8 Repos
Tiered memory systems using pluggable backends for context management.
Explore 8 awesome GitHub repositories matching artificial intelligence & ml · Composable Memory Architectures. Refine with filters or upvote what's useful.
LangChain is an orchestration framework designed for building, managing, and deploying applications powered by large language models. It provides a unified integration layer that normalizes disparate model provider APIs into a consistent set of primitives, enabling developers to build complex, multi-step AI workflows that manage state, memory, and tool execution. The project distinguishes itself through a durable execution runtime that maintains persistent state across long-running processes by checkpointing progress to external storage. It models agent workflows as directed graphs, allowing
Separates short-term context from long-term storage using pluggable backend interfaces for tiered memory architectures.
AgentMemory is a persistent knowledge store and memory server designed to provide AI coding agents with long-term memory. It functions as a knowledge graph engine and vector database store that saves and recalls project context, architectural decisions, and patterns across different sessions. The system distinguishes itself by using a tiered-memory consolidation pipeline that compresses raw observations into episodic, semantic, and procedural layers to optimize token usage. It employs a hybrid retrieval strategy combining keyword matching, vector embeddings, and graph traversal to surface rel
Organizes raw observations into episodic, semantic, and procedural memory tiers to optimize token usage.
Omi is an open-source wearable AI platform that captures audio and screen data to provide real-time conversational assistance and memory. It integrates a wearable hardware development kit with a vector memory database and large language model capabilities to create a persistent digital record of user interactions. The platform is distinguished by its BLE audio streaming pipeline, which transmits raw audio from wearable hardware for real-time transcription and speaker identification. It utilizes a plugin-based agent tool framework that allows AI assistants to autonomously invoke custom functio
Classifies user questions into distinct paths to determine if personal memory retrieval or specific persona logic is required.
memU is a long-term memory system for AI agents that provides a persistent knowledge base. It extracts facts and preferences from conversations into structured memories, organizing this information through a hierarchical knowledge base based on a file-system architecture of nested categories and linked resources. The system includes a multimodal data ingestion pipeline that converts audio, video, and images into standardized natural language for storage in large language model contexts. It also features a model provider abstraction layer, offering a unified interface to use interchangeable la
Implements a tiered memory system using pluggable providers for long-term storage and retrieval.
Evolver is a self-evolving AI agent framework that uses gene expression programming to autonomously improve agent behaviors through a continuous five-step loop of scanning, selecting, mutating, validating, and solidifying. It functions as an auditable evolution system that records every mutation and selection step, and can translate natural-language problems into executable Python code for automated grading and evaluation. The framework distinguishes itself through a distributed architecture that enables multiple agents to collaborate and share learned experiences across a network. It operate
Maintains a three-tier memory system with persistent facts, procedural knowledge, and session history.
Promptbase ist ein Prompt-Engineering-Framework, das für das Design, das Testen und die Optimierung von Prompts für Large Language Models entwickelt wurde. Es bietet ein System zur Messung der Modellgenauigkeit und -leistung durch ein Evaluierungs-Toolkit, das Ausgaben mit Ground-Truth-Datensätzen vergleicht. Das Projekt enthält zudem eine Orchestrierungspipeline für die Automatisierung von Machine-Learning-Aufgaben mit mehreren Komponenten über Cloud-basierte Endpunkte sowie ein Dienstprogramm zur Vorbereitung von RAG-Datensätzen (Retrieval-Augmented Generation). Das Framework zeichnet sich durch fortgeschrittene Optimierung der Antwortqualität aus und nutzt Chain-of-Thought-Generatoren, um zwischenzeitliche Argumentationsschritte zu erzeugen, sowie dynamisches Few-Shot-Beispiel-Retrieval unter Verwendung embedding-basierter semantischer Suche. Es implementiert Ensemble-Methoden zur Steigerung der Vorhersagegenauigkeit und verwendet komplexitätsbasiertes Query-Routing sowie Mehrheitsentscheidungs-Aggregation mehrerer Modellvariationen. Das System deckt breitere Funktionen im Datenmanagement und der Automatisierung ab, einschließlich der Formatierung externer Daten in strukturierte Dateien für das Training und der Orchestrierung von Modellausführungspipelines über Kommandozeilen-Dienstprogramme.
Analyzes query complexity to dynamically select the most effective prompting technique or reasoning path for a given input.
MemOS is an open-source persistent memory layer for AI agents and large language models, providing a self-hosted server that stores and retrieves structured memory across sessions. It enables AI systems to recall user preferences, history, and context without retraining, using a graph-based API and a web management interface for viewing, editing, and organizing memory items, skills, traces, and knowledge bases. The system distinguishes itself through a portable memory interchange protocol that allows memory to be transferred between different AI models, devices, and applications, along with a
Organizes memory into three tiers with separate retrieval channels for Skills, Traces, and World Models.
OpenSquilla ist ein LLM-Agent-Orchestration-Framework zur Koordination mehrstufiger KI-Workflows und Tool-Ausführungen mittels gerichteter azyklischer Graphen. Es fungiert als zentrales System zur Verwaltung spezialisierter Skill-Pakete und zur Ausführung komplexer Reasoning-Sequenzen. Das Projekt zeichnet sich durch ein Routing-Gateway aus, das Aufgaben basierend auf Komplexität, Kosten und Performance an verschiedene KI-Anbieter weiterleitet. Es nutzt ein mehrstufiges KI-Gedächtnissystem, das Arbeits-, episodisches und semantisches Wissen mittels lokaler Embeddings und SQLite organisiert, sowie eine sichere Ausführungsumgebung (Sandbox), die Agent-generierten Code über risikobasierte Berechtigungsprofile isoliert. Die Plattform deckt ein breites Spektrum an Funktionen ab, einschließlich Multi-Channel-Deployment für Web- und Messaging-Plattformen, automatisierter Aufgabenplanung via Cron und einer Model Context Protocol-Bridge zur Anbindung externer Tools. Zudem bietet sie umfassende Monitoring- und Observability-Tools zur Verfolgung von Token-Kosten, zum Auditing von Laufzeitentscheidungen und zur Verwaltung eines Katalogs wiederverwendbarer Skills. Das System enthält CLI-Utilities für die Workspace-Initialisierung und das Skill-Lifecycle-Management.
Implements a three-tier memory architecture using working, episodic, and semantic layers with automated consolidation.