Store and recall agent memories with FAISS, SQLite FTS5, and a knowledge graph.
npx skills add https://github.com/kyssta-exe/Ariadne --skill ariadne-kyssta-exe
Please help me install this Agent Skill. Skill: ariadne Source: https://github.com/kyssta-exe/Ariadne/tree/main/skills/ariadne Command: npx skills add https://github.com/kyssta-exe/Ariadne --skill ariadne-kyssta-exe
Ariadne provides a local-first memory system for AI agents, delivering fast, offline memory with a hybrid vector search, a rich knowledge graph, and cognitive retention without cloud dependencies.
Use Case: An AI agent can remember user preferences, recall past conversations, traverse relationships to infer new insights, and maintain privacy by staying local.
Install ariadne-memory and configure Hermes to use Ariadne as memory provider.
High-intent search queries and answers about installing and using this skill.
You can use a local-first memory system to store and recall agent memories offline. Ariadne provides this via a self-contained SQLite database with FAISS vector search and a typed knowledge graph, ensuring cognitive retention without cloud connectivity.
Hybrid search combines FAISS vector search with SQLite FTS5 keyword search and a typed knowledge graph. This approach enables both semantic similarity matching and exact keyword retrieval, supporting multi-hop reasoning and fast recall for local agents.
A forgetting curve manages cognitive retention by gradually deprioritizing stale memories while deduplication prevents redundant storage. Ariadne applies this locally within its SQLite database to maintain efficient, relevant memory recall for offline agents.
Yes, you can configure Hermes to use Ariadne as a local memory provider. It integrates as a self-contained plugin, utilizing a single SQLite database for memories, embeddings, and graph data to support offline agent workflows.
FAISS vector search and SQLite FTS5 keyword search work together within a local memory system to enable multi-hop reasoning. Ariadne combines these mechanisms alongside a typed knowledge graph to traverse relationships and infer new insights.
Local-first memory systems operate without cloud dependencies, which can complicate real-time cross-agent sharing. However, Ariadne provides a cross-agent surface within its local SQLite database, allowing optional embeddings and sharing workflows offline.
ToqsickImplement a local memory backend for Hermes Agent using fastembed and sqlite-vec.
whiteboardevClassify user queries to retrieve memories via recent, keyword, or semantic search.
ShadylukinDesign Neumann data models combining relational tables, graph nodes, and vector embeddings.
prince3626ezechiel-langAssess client environments and configure local-first private memory layers for offline AI deployment.
