Changelog

All notable changes to MemOS will be documented here
GitHub
v2.0.23

MemOS Cloud

Improvements
Preference memory model invocation:
  • Made preference-memory model calls compatible with Qwen3 hybrid-thinking models and disabled thinking by default

Plugin

New Features
OpenClaw local plugin:
  • Added local memory data export
  • Added automatic recall Phase 1 retrieval parameter configuration
  • Added Chinese keyword tokenization configuration
  • Added a dedicated model configuration for L3 abstraction
Improvements
OpenClaw local plugin:
  • Optimized the plugin startup and registration flow
  • Routed embedding inputs by usage
  • Reduced unnecessary background evolution calls in lightweight memory mode
  • Added Chinese README and usage documentation
Bug Fixes
OpenClaw local plugin:
  • Fixed CPU spikes caused by vector scans in large-data scenarios
  • Fixed recovery when SQLite native binding versions mismatch
  • Fixed stability issues around Hermes provider links, OpenAI-compatible endpoint probing, FTS query escaping, and plugin entry publishing
v2.0.22

MemOS Cloud

New Features
Preference memory extraction:
  • Supports independent model configuration
Bug Fixes
Preference memory extraction:
  • Tightened the preference memory prompt to fix over-extraction issues
  • Improved preference memory post-processing
Chat API compatibility:
  • Added string input support for the Chat API content field

Plugin

Bug Fixes
OpenClaw local plugin:
  • Fixed stability, recall accuracy, bridge process leaks, configuration compatibility, and local memory write issues in the OpenClaw local plugin flow
  • Added preference memory and automatic recall capabilities
v2.0.21

MemOS Cloud

Improvements
search context_format update:
  • Renamed the heuristic_routing mode in context_format to mixed, with behavior unchanged
v2.0.20

MemOS Cloud

Improvements
search context_format support:
  • memory: returns the extracted memory text with a shorter context, and is used by default
  • sources: returns the original sources corresponding to the memory, better suited for questions that need original details, time, quantities, or order
  • heuristic_routing: automatically selects memory or sources based on the query, preserving the benefits of the original text where possible while keeping the context length under control
Scheduler optimization:
  • Adjusted scheduler concurrency limits to avoid CPU, thread, and PolarDB contention from background jobs
  • Adjusted queue polling intervals to reduce idle polling and CPU usage
  • Added log and queue switches to reduce log writes on high-frequency search APIs
v2.0.19

MemOS Cloud

New Features
MemOS Cloud CLI Enhancements:
  • Added MemOS Cloud knowledge base APIs with support for up to 12 Agents
Improvements
Dashboard Experience:
  • Improved quick start display logic
  • Added one-click integration prompts

Documentation

Improvements
Documentation:
  • Added a conversational guide for integrating MemOS Cloud into Agents

Plugin

Improvements
Local Plugin:
  • Optimized the local plugin algorithm for domain-aware task handling and prompt injection
v2.0.17

MemOS Cloud

New Features
Dream Core Release:
  • Dream context nodes after fine-mode add
  • Context binding and summarization in Dream phase
  • Search can recall context nodes with a toggle
  • Dream diary fixes and context content
  • Dream registered in plugins with pipeline fixes
Time-sensitive Memory:
  • Batched LLM write-back for time-related information to reduce LLM calls
Bug Fixes
Skill Generation:
  • Skills now require at least 5 rounds of tool-call history before generation
  • Fixed scheduler task loss and Redis serialization issues
v2.0.16

MemOS Cloud

Improvements
Keyword Search Token Selection:
  • Dynamically adjusts token count during keyword search
Search-side Memory Version:
  • Added search-side memory version switch
  • Removed low-impact sidecar recall that hurt search performance
PolarDB Storage Optimization:
  • Horizontally partitioned user memory data with dual-write and read-switchover migration for hundred-million scale performance
Bug Fixes
Memory Version Fixes:
  • Fixed duplicate version updates from a single add
  • Fixed unit test issues

Documentation

Improvements
Documentation Updates:
  • Added open-source docs under docs/ and examples/README.md
  • OpenClaw Plugin docs improvements
  • Added CONTRIBUTING.md with contributor guidelines

Open Source

Improvements
Test Scripts:
  • Added tests/mem_feedback and tests/reranker
v2.0.15

MemOS Cloud

New Features
Time-sensitive Memory Version:
  • Extracts timespec and memory_form on write; time-aware version selection and rerank on search
Skill Upload:
  • Users can upload skills via the add API and retrieve them via search
Improvements
Keyword Search:
  • jieba extract_tags for Chinese queries (max 12 tokens)
  • Tokenizer/regex tokenization for English (max 12)
  • Mandatory user_name validation to prevent cross-user scans

Open Source

New Features
Skill Upload Support:
  • Build skill memories from downloadable skill links with retrieval support
v2.0.14

