memvid/memvid

Project overview

A serverless, single-file memory layer that packages data, embeddings, and search structures into a portable file for retrieval without relying on server-based databases. It is designed for developers who need to give AI agents persistent memory while avoiding heavy infrastructure.

Project type

RAG · Vector Database · Data Processing

Use cases

Knowledge Q&A · Automation

Deployment

Refer to project documentation

License

Apache-2.0

Best for

• Developers and AI engineers building retrieval workflows who require a library-based memory layer.
• Agent automation tasks requiring rewinding, replaying, or branching past memory states during execution.

Key capabilities

• Packages data, embeddings, search structure, and metadata into a single file without sidecar files.
• Rewind, replay, or branch any memory state for queries over past memory states.
• Uses CLIP visual embeddings for image search.
• Provides audio transcription with Whisper.

Limitations and risks

• Requires Rust 1.85.0+ for building from source.
• Cloud embeddings via OpenAI API are optional, which may introduce a paid service cost dependency.

Getting started

• Add the core crate to Cargo.toml dependencies, create a new memory file using the create method, add documents, and execute a search. Setup requires programming knowledge.

Evidence and sources

• GitHub project description: Memory layer for AI Agents. Replace complex RAG pipelines with a serverless, single-file memory layer.
• README: [dependencies] memvid-core = "2.0"
• README: Instead of running complex RAG pipelines or server-based vector databases, Memvid enables fast retrieval directly from the file.
• README: Memvid is a portable AI memory system that packages your data, embeddings, search structure, and metadata into a single file.
• Release: v2.0.140

Reviewed at: 2026-07-17T14:13:04Z