CaviraOSS/OpenMemory

Features

• Agent Memory Engines - Provides an embeddable memory engine that stores, retrieves, and manages conversational context and agent state using semantic indexing and temporal facts.
• Agent State Persistence - Provides a persistence layer that indexes agent state snapshots semantically for retrieval by meaning.
• Agent Memory Storage - Persists content with tags, metadata, user identity, and importance scores for later retrieval.
• Agentic Execution Loops - Orchestrates a cycle of observing input, recalling past experiences, planning actions, executing tools, and storing results for persistent context.
• Stateful - Provides a stateful chatbot builder that persists user context and conversation history across sessions using a dedicated memory store.
• Temporal Fact Management - Records and queries factual statements with subject-predicate-object structure and validity periods.
• Per-User Memory Isolation - Lists all memory entries scoped to a specific user identifier for retrieval.
• Memory Relevance Controls - Retrieves the most contextually relevant memories by comparing semantic similarity against stored vectors.
• Memory Decay Policies - Gradually reduces memory salience over time unless reinforced by subsequent queries.
• Stateful Agent Orchestrators - Coordinates a cycle of observation, recall, planning, tool execution, and storage to maintain persistent conversational context.
• Semantic Indexing for Agent State - Indexes agent state snapshots semantically for retrieval by meaning rather than by key.
• Temporal Fact Creation - Records factual statements with subject-predicate-object structure and validity periods.
• Temporal Fact Stores - Records factual statements with subject-predicate-object structure and explicit validity windows for time-aware knowledge queries.
• Temporal Knowledge Graphs - Records factual statements with subject-predicate-object triples and explicit validity windows for time-aware knowledge queries.
• Semantic Search - Retrieves relevant memories by comparing semantic similarity against stored vectors with filtering.
• Memory Search Engines - Finds relevant stored content by semantic query, optionally filtered by sector, user, or minimum salience.
• Salience-Based Decay Engines - Applies graduated salience reduction rules across recency tiers, compressing cold vectors and fingerprinting dormant entries to conserve storage.
• Vector Indexing - Stores memory content as high-dimensional vectors for similarity-based retrieval rather than exact keyword matching.
• Memory Record Creators - Stores new memory records with user-defined content, metadata, and salience settings for later retrieval.
• Memory Record Deleters - Removes memory records along with their vector embeddings and associated waypoints from the store.
• Semantic Memory Stores - Ships a vector-backed storage system that indexes memories by meaning and supports salience-based decay, compression, and reinforcement.
• Conversational Chatbots - Builds chatbots that retain user context and conversation history across sessions using persistent memory.
• Memory Record Updaters - Modifies the content or metadata of previously stored memories while preserving their vector representations.
• Memory Listers - Returns paginated collections of memories, optionally filtered by sector or user identifier.
• User Memory Deleters - Removes every memory entry associated with a given user identifier from the store.
• Memory Compression - Automatically condenses verbose memory content while preserving its core semantic meaning.
• Fingerprinted Memory Regenerators - Restores fingerprinted memories to full representation when accessed, enabling efficient cold storage with on-demand regeneration.
• Memory Reinforcement Mechanisms - Manually increases the salience score of a specific memory to boost its relevance in future queries.
• Temporal Fact Invalidation - Closes the validity period of a factual statement, marking it as no longer current.
• Temporal Fact Update - Modifies the confidence score or metadata of an existing factual statement.
• Temporal State Tracking - Maps each state change to a timeline, creating a temporal graph of how agent memory evolves.
• Temporal Knowledge Analyzers - Provides entity timelines, pattern searches, date comparisons, graph statistics, and identifies the most volatile facts.
• Importance-Ranked Retrievers - Ranks memories by importance so that the most relevant ones are surfaced first during recall.
• Query-Based Memory Reinforcers - Boosts the salience of memories matched during a query to prevent useful information from fading away.
• Temporal Fact Queries - Retrieves stored factual statements filtered by subject, predicate, object, or a specific point in time.
• Webpage Ingestion - Uploads files or scrapes URLs to extract and store content as searchable memories.
• Current Fact Retrieval - Returns the most recent version of a factual statement for a given subject and predicate pair.
• Memory Salience Boosters - Implements emotional salience boosting that adjusts memory importance scores based on detected sentiment intensity.
• Embedded Memory Engines - Runs the memory engine as an embedded library within a host application, eliminating the need for a separate backend server.
• Memory Fingerprint Compressors - Ships fingerprint compression for cold storage that replaces dormant memory vectors with compact fingerprints restorable on access.
• Memory Cluster Summarizers - Periodically clusters related memories and synthesizes condensed summaries through automated semantic reflection.
• Emotional Salience Boosters - Increases the importance of memories that carry strong positive or negative sentiment scores to prioritize emotionally charged content.

Star history