FaultLine aims to cut AI tokens by storing and recalling memory
FaultLine is a private AI memory system that stores what a person knows and brings it back when needed. It began as a two-level memory setup, with short-term memory in Qdrant and long-term memory in PostgreSQL. It now tries to build links between ideas as new information is added, and it grows an ontology of things and relationships over time.
When a question is asked, it aims to return only the information needed instead of adding extra background. It combines a vector database with PostgreSQL so it can search by meaning and still return fuller stored records. People can confirm, correct, or remove memories.
It also has an `/extend` option that can pre-learn relationships from an AI model or from online sources. The practical goal is better RAG memory and recall with fewer tokens.
Key points
- Short-term memory uses Qdrant, while long-term memory uses PostgreSQL.
- The system creates links between ideas when new information is added.
- It grows an ontology to organize things and relationships.
- It tries to answer with only the needed memory to reduce tokens.
- Users can confirm, edit, or remove stored memories.
Quick term guide
- AI memory
- A feature where an AI tool stores user details or preferences to use in later chats.
- memory system
- A setup that lets AI save useful information and use it again later.
- long-term memory
- The capability of an AI to store and retrieve information from previous interactions over extended periods.
- PostgreSQL
- A database used to store and retrieve app data.
- background
- Running out of sight while the main app or screen stays focused on something else.
- vector database
- A special type of storage that saves text as numbers so similar meanings can be found quickly, commonly used for AI memory
- AI model
- A program that can understand prompts and produce text, code, or answers.
- AI agent
- An AI program that can inspect information and suggest what to do next.