ИИ-агенты
agent-memory
Long-term memory runtime for AI agents — plain Markdown as truth, local ranked retrieval, no API key
★ 2,6 тыс. звёздMIT
Смотреть руководство →LongMemory — проект с открытым исходным кодом из категории «ИИ-агенты» от CaviraOSS (CaviraOSS/LongMemory на GitHub). У него 4 523 звезды на GitHub и 506 форков, он распространяется по лицензии Apache-2.0 и последний раз обновлялся на GitHub 20 сентября 2026.
LongMemory: local-first memory for AI agents with temporal truth, provenance, and hybrid recall
Use the npm library in-process, the global CLI, the Docker/HTTP server, or the MCP server for agent clients.
Install the package from npm (or the Python SDK from PyPI).
Create a memory instance or run the CLI init to create a local database.
Ingest a first fact or conversation, then recall it with the mode that fits your question.
Records keep both recorded time and valid time, so facts can change without rewriting history.
Source, author, confidence, and timestamps travel with memory items; nothing is silently lost.
Strict, historical, associative, and world-grounded recall over BM25, vector, and graph signals.
Tenant, user, project, session, and role scopes, with encryption at rest and a tamper-evident audit trail (per the README).
In-process library, CLI, HTTP server on port 7331, Docker image, MCP server, dashboard, and VS Code extension.
OpenAI-compatible endpoints, Google Gemini, Amazon Bedrock, Ollama, or a built-in local embedder.
Clone the project and enter its directory for the source/server path.
git clone https://github.com/CaviraOSS/LongMemory.git
cd LongMemoryAdd it to a Node project, or install the CLI globally.
npm install longmemory
# or: npm install --global longmemoryInitialize a database and run your first recall from the CLI.
longmemory init
longmemory recall "current project priorities" --mode associativeCaviraOSS publishes LongMemory (server images under ghcr.io/caviraoss); it is an open-source project.
Yes, Apache-2.0 licensed and local-first — your data stays in a local SQLite database unless you choose otherwise.
Not necessarily: local and Ollama embedding options avoid hosted APIs; choosing OpenAI, Gemini, or Bedrock embeddings uses your own keys.
A vector store returns similar chunks. LongMemory adds typed records, temporal validity, provenance, governance scopes, and recall modes that combine keyword, vector, and graph signals.
Run `npm install longmemory` and use `createMemory()`, or install the CLI globally and run `longmemory init` then `longmemory recall`.
Long-term memory runtime for AI agents — plain Markdown as truth, local ranked retrieval, no API key
Local-first AI memory that runs offline on just 8 GB of RAM.
Memory-LanceDB-Pro gives OpenClaw agents hybrid search, smart extraction, and intelligent forgetting
Basic Memory: local-first MCP memory so you and your AI tools build a shared knowledge graph in Markdown
Open-source AI memory that gives agents persistent long-term recall.
One portable, local-first memory layer for every AI agent - Markdown-native, user-owned and self-evolving.
Другие проекты из категории «ИИ-агенты» с наибольшим числом звёзд в каталоге LibreAI — реалистичный список альтернатив, если LongMemory вам не подходит.
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
Соседние проекты из категории «ИИ-агенты» в каталоге, ближайшие по числу звёзд к LongMemory — продолжайте изучать полку по обе стороны от него.
Convene 18 legendary thinkers as an AI council that debates your hardest questions — inside Claude Code
M-flow is a bio-inspired cognitive memory engine for AI agents — a graph alternative to flat RAG
Microsoft's open framework for red-teaming generative AI systems.
Pydantic Logfire: Python and AI observability on OpenTelemetry — SQL-queryable traces, metrics, and logs
Koog is a JVM (Java and Kotlin) framework for building predictable, fault-tolerant and enterprise-ready AI agents across all platforms – from backend service…
Agent framework for the JVM. Pronounced Em-BAY-bel /ɛmˈbeɪbəl/
Бесплатно и открыто — забирайте исходный код, ставьте звезду репозиторию и начинайте создавать.