Visual reference for the three-layer Jarvis-like memory system. Redis buffer, Markdown logs, and Qdrant vector database working together.
Session JSONL → Live Context
Lost on /reset or session expiration • ~8k-32k tokens
Key: mem:rob
List data structure • LPUSH append
Flush: Daily 3:00 AM
cron_backup.py → Qdrant
Location: memory/YYYY-MM-DD.md
Human-readable Markdown
Git-tracked
Never lost • Always accessible
Embeddings
snowflake-arctic-embed2
1024 dimensions
Collections
kimi_memories
kimi_kb
private_court_docs
User-Centric
user_id: "rob"
Cross-session search
Every message triggers automatic append to Redis and file log.
Periodic checks accumulate new turns since last save.
Nightly cron jobs flush buffers to permanent storage.
10.0.0.10:11434
Embeddings
10.0.0.40:6333
Vector DB
10.0.0.36:6379
Task Queue
10.0.0.8:8888
Search
| Command | Action | Layer |
|---|---|---|
| save mem | Save all context | Redis + File |
| save q | Store to Qdrant | Vector DB |
| q topic | Semantic search | Vector DB |
| remember | Quick note | File |
| Schedule | Script | Action |
|---|---|---|
| Heartbeat | hb_append.py | New turns → Redis |
| 3:00 AM | cron_backup.py | Redis → Qdrant |
| 3:30 AM | sliding_backup.sh | File → Backup |
Download the full Jarvis Memory blueprint with all scripts, documentation, and installation guide.
Go to Downloads