Save user facts, preferences, or important details to PCG (Personal Context Graph). Use when the user explicitly tells you something you should remember. Unified PIC + KG-API + LIAM service.
Save user facts, preferences, or important details to the Personal Context Graph (PCG) for persistent memory across all homelab agents.
Nova Agent
│
└─► PIC (port 8765)
│
├─► Neo4j (graph storage)
│ └─► :Observation nodes → consolidated into :Preference nodes
│
└─► ChromaDB (vector search)
└─► Semantic search over preferences
Never save_memory without explicit user confirmation. Always ask first.
Say: "Should I remember that for future conversations?" or "Want me to save that preference?"
Write the fact in third person. Pick the most appropriate category.
Tell the user it's been saved to their personal memory.
| Category | Use For |
|---|---|
communication | Response style, address preferences, language |
work | Work habits, schedule, tools, preferences |
scheduling | Calendar preferences, meeting habits, reminders |
learning | Learning style, topics of interest, goals |
health | Health preferences, diet restrictions, exercise |
social | Social preferences, family context, relationships |
creative | Creative preferences, tools, style |
finance | Financial preferences, budgeting, priorities |
technology | Tech preferences, devices, platforms |
food | Food preferences, allergies, favorites |
family | Family details, events, relationships |
other | Anything that doesn't fit other categories |
/api/pic/learn (observation) or /api/pic/preferences (direct)skills/pic-memory/scripts/pic_memory.pynova/tools.py → handle_save_memory