Oubli offers a persistent fractal memory for AI, ensuring that vital information is never lost. Quickly bootstrap your AI's memory, seamlessly export and import knowledge, and visualize your mind with powerful synthesis tools. Transform raw data into actionable insights for richer interactions.
Oubli: A Persistent Memory System for AI
Oubli, which translates from French to "forgetting," offers a groundbreaking memory system designed for AI applications. Drawing inspiration from the sweet West African fruit that symbolizes the act of forgetting, Oubli ensures that vital information is never lost, supporting the Claude Code interface with a unique fractal memory framework.
Key Features
- Fractal Memory Hierarchy: Enables both upward synthesis of raw memories into insights and downward drill-down access to detailed source memories.
- Hybrid Search Functionality: Utilizes a combination of BM25 keyword search and semantic embeddings for intelligent memory retrieval, allowing for contextual understanding beyond exact matches.
- Core Memory Management: Maintains essential facts (~2K tokens) that define personal identity, updated in real-time during interactions with Claude.
- Proactive Memory Updates: Automatically saves relevant information without requiring prompts, ensuring memories stay current.
- Instantaneous Updates: Changes to family, work, or identity are reflected immediately in Core Memory.
- Background Operation: Memory management occurs silently, providing a seamless user experience.
- Local Storage: All data is stored locally, ensuring privacy and control over personal information.
- Interactive Visualization: Create visual representations of memory hierarchies using the command
/visualize-memory.
Installation Overview
To set up Oubli, execute the following commands in your project directory:
pip install oubli
cd /your/project\noubli setup
This establishes a local installation that maintains isolated memories specific to each project.
How Oubli Works
Oubli utilizes a fractal memory system structured in layers:
- Core Memory: Contains essential identity tokens, always available in context.
- Synthesis Levels: Higher-level insights consolidate themes from raw memories, categorized to enable efficient retrieval.
- Raw Memories: Full text from past interactions can be accessed to retrieve specific details or quotes as needed.
Usage
Oubli facilitates natural interaction, automatically saving memories derived from casual conversations. Commands include:
/synthesize: Merges duplicate memories and updates Core Memory./clear-memories: Clears all entries (confirmation required).oubli viz: Opens an interactive graph of memories in the browser.
In summary, Oubli presents an innovative approach to memory management, enabling AI to maintain a personalized, evolving understanding of users while organizing information efficiently for enhanced interactions.
No comments yet.
Sign in to be the first to comment.