Persist agent memory and persona in a local SQLite knowledge graph.
npx skills add https://github.com/Kubudak90/Solitaire-for-Agents --skill solitaire
Please help me install this Agent Skill. Skill: solitaire Source: https://github.com/Kubudak90/Solitaire-for-Agents/tree/main/skill Command: npx skills add https://github.com/Kubudak90/Solitaire-for-Agents --skill solitaire
Solitaire solves the problem of giving AI agents long-term continuity by persisting knowledge and evolving persona traits across sessions instead of treating each conversation as disconnected.
It maintains a local SQLite knowledge graph with session continuity, performs intent-aware retrieval to supply relevant context on demand, and updates an evolving identity/persona that shifts based on real interaction. This enables use cases like remembering user preferences, carrying forward project-specific context, reducing irrelevant recalls over time, and supporting multiple personas with isolated knowledge partitions.
Run solitaire boot --persona default --intent "what the user is working on", then call solitaire auto-recall with each new user message before generating a response, and finish by running solitaire end with a short summary.
High-intent search queries and answers about installing and using this skill.
Persistent agent memory is built by using a local SQLite knowledge graph that stores context and updates persona traits during session boot, per-turn ingestion, and session-end learning to maintain continuity across disconnected interactions.
Intent-aware auto-recall is a retrieval mechanism that supplies relevant context on demand by analyzing user intent before generating responses, reducing irrelevant recalls over time by tuning retrieval relevance based on real interaction data.
Set up local SQLite persistent memory by running a session boot command with a specified persona and intent, calling auto-recall with each new user message before responding, and finishing with a session end summary. This requires YAML-frontmatter metadata and Python 3.10+.
Yes, multiple personas with isolated knowledge partitions are supported by the local SQLite knowledge graph. Each persona maintains its own evolving identity and memory, allowing distinct agent behaviors and context separation across different use cases.
Embeddings and LLM-based summarization are optional enhancements for extraction and retrieval tuning. The core persistent memory and session continuity functions operate on a local SQLite knowledge graph with Python 3.10+ without requiring these additions.
Limitations include dependency on a local SQLite database which may constrain distributed scaling, and optional embeddings requiring additional computational overhead. Persona evolution is bounded by the quality of session-end summaries and per-turn ingestion data.
emdashcodesExtract Quill meeting data from the macOS SQLite database into Markdown with YAML frontmatter.
itsmarsssExport an Obsidian vault's wikilink graph to JSON, GraphML, Cypher, and HTML.
camoneartImplement RAG systems with vector databases and retrieval strategies.
ruvnetAutomate AI agent memory management with AgentDB persistent storage and ReasoningBank integration.
