Open Issues Need Help
View All on GitHubAI Summary: This GitHub issue outlines the implementation of a PostgreSQL storage backend for 'memory metadata' and 'procedural memory'. The work involves designing database schemas for memories and procedures, creating an abstract `MetadataStore` base class with a concrete `PostgresStore` implementation, and developing robust CRUD, bulk operations, schema migration handling, and connection pooling.
Memory that forgets, like humans do. A production-grade memory system for AI agents with intelligent forgetting.
AI Summary: This issue proposes adding a `pgvector` backend to allow users to store both metadata and vector embeddings within a single PostgreSQL database. The implementation requires a `PgVectorBackend` class supporting standard vector store operations like upsert, search, and delete, along with advanced features such as HNSW and IVFFlat index types, `asyncpg` connection pooling, batch operations, configurable distance metrics, and comprehensive index management (create, rebuild, vacuum). The primary benefit is a simplified deployment by consolidating data storage.
Memory that forgets, like humans do. A production-grade memory system for AI agents with intelligent forgetting.
AI Summary: This issue proposes adding a SQLite storage backend for the `MetadataStore` component, specifically designed for simple, single-user deployments without external dependencies. The implementation will include full CRUD operations, schema creation, safe handling of concurrent access within a single process, and an in-memory mode for testing.
Memory that forgets, like humans do. A production-grade memory system for AI agents with intelligent forgetting.
AI Summary: This issue requests the creation of eight distinct example scripts and notebooks to demonstrate various features of the 'Cognitive Memory' library, including basic usage, custom configuration, integrations with LangGraph and LangChain, multi-session conversations, multi-user isolation, memory consolidation, and REST API usage. Each example must be self-contained, well-commented, show expected output, handle errors, and use environment variables for secrets.
Memory that forgets, like humans do. A production-grade memory system for AI agents with intelligent forgetting.
AI Summary: This issue requests the creation of comprehensive unit and property-based tests for the core engines: Decay, Importance, Retrieval, and Consolidation. It outlines specific behaviors and properties to test for each engine, including aspects like monotonic decay, factor computation, MMR diversity, and mocked LLM fact extraction, utilizing pytest, hypothesis, and pytest-asyncio.
Memory that forgets, like humans do. A production-grade memory system for AI agents with intelligent forgetting.
AI Summary: This issue proposes implementing a LangChain-compatible memory class, `CognitiveMemory`, that integrates with an existing Cognitive Memory system. The class will extend `BaseMemory`, providing methods to load relevant context, save conversation history, and clear memory, while also supporting configurable prefixes and memory keys. The goal is for it to serve as a drop-in replacement for `ConversationBufferMemory`.
Memory that forgets, like humans do. A production-grade memory system for AI agents with intelligent forgetting.
AI Summary: This issue aims to integrate Qdrant as a vector store backend for episodic memory, requiring the creation of an abstract `VectorStore` base class and a concrete `QdrantStore` implementation. The implementation must support full CRUD operations, metadata filtering, connection pooling, and both local and cloud Qdrant instances, along with robust error handling and comprehensive integration tests.
Memory that forgets, like humans do. A production-grade memory system for AI agents with intelligent forgetting.
Memory that forgets, like humans do. A production-grade memory system for AI agents with intelligent forgetting.
AI Summary: This GitHub issue outlines the implementation of a comprehensive configuration system using Pydantic models. The system will define several specific configuration models like `DecayConfig` and `ImportanceConfig`, culminating in a top-level `MemorySystemConfig`, all supporting validation, environment variable loading, and sensible defaults. The task requires creating a single configuration file, extensive unit testing, and ensuring clear error messages for invalid configurations.
Memory that forgets, like humans do. A production-grade memory system for AI agents with intelligent forgetting.
AI Summary: This issue focuses on implementing the foundational data models for a cognitive memory system. It requires creating several Python dataclasses (e.g., `Memory`, `Fact`, `Entity`) and enums (`MemoryType`, `MemorySource`) with type hints and Pydantic validation, based on specifications in an architecture document. The task includes writing unit tests and docstrings for these new data structures.
Memory that forgets, like humans do. A production-grade memory system for AI agents with intelligent forgetting.