Smriti FAQs

Question 1

What is Smriti and its core purpose?

Accepted Answer

Smriti is a neuro-inspired long-term memory architecture for AI agents. Its core purpose is to provide persistent, scalable memory by decoupling fast real-time ingestion from slower, analytical background consolidation, ensuring non-blocking interactions.

Question 2

How does Smriti's Dual-Process architecture work?

Accepted Answer

Inspired by human cognition, Smriti separates memory operations into System 1 (fast, real-time ingestion into a Working Memory/Episode Buffer) and System 2 (slow, background consolidation into a Semantic Palace for chunking, conflict resolution, and skill extraction).

Question 3

Which Large Language Models (LLMs) does Smriti support?

Accepted Answer

Smriti is provider-agnostic and offers native integration with a wide range of LLMs, including local Ollama models (e.g., Mistral), OpenAI (gpt-*), Anthropic (claude-*), and Google Gemini models.

Question 4

Can I visualize or export my AI agent's memory using Smriti?

Accepted Answer

Yes, Smriti includes an interactive, zero-dependency Memory Browser UI to visualize the Semantic Palace graph. It also supports exporting your agent's memory to an Obsidian vault, allowing its graph view to mirror your memory structure.

Question 5

What are the main benefits of using Smriti for AI agents?

Accepted Answer

Smriti provides scalable, persistent, and non-blocking long-term memory, enabling complex reasoning without hindering real-time interactions. It enhances memory organization through asynchronous consolidation and offers flexible LLM and agent framework integrations like LangChain and Claude Code.

Smriti

Smriti

主要功能

使用案例

主要功能

使用案例