memvid-cli

Command-line interface for Memvid v2 - AI memory with crash-safe, single-file storage.

Installation

cargo install memvid-cli

Or with specific features:

# Without llama.cpp (for Docker/ARM emulation, Windows ARM, or systems without LLVM)
cargo install memvid-cli --no-default-features --features temporal_track,temporal_enrich,parallel_segments,candle-llm

# With Metal acceleration (macOS)
cargo install memvid-cli --features metal

# With audio playback
cargo install memvid-cli --features audio-playback

Note: The default installation includes llama.cpp which requires LLVM/libclang. If you encounter build errors related to libclang on Windows or ARM platforms, use the --no-default-features command above.

Quick Start

# Create a new memory file
memvid create journal.mv2

# Add content
memvid put journal.mv2 --input document.pdf --embeddings
memvid put journal.mv2 --input notes/ --embeddings

# Search
memvid find journal.mv2 --query "meeting notes" --json

# Q&A with retrieval
memvid ask journal.mv2 --question "What was discussed in the Q3 review?"

# Timeline queries
memvid timeline journal.mv2 --since 1706745600 --limit 10

# View a specific frame
memvid view journal.mv2 --frame-id 3 --json

Commands

Features

• Single File: Everything in one portable .mv2 file
• Crash Safe: Embedded WAL ensures data durability
• Hybrid Search: Lexical (Tantivy) + Semantic (HNSW) search
• Multi-Format: PDF, DOCX, XLSX, images, audio, video
• Local LLM: Built-in Q&A with Phi-3 or llama.cpp
• Timeline: Time-based document organization
• Offline: Works completely offline after model download

Environment Variables

License

Apache-2.0

Links

• Documentation
• GitHub