MemOS Cloud

Bug Fixes
Memory Version:
  • Fixed memory version applying across chunks from the same source text

Open Source

New Features
Plugin Framework:
  • Introduced a lightweight plugin framework with @hookable hooks and entry_points discovery
Improvements
Scheduler Optimization:
  • Cache caps, lock splitting, refill race fixes, stream cleanup, and Redis calls moved outside locks
Bug Fixes
PolarDB & MemReader:
  • Fixed PolarDB metadata nesting bug
  • Fixed memreader chunking bug
LLM JSON Parsing:
  • Fixed LLM JSON output parsing issue (#1526)
v2.0.13

MemOS Cloud

Improvements
Service Optimization:
  • Scheduling log level tuning
  • Structured queue logs
  • Key-path observability for submit -> dequeue -> dispatch
v2.0.12

MemOS Cloud

Improvements
Service Optimization:
  • Configurable minimum thread pool size
  • LLM-based memory version routing with improved skip recall
v2.0.11

Open Source

New Features
LLM & Search:
  • MiniMax LLM (#1291)
  • Health check API (#1248)
  • delete_memory filters (#1104)
  • Tavily search (#1357)
  • Dockerfile & Helm chart (#1293)
Improvements
Scheduler & API:
  • Thread pool and PolarDB optimizations
  • Better API validation errors
  • Consolidated APIs into server_api (#1344)
Bug Fixes
Database & Memory:
  • Neo4j pre-filter fix (#1359)
  • Nested metadata write fix (#1178)
  • Feedback archival fix (#1243)
MCP & API:
  • MCP add_memory format fix (#1275)
  • get_further_suggestion string input fix (#1218)
v2.0.10

MemOS Cloud

Improvements
PolarDB & MemReader:
  • PolarDB multi-filter query optimization
  • MemReader updated to 0.6B model with lower latency
Memory Version Service:
  • Memory version pluginization
  • NLI high availability
  • MemReader routing for memory version plugin
Scheduler:
  • Scheduler improvements for better memory usage
v2.0.9

MemOS Cloud

Improvements
Memory Version:
  • Feedback-compatible memory version
  • Improved independent-fact recovery during conflict resolution
Bug Fixes
Memory Version & Feedback:
  • Parameter validation and timestamp fixes
  • Fixed empty sources during feedback
v2.0.8

MemOS Cloud

New Features
Memory Version:
  • Automatic conflict/duplicate detection, merge, and archive on add (enabled in this release)
Improvements
Preference Memory Migration:
  • Preference memories migrated to the graph database
Memory Version Async:
  • Memory version fully async without impacting addMessage latency
Memory Search:
  • Reduced redundant get_by_metadata calls during search
Bug Fixes
Add & Search Fixes:
  • Fast-mode embedding fix for MultiModalStruct
  • Image memory parsing fix
  • Concurrent add merge ordering fix

Open Source

Improvements
Performance:
  • Full-text warmup
  • Semaphore-based concurrency limiting
  • Database log trimming
v2.0.7

MemOS Cloud

New Features
Memory Version:
  • Automatic conflict/duplicate detection, merge, and archive on add
Improvements
Relativity Defaults:
  • Default relativity threshold set to 0.45 for search and chat
Logging:
  • Adjusted log levels and suppressed noisy errors
Bug Fixes
Preference Embedding:
  • Fixed preference embedding recomputation to only process missing entries
v2.0.6

MemOS Cloud

New Features
Keyword Search:
  • Added keyword recall path while keeping embedding-based relativity
Chat Relativity:
  • Chat API accepts relativity to control retrieval threshold
Bug Fixes
API & Model Fixes:
  • Preference threshold field fix
  • get_memory filter logic fix
  • LLM extra_body / enable_thinking compatibility

Open Source

New Features
Multi-perspective Memory:
  • Each agent memory cube can maintain its own subjective perspective
Multi-perspective Demo:
  • Ao Tai Line hiking mini-game demo
v2.0.5

MemOS Cloud

New Features
Search Relativity:
  • Added relativity field (0~1) to filter recalled memories by relevance score
Improvements
Skills:
  • Local skill document storage with URLs
  • Richer skill generation from user history

Playground

Bug Fixes
Playground:
  • Fixed known UX issues and improved environment stability

Plugin

New Features
MindDock Cloud Skills:
  • Skill page entry in plugin UI
  • Real-time Skill prompt injection across chatbots

MCP

New Features
MCP Memory Deletion:
  • Delete intent triggers both deleteMemory and addFeedback

Open Source

New Features
Knowledge Base Search:
  • Dual-track raw text + memory retrieval
  • Context neighbor recall for long documents
Improvements
Search Pipeline Refactor:
  • Search -> Enhance -> Rerank -> Filter pipeline
  • Modular schedulers with Redis Streams fixes
v2.0.4

