Beta v1.3.0

Changelog

api simplification

• python sdk: simplified to zero-config Memory() api matching javascript
  • from openmemory.client import Memory → mem = Memory()
  • works out of the box with sensible defaults (in-memory sqlite, fast tier, synthetic embeddings)
  • optional configuration via environment variables or constructor
  • breaking change: moved from OpenMemory class to Memory class

benchmark suite rewrite

• implemented comprehensive benchmark suite in temp/benchmarks/
  • typescript-based using tsx for execution
  • supports longmemeval dataset evaluation
  • multi-backend comparison (openmemory, mem0, zep, supermemory)
• created src/main.ts consolidated benchmark runner
  • environment validation
  • backend instantiation checks
  • sequential benchmark execution with detailed logging

✨ features

core improvements

• Memory.wipe(): added database wipe functionality for testing

  • clear_all implementation in db.ts for postgres and sqlite
  • clears memories, vectors, waypoints, and users tables
  • useful for benchmark isolation and test cleanup

• environment variable overrides:

  • OM_OLLAMA_MODEL: override ollama embedding model
  • OM_OPENAI_MODEL: override openai embedding model
  • OM_VEC_DIM: configure vector dimension (critical for embedding compatibility)
  • OM_DB_PATH: sqlite database path (supports :memory:)

vector store enhancements

• added comprehensive logging to PostgresVectorStore
  • logs vector storage operations with id, sector, dimension
  • logs search operations with sector and result count
  • aids in debugging retrieval issues

🐛 bug fixes

• embedding configuration:

  • fixed models.ts to respect OM_OLLAMA_MODEL environment variable
  • resolved dimension mismatch issues (768 vs 1536) for embeddinggemma
  • ensured OM_TIER=deep uses semantic embeddings (not synthetic fallback)

• benchmark data isolation:

  • implemented proper database reset between benchmark runs
  • fixed simhash collision issues causing cross-user contamination
  • added resetUser() functionality calling Memory.wipe()

• configuration loading:

  • fixed dotenv timing issues in benchmark suite
  • ensured environment variables load before openmemory-js initialization
  • corrected dataset path resolution (longmemeval_s.json)

📚 documentation

• comprehensive readme updates:

  • root README.md: language-agnostic, showcases both python & javascript sdks
  • packages/openmemory-js/README.md: complete api reference, mcp integration, examples
  • packages/openmemory-py/README.md: zero-config usage, all embedding providers

• api documentation:

  • environment variables with descriptions
  • cognitive sectors explanation
  • performance tiers breakdown
  • embedding provider configurations

🔧 internal improvements

• type safety: added lint error handling in benchmark adapters
• code organization: separated generator, judge, and backend interfaces
• debug tooling: created dimension check script (check_dim.ts)
• logging standardization: consistent [Component] prefix pattern

⚠️ breaking changes

• python sdk now uses from openmemory.client import Memory instead of from openmemory import OpenMemory
• Memory() constructor signature changed to accept optional parameters (was required)
• benchmark suite moved to typescript (was python)

---

• Feature - Add Frontend Docker Image by @dflor003 in #100

New Contributors

• @dflor003 made their first contribution in #100

Full Changelog: v1.2.3...v1.3.0

v1.2.3

1.2.3 - 2025-12-14

Added

• Temporal Filtering: Enables precise time-based memory retrieval
  • Added startTime and endTime filters to query method across Backend, JS SDK, and Python SDK.
  • Allows filtering memories by creation time range.
  • Fully integrated into hsg_query logic.

Fixed

• JavaScript SDK Types: Fixed IngestURLResult import error and v.v property access bug in VectorStore integration.
• Python SDK Filtering: Fixed missing implementation of user_id and temporal filters in hsg_query loop.

What's Changed

• fix: Add PATCH to CORS allowed methods by @aziham in #93

New Contributors

• @aziham made their first contribution in #93

Full Changelog: v1.2.2...v1.2.3

v1.2.2

Fixed

• MCP Server Path Resolution: Fixed ENOENT error in stdio mode (Claude Desktop)

  • Enforced absolute path resolution for SQLite database
  • Ensures correct data directory creation regardless of working directory
  • Critical fix for local desktop client integration

• VectorStore Refactor: Fixed build regressions in backend

  • Migrated deprecated q vector operations to VectorStore interface
  • Fixed users.ts, memory.ts, graph.ts, mcp.ts, and decay.ts
  • Removed partial SQL updates in favor of unified vector store methods

Added

• Valkey VectorStore Enhancements: Improved compatibility and performance
  • Refined vector storage implementation for Valkey backend
  • Optimized vector retrieval and storage operations

Changed

• IDE Extension:

  • Updates to Dashboard UI (DashboardPanel.ts) and extension activation logic (extension.ts)
  • Configuration and dependency updates

• Python SDK:

  • Refinements to embedding logic (embed.py)
  • Project configuration updates in pyproject.toml

• Backend Maintenance:

  • Dockerfile updates for improved containerization
  • Updates to CLI tool (bin/opm.js)

New Contributors

• @ajitam made their first contribution in #69
• @fparrav made their first contribution in #80
• @DAESA24 made their first contribution in #83
• @oantoshchenko made their first contribution in #84
• @therexone made their first contribution in #85

