embedchain/embedchain
Embedchain
Embedchain is an LLM memory management framework and RAG orchestration engine designed to provide AI agents with a persistent storage layer. It functions as a long-term memory pipeline that extracts facts from unstructured interactions and stores them as permanent knowledge base entries to retain user preferences and interaction history across sessions.
The system employs a hybrid vector database interface that combines semantic embeddings with traditional keyword search. It utilizes an entity-linking knowledge graph to connect related information points and applies temporal ranking to distinguish current states from historical data.
The framework covers multi-level state management across user, session, and agent tiers and implements multi-signal retrieval to surface relevant context. It includes a command line interface for administering stored data and interaction history.
Features
- Long-term Memory Stores - Provides a persistent long-term memory store that extracts facts from interactions to maintain a knowledge base across sessions.
- AI Memory Layers - Provides a persistent storage layer for AI agents to remember user preferences and interaction history.
- Fact Extraction Pipelines - Ships a pipeline that extracts confirmed facts from unstructured interactions to build a permanent knowledge base.
- Agent Memory Managers - Acts as a comprehensive agent memory manager providing persistent storage for user preferences and interaction history.
- Context-Aware Retrieval - Implements context-aware retrieval by combining temporal ranking and multi-level state management to optimize AI agent responses.
- RAG Pipelines - Functions as a RAG pipeline that orchestrates vector and keyword search to provide relevant context for LLMs.
- Context Partitioning - Partitions context into distinct user, session, and agent layers to maintain state across interaction scales.
- Multi-Level Memory Management - Manages hierarchical memory layers across user, session, and agent tiers to maintain interaction context.
- Fact Extraction Pipelines - Provides a pipeline to process unstructured interactions and isolate confirmed facts as permanent long-term memory entries.
- Hybrid Search Engines - Integrates vector-based semantic retrieval with keyword-based indexing for finding relevant stored information.
- Hybrid Vector-Keyword Indexing - Combines semantic vector embeddings with inverted keyword indices to improve information recall.
- Information Retrieval - Implements a retrieval mechanism to query stored information based on semantic and keyword context.
- Stateful Weight Management - Maintains stateful weight management of memories based on temporal timestamps.
- Retrieval Re-ranking - Applies temporal weighting to re-rank retrieved information, distinguishing current states from historical data.
- Signal Pipelines - Uses signal-based processing stages to aggregate semantic and keyword data for context retrieval.
- Entity Relationships - Provides configurations for defining associations between entities to enable retrieval of complex associated information.
- Temporal - Utilizes temporal tracking to rank retrieved data and distinguish current states from past events.
- Knowledge Graph Retrieval - Utilizes a knowledge graph for entity-relationship retrieval to connect associated data points for AI agents.
- Multi-Signal Retrieval Aggregators - Aggregates semantic, keyword, and entity signals to surface the most relevant context for given queries.
- Graph Linking Systems - Implements a graph-based system to link related information nodes across memory entries for faster retrieval.
- Agent Frameworks - Simplifies the creation of RAG-optimized AI bots.
- Application Development - Framework for creating bots over custom datasets.
- Development Frameworks - Framework for building bots over custom datasets.
- Development Platforms - Simplifies creating custom bots over specific datasets.
- Infrastructure and Utilities - Simplifies creating bots over custom datasets.
- Prompt Engineering Resources - Framework for creating ChatGPT-like bots over custom datasets.
- Prompting Frameworks - Tool for creating custom chatbots over specific datasets.
- Templates and Boilerplates - Framework for creating LLM bots over datasets.