MemOS Cloud

New Features
Skill Memory:
  • Extract skill memories from conversation history with downloadable skill files
chat/completions:
  • OpenAI-compatible chat/completions endpoint for no-code platforms
Bug Fixes
Cloud Stability:
  • Fixed occasional add-memory stalls
  • Improved graph database efficiency

Playground

Bug Fixes
Playground:
  • Fixed system-busy error in Playground chat

Open Source

New Features
Conflict / Duplicate Detection:
  • Related-memory recall and NLI-based conflict detection on add
Improvements
Search Deduplication:
  • Reduced duplicate fact and preference recalls
v2.0.3

MemOS Cloud

Improvements
Multimodal Memory:
  • Image parsing now uses conversational context for better memory extraction
Knowledge Base:
  • JSON / Markdown / XML file support
  • knowledgebase_ids: all for cross-project retrieval
Bug Fixes
search/memory:
  • Fixed failures when search input text is too long

Documentation

New Features
Open Source Docs:
  • Documentation fixes and tips
  • Server router startup examples
  • API docs published on Apifox
v2.0.2

MemOS Cloud

Improvements
Knowledge Base Recall:
  • Better fine-grained document detail recall
Tool Memory:
  • Procedural tool-call experience
  • Tool schema compression (>50% token reduction)
Fact Memory Merge:
  • Merge and archive mechanism to reduce duplicate writes/recalls
get/memory:
  • Tool memory in responses
  • Filter-based memory listing for debugging
Bug Fixes
get/memory:
  • Fixed error when include_preference=False

Documentation

New Features
Knowledge QA Assistant Guide:
  • Added best-practice tutorial for building a MemOS knowledge-base assistant
v2.0.1

MemOS Cloud

New Features
Get Memory API:
  • List all user memories with pagination
Chat on Cloud:
  • Chat API with memory-aware replies
Improvements
Delete Memory API:
  • Delete any memory type without specifying user
Preference Feedback:
  • Feedback API now updates preference memories

Playground

New Features
Memory Deletion:
  • Manual memory deletion in Playground memory management

Open Source

Improvements
Search Dedup:
  • Optional dedup parameter (no/sim) for semantic deduplication
Bug Fixes
Scheduler & KV Cache:
  • Local scheduler without Redis config fix
  • KV cache compatibility fix
v2.0.0

MemOS Cloud

New Features
Knowledge Base:
  • Multi-dimensional knowledge bases from documents or URLs
  • Share knowledge bases across projects
Feedback & Deletion:
  • add_feedback for natural-language memory edits
  • Cloud delete memory API by memory id
Multimodal & Tool Memory:
  • Image message memory
  • Tool call trajectory memory
  • Document and URL parsing via MemReader
Improvements
Database Upgrade:
  • Connection management and batch insert improvements
  • Custom tag/info filters on Search
Scheduler Upgrade:
  • Redis Streams multi-queue fair scheduling
  • Pending task auto-recovery

MCP

New Features
MCP Delete & Feedback:
  • MCP supports delete_memory and add_feedback

