DejaVu offers a private, local-first AI memory layer for agents and assistants, solving context loss between sessions and reliance on third-party cloud storage. It provides a unified memory store running entirely on the user's machine, accessible via multiple interfaces. This empowers users with portable, private, and consistently available context across AI applications, eliminating vendor lock-in and sync complexities.

How It Works

The system features three layers: an interface (Python SDK, CLI, REST, MCP), a local engine for extracting durable facts, and a SQLite database for storage. DejaVu uses Venice's API for LLM-driven memory extraction/search, with the core engine and data residing locally. This ensures all memory operations, except LLM calls, run on the user's machine, guaranteeing privacy and seamless context sharing across tools.

Quick Start & Requirements

• Installation: pip install dejavu-memory
• Prerequisites: Venice API key required (export VENICE_API_KEY="your-key").
• Initialization: Run dejavu init to set up the local memory store.
• Usage: Examples for CLI (dejavu add, dejavu search), Python SDK, and local REST API (dejavu serve).
• Storage: Memories saved locally in ~/.dejavu/.
• Links: Venice API key: venice.ai.

Highlighted Details

• Unified Memory Store: Single SQLite database (~/.dejavu/memories.db) ensures consistent context across all integrated tools.
• Multi-Interface Compatibility: Supports Python SDK, CLI, REST API, and MCP server for integration with AI agents/editors (e.g., Claude Desktop, Cursor).
• Privacy by Design: Local storage, no hosted service, telemetry off by default, LLM calls via Venice's privacy-focused API.
• Local-First Architecture: Core components run locally, promoting data ownership, offline access, and freedom from vendor lock-in.

Maintenance & Community

DejaVu is developed and maintained by JSingletonAI. The README does not specify community channels (e.g., Discord, Slack) or a public roadmap.

Licensing & Compatibility

The README does not explicitly state the software license. This omission requires clarification for commercial applications or integration into closed-source projects.

Limitations & Caveats

Presented as a "first release" shipping the "foundation," DejaVu is an initial iteration. Its LLM-dependent memory extraction/search relies on the external Venice API.