Llongterm: Mind as a Service for AI Apps & Agents

LLongterm provides long-term memory capabilities for AI apps and agents, enabling persistent, human‑readable memories across conversations and sessions. It is designed to be compatible with all AI chatbots and agents and offers a sandbox for experimentation. The platform emphasizes memory that is easy to inspect and reason about, with structured mind objects and memory APIs to store and retrieve information.

Overview

• Create a persistent “mind” for an AI user or agent and attach conversations, profiles, milestones, session history, and other contextual data.
• Memory operations are accessible via a simple API, allowing you to remember threads, topics, goals, and user preferences.
• The system supports human-readable data representations (JSON-like structures) to facilitate debugging and auditing.
• Includes a sandbox environment for experimentation and learning how memory affects behavior.

How it works (conceptual)

• You create a mind instance: const mind = llongterm.create().
• You store memory by passing conversational threads or structured objects to mind.remember(thread).
• The memory is retrieved and organized into a memory bag (e.g., const { memory } = await mind.remember(thread)).
• Minds can be used with any AI chatbot or agent, enabling a shared context across platforms and sessions.

Use Cases

• AI Teaching Assistants that recall student profiles, progress, and preferences.
• AI Customer Support agents that remember customer history and prior requests.
• Therapy or coaching assistants that track progress and milestones over time.
• Product managers or students who need long-term timelines, goals, and learning history.

Example: Student Profile (illustrative)

Alex Chen, Master’s CS student, visual learner, strong math/problem solving, needs writing skills improvement. Learning preferences include visual diagrams, practice problems, and interactive demos. Current modules include Advanced Algebra with a plan to cover Complex Numbers and Matrices. The memory model supports keys like name, learning_style, strengths, areas_for_improvement, learning_preferences, academic_background, course_progress, learning_history, needs_review, goals, milestones, and support needs.

How to Use LLONGTERM (typical flow)

1. Create a mind for a user or agent:

• const mind = llongterm.create()

1. Remember a conversation or profile:

• const thread = [{ author: "assistant", message: "How do you feel about apples?" }, { author: "user", message: "I finally enjoy them 🍏" }]
• const { memory } = await mind.remember(thread)

1. Retrieve and use memory in conversations:

• Use memory data to tailor responses, recall goals, remind about milestones, or surface relevant context.

Safety and Ownership

• Memory data is human-readable and structured for auditing and understanding.
• Ensure proper consent and data governance when storing personal or sensitive information.

---

Feature List

• Create and manage per-user or per-agent minds (Mind as a Service).
• Long-term memory persistence across sessions and conversations.
• Human-readable memory representations for easy inspection and debugging.
• API-based memory operations (remember, retrieve, and reason over memory).
• Compatibility with all AI chatbots and agents across platforms.
• Sandbox environment for experimentation and learning.
• Structured memory schemas (profiles, conversation threads, milestones, goals, preferences).
• Scalable memory management with clear ownership and modular data blocks.
• Easy integration with existing AI stacks and workflows.