get_memory() method to database interface
    • Enhanced FTS table population with INSERT OR REPLACE for reliable updates
    • Created fix_fts_schema.py script to repair existing corrupted databases
    • Added comprehensive testing for new database creation with correct FTS schema
    • Improved error handling and logging for FTS operations
    • All memento tools now fully functional with proper FTS support
  • 2026-03-15: v0.1.13 - ID generation pattern implementation and database schema initialization fix (Hannibal)
    • Implemented functional ID generation pattern with automatic UUID generation and optional override
    • Fixed database schema initialization bug in server startup (missing initialize_schema() call)
    • Added import asyncio to server.py to fix missing dependency
    • Enhanced store_memento tool with automatic ID generation when no ID provided
    • Added support for custom IDs with validation and duplicate detection
    • Updated tool schema to include optional id field in store_memento
    • Improved error handling for ID validation and conflict scenarios
    • All 128 tests passing with comprehensive ID pattern validation
  • 2026-03-15: v0.1.12 - Zed MCP integration fix and character encoding issues (Hannibal)
    • Fixed Zed MCP server integration by resolving hidden character issues in configuration
    • Identified and resolved copy-paste invisible character problems in environment variables
    • Enhanced Windows compatibility with proper UTF-8 encoding for tool descriptions
    • Improved error handling for MCP protocol communication with Zed editor
    • Added comprehensive debugging tools for Zed MCP integration testing
  • 2026-03-15: v0.1.11 - Relationship system bug fixes and API consistency (Hannibal)
    • Fixed critical bug in relationship creation: added missing create_relationship method to SQLiteMemoryDatabase
    • Added get_related_memories method with BFS traversal support for max_depth parameter
    • Added update_relationship_properties method for modifying relationship attributes
    • Added search_relationships_by_context method for structured context-based filtering
    • Fixed API inconsistency between tool handlers and database interface methods
    • Enhanced get_related_memories to properly handle relationship type filtering and depth traversal
    • Improved error handling and validation for relationship operations
    • All relationship tools now functional: create_memento_relationship, get_related_mementos, etc.
  • 2026-03-14: v0.1.10 - Complete test suite reconstruction and bug fixes (Hannibal)
    • Fixed 18 failing tests in the test suite (from 18 to 0 failures)
    • Resolved event loop issues in CLI health check and export/import tests
    • Fixed database test failures with proper ID handling and SearchQuery usage
    • Corrected sys.exit assertion problems in CLI show-config tests
    • Fixed string matching issues in health check timeout tests
    • Resolved error handling class string representation assertions
    • Fixed server test issues with mock handlers and cleanup expectations
    • Added missing Config.ALLOW_RELATIONSHIP_CYCLES attribute
    • Fixed await syntax bug in SQLiteMemoryDatabase.update_memory()
    • Added missing RelationshipProperties import to database interface
  • 2026-03-14: v0.1.9 - Path resolution and Windows compatibility fixes (Hannibal)
    • Fixed SQLite database path issues on Windows systems
    • Changed default database path from .db to context.db in configuration
    • Fixed indentation error in memento.py wrapper script
    • Improved environment variable handling for configuration overrides
    • Enhanced Windows compatibility for database file operations
  • 2026-03-14: v0.1.8 - Code quality analysis and dead code removal (Hannibal)
    • Removed unused SimpleGraph and graph algorithms classes
    • Removed unused MemoryNode and graph-related models
    • Removed unused utilities like datetime_utils.py and error_handling.py
    • Migrated inline tools definitions from server.py to tools/definitions.py
    • Fixed test suite by using explicit asserts instead of return values
    • Cleaned up leftover testing artifacts and databases
  • 2026-03-14: v0.1.7 - File reorganization and structure cleanup (Hannibal)
    • Renamed database files for better clarity:
      • sqlite_backend.py → engine.py (database engine/connection management)
      • sqlite_database.py → interface.py (domain interface for memories and relationships)
    • Renamed backends directory to database for accurate representation
    • Updated all imports to reflect new file structure
    • Improved code organization and separation of concerns
    • Removed unused work files (simple_sqlite_db.py)
    • Cleaned up import paths after file reorganization
    • Ensured all tests pass with new file structure
  • 2026-03-14: v0.1.6 - Memory tools guidance and usage clarification (Hannibal)
    • Added help_memento_tools_usage tool for comprehensive guidance on memento vs session memory usage
    • Added clear distinction guidance to prevent confusion with Serena Context Server tools
    • Added decision matrices and practical examples for tool selection
    • Renamed guide tool from get_memento_guide to help_memento_tools_usage for better LLM discoverability
    • Added best practices documentation for memento usage
    • Updated tool count: Core profile now has 10 tools (including guide tool)
    • Enhanced documentation with memento vs session memory distinction
    • Improved README.md with usage guidance and common mistakes to avoid
    • Fixed potential confusion between memento tools (_persistent suffix) and session memory tools (no suffix)
    • Fixed documentation inconsistencies regarding tool naming conventions
    • Fixed integration guidance for using multiple MCP context servers together
  • 2026-03-14: v0.1.5 - Memento tool naming convention and configuration updates (Hannibal)
    • Added memento tool naming convention to avoid conflicts with Serena Context Server
    • Added clear distinction between memento (cross-session) and session memory storage
    • Updated environment variable prefixes from MEMORY_ to MEMENTO_
    • Changed default database path to ~/.mcp-memento/context.db
    • Updated configuration file names to context-keeper.yaml
    • Fixed module entry point to properly handle CLI arguments
    • Improved Zed editor MCP configuration compatibility
    • BREAKING CHANGE: Renamed all MCP tools with _persistent suffix:
      • Core tools: store_memory → store_memento, get_memory → get_memento, etc.
