HomeProjectsArticlesPK BoardAbout
Explore Projects

MemVid

Normal
GitHub Rust Apache-2.0

Description

MemVid is a long-term memory layer for AI agents that uses video encoding for lightweight single-file storage, replacing complex RAG pipelines with instant retrieval.

Key Features

  • Single-file portable memory layer — no database or server infrastructure required
  • Smart Frame architecture inspired by video encoding for append-only, crash-safe storage
  • Sub-5ms retrieval latency at scale with predictive caching (0.025ms P50)
  • Multi-language SDKs: Rust core, Python SDK, Node.js SDK, and CLI tool
  • Time-travel debugging to rewind, replay, or branch any memory state
  • Feature-rich with full-text search (BM25), vector similarity, CLIP image search, and encryption

Use Cases

💡 Providing persistent long-term memory for AI agents across sessions
💡 Building enterprise knowledge bases with instant retrieval and versioning
💡 Enabling offline-first AI systems that carry memory without cloud dependencies
💡 Supporting customer support agents with context-aware conversation history
💡 Auditing and debugging AI workflows with traceable memory snapshots

Strengths & Limitations

✅ Strengths

  • • High community interest (16.2k stars)
  • • Permissive open-source license (Apache-2.0)
  • • Responsive to issues, low backlog

Tags

memory rag vector-database agent python

Categories

🧩 Agent Memory 📚 RAG Tools

Quick Start

Install with `cargo add memvid-core` (Rust), `pip install memvid-sdk` (Python), or `npm install @memvid/sdk` (Node.js). Create a `.mv2` file, add documents with metadata, commit, then search. Supports feature flags like `lex`, `vec`, `clip`, and `encryption`.
Visit GitHub Visit Website

Related Projects

HyperDB

1.4k · Python
Stale C

A hyper-fast local vector database for use with LLM Agents, providing lightweight vector storage and similarity search capabilities for embedding as instant memory and knowledge retrieval components in agent applications.

vector-databasepythonagent +2
  • · Hyper-fast local vector database designed specifically for LLM Agents
  • · C++ high-performance backend with hardware-accelerated operations via MKL BLAS
  • · Simple unified interface compatible with all major LLM agent frameworks
View on GitHub

MemAgent

1.1k · Python
Stale B

A MemAgent framework that can extrapolate to 3.5M context tokens, along with a training framework for RL training of any agent workflow.

memoryagentrag +2
  • · Ultra-Long Context Processing: Extrapolate from 8K training context to 3.5M tokens with performance loss under 5%
  • · Reinforcement Learning Driven: Trained with RLVR (Reinforcement Learning from Verifiable Rewards), extends DAPO algorithm for end-to-end multi-turn conversation optimization
  • · Linear Time Complexity: Breaks through computational bottlenecks in long-text processing with linear resource scaling
View on GitHub

TrustGraph

2.5k · Python
Active A+

A graph-native context development platform for storing, enriching, and retrieving structured knowledge with semantic search and portable context cores, supporting RDF, SPARQL, and other standards for AI agent knowledge management.

ragmemorydatabase +3
  • · Context Graph engine with automated entity/relationship extraction and ontology-driven graph construction
  • · Out-of-the-box RAG pipelines: DocumentRAG, GraphRAG, and OntologyRAG
  • · Multi-model database system supporting tabular, document, graph, vector, image, video, and audio data
View on GitHub

Embedchain

63.4k · Python
Active A+

Embedchain is a universal memory layer for AI agents, enabling quick integration of diverse data sources into LLMs for context-aware AI applications.

memoryragembeddings +2
  • · Universal memory layer — quickly integrates diverse data sources (web, PDFs, YouTube, Notion) into LLMs for context
  • · Vectorized storage — auto-chunks, embeds, and stores data in vector databases with semantic retrieval support
  • · Multi-LLM backend support — compatible with OpenAI, Cohere, Ollama and other LLM and embedding models
View full details → View on GitHub

Related Articles

Memory记忆系统长期记忆

Agent Memory Architecture: Working, Long-term, and Shared Memory Trade-offs

A systematic comparison of the three categories of agent memory -- working, long-term, and shared -- covering storage media, lifecycle, retrieval methods, typical frameworks, and design patterns, fully addressing agent personalization and multi-agent collaboration engineering.

Read more →