Full Changelog: 1.2.1...v1.2.2

V1.2.1

1.2.1 - Standalone Sdk

Added

• Python SDK (sdk-py/): SDK Overhaul, it can now perform as a standalone version of OpenMemory

  • Full feature parity with Backend
  • Local-first architecture with SQLite backend
  • Multi-sector memory (episodic, semantic, procedural, emotional, reflective)
  • All embedding providers: synthetic, OpenAI, Gemini, Ollama, AWS
  • Advanced features: decay, compression, reflection
  • Comprehensive test suite (sdk-py/tests/test_sdk_py.py)

• JavaScript SDK Enhancements (sdk-js/): SDK Overhaul, it can now perform as a standalone version of OpenMemory

  • Full feature parity with Backend
  • Local-first architecture with SQLite backend
  • Multi-sector memory (episodic, semantic, procedural, emotional, reflective)
  • All embedding providers: synthetic, OpenAI, Gemini, Ollama, AWS
  • Advanced features: decay, compression, reflection

• Examples: Complete rewrite of both JS and Python examples

  • examples/js-sdk/basic-usage.js - CRUD operations
  • examples/js-sdk/advanced-features.js - Decay, compression, reflection
  • examples/js-sdk/brain-sectors.js - Multi-sector demonstration
  • examples/py-sdk/basic_usage.py - Python CRUD operations
  • examples/py-sdk/advanced_features.py - Advanced configuration
  • examples/py-sdk/brain_sectors.py - Sector demonstration
  • examples/py-sdk/performance_benchmark.py - Performance testing

• Tests: Comprehensive test suites for both SDKs

  • tests/js-sdk/js-sdk.test.js - Full SDK validation
  • tests/py-sdk/test-sdk.py - Python SDK validation
  • Tests cover: initialisation, CRUD, sectors, advanced features

• Architecture Documentation

  • Mermaid diagram in main README showing complete data flow
  • Covers all 5 cognitive sectors
  • Shows embedding engine, storage layer, and recall engine
  • Includes temporal knowledge graph integration
  • Node.js script to regenerate diagrams

1.2.0

What's New

• Web UI to control OpenMemory
• HYBRID Tier Performance Mode
• Memory Compression Engine

What's Changed

• Decay system
• perf(vector): Optimized the aggregateVectors function by @DKB0512 in #26
• perf(embedding): Optimize embedWithLocal by @DKB0512 in #29
• perf(chunk): Optimized the combineChunk function by @DKB0512 in #27
• Add permissions for content read access by @recabasic in #31

New Contributors

• @recabasic made their first contribution in #31

Full Changelog: 1.1.1...1.2.0

1.1.1

Changelog

Added

• Memory Compression Engine: Auto-compresses chat/memory content to reduce tokens and latency

  • 5 compression algorithms: whitespace, filler, semantic, aggressive, balanced
  • Auto-selects optimal algorithm based on content analysis
  • Batch compression support for multiple texts
  • Live savings metrics (tokens saved, latency reduction, compression ratio)
  • Real-time statistics tracking across all compressions
  • Integrated into memory storage with automatic compression
  • REST API endpoints: /api/compression/compress, /api/compression/batch, /api/compression/analyze, /api/compression/stats
  • Example usage in examples/backend/compression-examples.mjs

• VS Code Extension with AI Auto-Link

  • Auto-links OpenMemory to 6 AI tools: Cursor, Claude, Windsurf, GitHub Copilot, Codex
  • Dual mode support: Direct HTTP or MCP (Model Context Protocol)
  • Status bar UI with clickable menu for easy control
  • Toggle between HTTP/MCP mode in real-time
  • Zero-config setup - automatically detects backend and writes configs
  • Performance optimizations:
    • ESH (Event Signature Hash): Deduplicates ~70% redundant saves
    • HCR (Hybrid Context Recall): Sub-80ms queries with sector filtering
    • MVC (Micro-Vector Cache): 32-entry LRU cache saves ~60% embedding calls
  • Settings for backend URL, API key, MCP mode toggle
  • Postinstall script for automatic setup

• API Authentication & Security

  • API key authentication with timing-safe comparison
  • Rate limiting middleware (configurable, default 100 req/min)
  • Compact 75-line auth implementation
  • Environment-based configuration

• CI/CD

  • GitHub Action for automated Docker build testing
  • Ensures Docker images build successfully on every push

Changed

• Optimized all compression code for maximum efficiency

1.1.0

• Add Pluggable vector dbs and PostgreSQL support

Full Changelog: 1.0.0...1.1.0

MCP

What's Changed

• Add Model Context Protocol (MCP)
• Add Tag and Metadata Filtering to HSG + API Query by @ammesonb in #1
• refactor: Dockerfile to install all dependencies and prune dev by @josephgoksu in #4
• Fix Docker build by @pc-quiknode in #5

New Contributors

• @ammesonb made their first contribution in #1
• @josephgoksu made their first contribution in #4
• @pc-quiknode made their first contribution in #5

Full Changelog: https://github.com/CaviraOSS/OpenMemory/commits/1.0.0