      • Extended tools: get_memory_statistics → get_memento_statistics, etc.
      • Advanced tools: find_memory_path → find_path_between_mementos, etc.
    • Fixed module execution with python -m memento now works correctly
    • Fixed CLI arguments like --health and --show-config are properly handled
    • Fixed environment variable support for Zed editor integration
    • Fixed configuration file path references
    • Fixed tool registry mapping with new memento tool names
  • 2026-03-13: v0.1.4 - NetworkX dependency removal and SimpleGraph implementation (Hannibal)
    • Added SimpleGraph class to replace NetworkX dependency
    • Added lightweight in-memory graph structure for SQLite backend
    • Added JSON serialization support for SimpleGraph
    • Added integration tests for SimpleGraph with SQLite backend
    • Replaced NetworkX dependency with custom SimpleGraph implementation
    • Updated SQLite backend to use SimpleGraph instead of NetworkX
    • Reduced external dependencies from 6 to 5 packages
    • Updated project documentation to reflect NetworkX removal
    • Fixed JSON serialization issues with edge tuple keys
    • Fixed graph loading when tables don't exist yet
    • Fixed documentation inconsistencies regarding dependencies
    • Removed NetworkX dependency (networkx>=3.0.0)
    • Removed external graph library dependency
  • 2026-03-13: v0.1.3 - SQLite-only backend simplification (Hannibal)
    • Added SQLite-only backend support (removed all other backends)
    • Added simplified configuration with YAML + env vars
    • Added MemoryNode.to_database_properties() method for SQLite compatibility
    • Added updated documentation reflecting SQLite-only architecture
    • Simplified project structure by removing non-SQLite backends
    • Updated README.md to reflect SQLite-only support
    • Removed references to Neo4j, FalkorDB, Memgraph, Turso, and Cloud backends
    • Updated memory_parser.py documentation
    • Fixed critical bug preventing memory storage in SQLite backend
    • Fixed documentation inconsistencies
    • Fixed import/export functionality for SQLite
    • Removed multi-tenant support
    • Removed all non-SQLite backend implementations
    • Removed admin-only tools and migration tools
    • Removed advanced analytics modules not essential for core functionality
  • 2026-03-12: v0.1.2 - Initial simplified version for Zed editor (Hannibal)
    • Added initial simplified version for Zed editor integration
    • Added core MCP tools for memory management
    • Added SQLite backend with zero dependencies
    • Added basic configuration system
    • Project renamed to mcp-memento
    • Focus on context-aware memory management for MCP servers
    • Simplified architecture for Zed editor compatibility
  • 2026-03-11: v0.1.1 - Export/import and health check functionality (Hannibal)
    • Added export/import functionality for memories
    • Added health check CLI command
    • Added basic relationship management
    • Fixed SQLite schema initialization issues
    • Fixed memory parsing utilities
  • 2026-03-10: v0.1.0 - Initial SQLite backend implementation (Hannibal)
    • Added initial SQLite backend implementation
    • Added core memory models (Memory, MemoryContext, Relationship)
    • Added basic tool handlers for MCP protocol
    • Ported from original MemoryGraph project
    • Simplified for single-backend operation
  • 2026-03-09: v0.0.1 - Initial project setup (Hannibal)
    • Added initial project setup
    • Added MCP server foundation
    • Added basic project structure

Versioning Scheme

  • Major version (0.x.y): Breaking changes to API or architecture
  • Minor version (0.x.y): New features and enhancements
  • Patch version (0.x.y.z): Bug fixes and minor improvements