milla-jovovich/mempalace
Mempalace
Mempalace is a local-first long-term memory store for large language models and AI agents. It provides a persistent storage system for verbatim conversation history and agent data, utilizing a local-first knowledge graph to track evolving entity relationships and timelines.
The project implements a standardized memory protocol that allows external AI clients to read and write persistent memory via standard input and output. It features a hybrid semantic search engine that combines keyword boosting and reranking to find precise historical information across scoped categories.
The system includes a pluggable vector database interface to connect agents to local files or remote stores. Capability areas cover agent context management through scoped namespace isolation, high-fidelity conversation indexing, and LLM-based result reranking to prioritize contextually relevant sessions.
Features
- Long-term Memory Stores - Implements a persistent storage system for retaining user preferences and interaction history across multiple sessions.
- Local-First Storage - Provides a local-first storage system for the long-term retention of verbatim conversation history and agent data.
- Agent Memory Systems - Manages persistent storage and context isolation to maintain interaction history for autonomous agents.
- Agentic Context Management - Provides mechanisms to isolate memory and knowledge scopes for different AI agents to prevent context bloat.
- Standardized Memory Protocols - Implements a standardized server protocol allowing external AI clients to interact with persistent memory via standard I/O.
- Temporal Knowledge Graphs - Tracks evolving entity relationships and timelines using a local temporal knowledge graph.
- Hybrid Retrieval - Combines vector similarity and keyword boosting to accurately retrieve verbatim historical records.
- Verbatim Transcript Indexing - Indexes original conversation transcripts verbatim to maintain maximum retrieval accuracy for long-term memory.
- Verbatim Conversation Storage - Saves conversation data as exact text without summarizing to ensure high-fidelity retrieval.
- Namespace-Based Isolation - Uses isolated namespaces to separate storage for different agents, preventing context bloat and data leakage.
- Local-First Architectures - Prioritizes local data storage to ensure low latency and high-fidelity memory without external API dependencies.
- Remote Vector Store Namespacing - Links remote vector storage backends using namespaces to isolate tenants and manage long-term memory.
- Model Context Protocol Integrations - Implements the Model Context Protocol to expose memory structures and agent diaries to AI models.
- Semantic Search Engines - Ships a hybrid retrieval pipeline combining keyword boosting and reranking for context-aware historical search.
- Chat History Indexing - Indexes historical conversational data from transcripts to enable precise information retrieval.
- Hybrid Search Infrastructure - Combines vector embeddings with keyword and regex matching to retrieve precise historical information.
- Natural Language Memory Queries - Allows retrieval of verbatim historical records from the memory store using natural language queries.
- Pluggable Storage Backends - Provides an abstraction layer for swapping between different database engines and remote vector stores.
- Local Filesystem Indexers - Scans local directories to build searchable catalogs of conversation logs and files as verbatim records.
- AI Client-Server Communication Protocols - Implements a bidirectional messaging protocol for communication between AI clients and the memory server.
- Storage Layer Coordination - Coordinates different storage layers to manage how AI agents archive and access historical data.
- Memory Management - High-performance AI memory system.
Star history
How this analysis was created: This summary and feature list are AI-generated from collected project material and can contain mistakes. Stars, license and language are imported from GitHub. Inclusion does not mean that we have tested or audited this project. Check the source documentation for any feature you depend on. Learn more on our About page.