Open Source

New Features
Chat API:
  • Streaming and non-streaming Chat APIs
Improvements
Lightweight Deployment:
  • Fast and full deployment modes
Bug Fixes
Scheduler Tasks:
  • Legacy scheduler compatibility fix
  • Memory update task logging fix
v1.1.3
New Features
Memory Add & Search:
  • Added async mode (plaintext & preference)
  • Preference Memory now supported
  • Reranker strategy suite
  • BM25 for TreeTextMemory
Scheduler:
  • Modularized API scheduler
  • Redis ORM optimized for history sync & hybrid search
Data & Infra:
  • PolarDB graph backend connection pool/timeout & fixes
  • Unified graph factory (Neo4j/PolarDB/Nebula)
  • Milvus interface & item optimizations
  • Enhanced logging pipeline
  • Nacos-based dynamic configuration
Evaluation:
  • PrefEval field standardization
  • LoCoMo/LongMemEval/PrefEval/PersonaMem evaluation upgrades
Improvements
Plaintext Memory:
  • Standardized preference fields
Framework:
  • Updated API routes
  • More robust error handling & context tracing
Bug Fixes
Scheduler:
  • Fixed query scheduling edge case
  • Corrected message schema inconsistencies
Plaintext Memory:
  • Graph/DB：corrected PolarDB issues
Framework:
  • Fixed SQLite user listing bug
v1.1.1
New Features
Cloud Service 1.0 Beta Release:
  • 🛠️ Ready-to-Use：Deploy directly via cloud API without complex setup
  • 🔄 Cross-Session Memory：Automatically retrieve user profiles, preferences, and behaviors for continuous personalization
  • 📖 Developer-Friendly：Comprehensive documentation and example code support
  • 🎁 Free Quota：Every developer can receive basic quota to explore
Enterprise Group Q&A bot:
  • Supports Q&A based on cloud service knowledge base
v1.0.1
New Features
Group Q&A Bot:
  • Launched group Q&A bot based on MemOS Cube
KV-Cache Performance Optimization:
  • Updated comparative experimental data for KV-Cache on different GPU deployment schemes
  • Optimized test benchmarks and statistical methods
Plaintext Memory Enhancement:
  • Added Reranker sorting functionality for plaintext memory
Playground Updates:
  • Updated Playground version with all the above new features
Improvements
Plaintext Memory Optimization:
  • Optimized plaintext memory hallucination issues
v1.0.0
New Features
Playground:
  • Expanded Playground features and algorithm performance.
MemCube Construction:
  • Added a text game demo based on the MemCube novel.
Extended Evaluation Set:
  • Added LongMemEval evaluation results and scripts.
Improvements
Plaintext Memory:
  • Integrated internet search with Bocha.
  • Added support for Nebula database.
  • Added contextual understanding for the tree-structured plaintext memory search interface.
Bug Fixes
KV Cache Concatenation:
  • Fixed the concat_cache method.
Plaintext Memory:
  • Fixed Nebula search-related issues.
v0.2.2
New Features
Explicit Memory:
  • Implemented internet search integration with Nebula database support
  • Enhanced contextual understanding for memory extraction
  • Deployed internet search interface integrated with memreader for web message processing
KV Cache:
  • Conducted KV Cache evaluation including LMCache comprehensive research and performance testing
  • Completed stress testing for MemOS vllm (v0.2.1) in specific environments and models
  • Refined ttft test data for preloaded KV Cache on Qwen2.5-72B-Instruct model and updated report
  • Completed initial research and design for high-ROI inference service system using domestic hardware
Memory Operation Model:
  • Completed model training evaluation and release
  • Released 4b, 1.7b and 0.6b models supporting memory extraction and integration
Improvements
Documentation:
  • Added English version for first three chapters of Cookbook
Memory Scheduling:
  • Performed code refactoring and functionality improvements
  • Refactored functional modules including monitor, dispatcher and retriever
  • Enhanced network logging functionality
  • Improved scheduler robustness with new exception-catching decorators
  • Implemented locking for shared resources
Development Environment:
  • Updated Docker configuration
  • Updated dim environment configuration
API Enhancements:
  • Added playground context support
  • Enhanced product API functionality
  • Rewrote query module
  • Implemented chat history feature
Bug Fixes
Data Parsing:
  • Fixed date parsing error
  • Fixed memos_w_scheduler example code issues
  • Fixed fine-grained search bug in Nebula graph database
  • Fixed metadata filter retrieval logic
System Integration:
  • Aligned MOSProduct._build_system_prompt signature with MOSCore
Explicit Memory:
  • Fixed general text memory processing logic
  • Fixed memreader component issues
v0.2.1
New Features
MemCube Features:
  • Added plaintext memory + KV Cache functionality with reasoning and decoding performance report
  • Completed development of interactive cube feature with full workflow support and embedding model switching
MemOS System:
  • Released MemOS Neo lightweight version with simplified architecture and core API modules
  • Added MCP support research and preparation work, expanding MCP capabilities
  • Created embedded Agent flow Pipeline documentation with Coze framework integration
Deployment & Environment:
  • Added Docker deployment support
  • Enhanced API layer with multi-model support, compatible with OpenAI/Qwen/DeepSeek and other mainstream models
  • Improved embedding model support
  • Added neo4j Community Edition/Nebular database support
  • Implemented multi-tenant architecture support within single database
Evaluation & Testing:
  • Adapted and evaluated memos API interface format, updated evaluation code
Documentation & Examples:
  • Added new Cookbook content
  • Included Mud game example
v0.2.0
New Features
Website Usability Updates:
  • Added documentation search functionality
  • Implemented Chinese/English documentation switching
  • Enhanced footer navigation links and page editing features
Memory Operator Model:
  • Introduced MemReader-4B small model for memory extraction operations
  • Enabled fully local deployment for restricted network environments
  • Achieved lower cost and faster memory operations with performance exceeding GPT-4o-mini
  • Fine-tuned based on Qwen3-4b with human and model annotation data using supervised fine-tuning
Cross-Platform Framework Adaptation:
  • Added Windows platform deployment support
  • Added Mac platform deployment support
  • Completed adaptation for Linux, Windows, and macOS mainstream operating systems
  • Passed testing on Ubuntu 20.04+/CentOS, Windows 10+/11, macOS 14 Ventura+ platforms
  • Supports stable operation of core modules (memory lifecycle management, MIP protocol, memory cache scheduling)
Playground Progress:
  • Completed engineering, front-end, and algorithm end-to-end connection, and bug fixing and optimization in progress