DATABASE V13 QDRANT REDIS NEO4J POSTGRES BLOB NAS γ₁=14.134725141734693 · DAY 97

EOSE LABS · DAY 97 · DATA LAYER REBASELINE

DATABASE V13 BONIXER

Qdrant × Redis × Neo4j × PostgreSQL × CosmosDB × Blob × NAS · 7 Sovereign Data Types

γ₁ = 14.134725141734693

DATABASE TOPOLOGY · APP → ROUTING → DB · τ=337ms QUERY FLOW

7 DATABASE TYPES

THE 7 DATABASE TYPES IN THE FLEET

QDRANT

LIVE

BELT64 SEG 25

TypeVector DB

Locationyone:6333 (Docker restart:always)

Connecthttp://192.168.2.23:6333

pemclau-v1118,366 pts · dim=768

sessions-v1115,955 pts · dim=768

pemclau-v13-day9679 pts SEEDING

Purpose: Vector similarity search for PEMCLAU sovereign memory. Embeddings generated by nomic-embed-text via yone Ollama. 2-hop GraphRAG queries return 60–80 causally connected nodes. V13 use: primary PEMCLAU vector store, session memory retrieval, FastMCP tool endpoint at yone:9342.

REDIS

LIVE

BELT64 SEG 26

TypeCache / Pub-Sub

Locationforge:6379 + yone

Connectredis://192.168.2.12:6379

Primary keyfleet:engine

Purpose: Sub-millisecond fleet state management and pub/sub messaging. The fleet:engine Redis key holds live silo state. WPA threshold decisions are cached here before propagating to silos. forge is the fastest silo (70.7477ms τ) making it ideal for the Redis host. V13 use: WPA routing cache, silo heartbeat pub/sub, live graph state.

NEO4J

PLANNED

BELT64 SEG 27

TypeGraph DB

Locationyone or forge (planned)

Connectbolt://192.168.2.23:7687

Nodes17,283

Edges80,979

Purpose: Native graph traversal for PEMCLAU knowledge graph. Currently the graph lives in Qdrant payload fields — V13 plans Neo4j as the dedicated graph backend. 4 edge types: theorem_dependency · phase_coherence · temporal_proximity · crew_provenance. Cypher queries would replace manual 2-hop traversal. Belt64 Seg 27 reserved. Deploy with yone colocation for unified PEMCLAU stack.

POSTGRESQL

PENDING · SEG 24 RAINCHEQUE

BELT64 SEG 24

TypeRelational ACID

Locationyone:5432 (planned)

Connectpostgresql://yone:5432/eose

AltGCP Cloud SQL db-f1-micro ~$30/mo

Purpose: ACID relational storage for structured fleet data. Use cases: Lean4 theorem CRUD (pcdev joffe-math API needs persistent theorem storage), CRM contact records (kayyo@pemos.ca + ayyo@pemos.ca identity graph), actuarial tables (health + finance models), cost tracking (replace ad-hoc with proper relational). ESO will inject credentials from Azure Key Vault (L3 KMS ready). V13 raincheque — deploy Day 98.

COSMOSDB

LIVE

BELT64 SEG 28

TypeAzure NoSQL (Core SQL API)

LocationAzure Canada East

DistributionGlobal · multi-region ready

Purpose: Azure-native NoSQL for non-relational fleet metadata. NP cache (non-personalized content), cost event streaming, audit logs. Global distribution available for AKS-adjacent data that needs geo-redundancy. V13 use: cost tracking events, fleet telemetry aggregation, AKS deployment metadata.

AZURE BLOB

LIVE

BELT64 SEG 29

TypeObject Store

LocationAzure Canada East

Containersmodel-weights, docker-layers, clo-docs

ACReosefleetacrdev.azurecr.io

Purpose: Sovereign object store for binary/large data. Model weights (qwen3:32b, nomic-embed-text), Docker image layers via ACR, CLO brief documents (Harvey + Ruth + Cochran + Amani opinions), PEMCLAU old edition backups. V13 use: mefine-static container image layers, PEMCLAU edition archives, agent weight checkpoints.

NAS DISKPOOL

LIVE · 14TB

BELT64 SEG 30

TypeSovereign Cold Store

Location192.168.2.20

msi01 mount/mnt/nas-diskpool/

msclo mount/mnt/deseof/

Transfer path/mnt/nas-diskpool/eose/msclo/

Purpose: Cross-silo file transfer bus and long-term sovereign archive. NEVER use SCP or HTTP servers for cross-silo transfer — always use NAS. Contains: Ansible/IaC ground truth, PEMCLAU FC3 sealed archives, model weight backups, fleet snapshot history. V13 use: belt64 cold archive for Segs 29–30, cross-silo workspace sync.

DATA ROUTING

DATA ROUTING TABLE V13

DATA TYPE	PRIMARY DB	BELT64	STATUS	QUERY PATTERN
PEMCLAU vectors	Qdrant yone:6333	25	LIVE	cosine similarity over 768-dim nomic embeddings
Session memory	Qdrant sessions-v1	25	LIVE	115,955 pts — semantic session retrieval
Graph edges	Qdrant payload / Neo4j (planned)	27	PARTIAL	2-hop GraphRAG — theorem_dependency chains
Fleet state	Redis forge:6379	26	LIVE	GET/SET fleet:engine — sub-ms
Silo heartbeats	Redis pub/sub + /tmp/laam-silo-state/	26	LIVE	JSON files per silo + Redis channel
Lean4 theorems	joffe-math API pcdev:9384	24	PENDING POSTGRES	REST CRUD → future PostgreSQL backend
CRM contacts	PostgreSQL (planned)	24	PENDING	kayyo@pemos.ca identity graph CRUD
Cost events	CosmosDB	28	LIVE	CA$11,600/mo AKS + ~$352/day Anthropic
Docker images	ACR eosefleetacrdev	29	LIVE	docker pull eosefleetacrdev.azurecr.io/...
CLO docs	Azure Blob + NAS	29/30	LIVE	Amani + Harvey + Ruth + Cochran briefs
Secrets	Azure Key Vault	31	LIVE (ESO)	ESO ExternalSecret → k8s Secret auto-sync
IaC / Ansible	Git (NAS + GitHub)	30	LIVE	NAS = ground truth, GitHub = mirror

POSTGRESQL V13 PATH

BELT64 SEG 24 — POSTGRESQL DEPLOYMENT PLAN

Belt64 Segment 24 is the PostgreSQL reservation. V13 formalizes what needs PostgreSQL and where it should land.

USE CASE	TABLE DESIGN	PRIORITY	LANDING
Lean4 theorems	theorems(id, name, proof_text, status, timestamp)	HIGH	yone:5432/eose
CRM contacts	contacts(id, email, name, role, org, created_at)	MEDIUM	yone:5432/eose
Actuarial tables	mortality(age, gender, rate, edition, source)	MEDIUM	yone:5432/eose
Cost tracking	costs(service, amount_cad, ts, tag, cloud)	HIGH	yone or AKS
Fleet events	fleet_events(silo, event_type, ts, payload)	LOW	yone:5432/eose

DEPLOY COMMAND (Day 98):

docker run -d --name yone-postgres -e POSTGRES_DB=eose -e POSTGRES_PASSWORD_FILE=/run/secrets/pgpass -p 5432:5432 --restart always postgres:16-alpine