Nram FAQs

Question 1

What is Nram and what problem does it solve for AI agents?

Accepted Answer

Nram is a self-hosted persistent memory layer for AI agents. It solves the challenge of limited context windows by enabling agents to store, recall, and enrich vast amounts of long-term information, improving coherence and recall across conversations.

Question 2

Is Nram designed for multi-user or multi-agent environments?

Accepted Answer

Absolutely. Nram features multi-tenancy with organizations and projects, Role-Based Access Control (RBAC) with five distinct roles, and robust OAuth 2.0 for secure user and client management, making it ideal for complex, collaborative AI systems.

Question 3

How does Nram facilitate intelligent recall of memories?

Accepted Answer

Nram utilizes semantic search through vector embeddings, supporting pgvector (PostgreSQL), HNSW (SQLite), or Qdrant. This allows AI agents to recall context based on meaning and relevance, rather than just keywords.

Question 4

What is the role of the Knowledge Graph and 'Dreaming' in Nram?

Accepted Answer

Nram automatically builds a knowledge graph from extracted facts and entities, enabling multi-hop traversal for deeper context. The 'Dreaming' process is an offline mechanism that refines this graph by consolidating entities, resolving contradictions, deduping, and inferring new relationships.

Question 5

Does Nram support different LLM providers and database backends?

Accepted Answer

Yes, Nram is LLM provider agnostic, supporting OpenAI, Anthropic, Google Gemini, Ollama, OpenRouter, and any OpenAI-compatible endpoint. For databases, it offers dual support for SQLite (default) and PostgreSQL, both fully supporting enrichment and knowledge graph features.

Nram

Nram

主要功能

使用案例

主要功能

使用案例