Data Pipelines
Offline
LB-102schizophrenia

Ingestion flow

Every source runs the same template

fetchnormalizeresolve entity_idPostgresMongoNeo4jembed → Mongo vectorsSSE + metrics

A canonical entity_id links the same record across Postgres, Mongo and Neo4j. After each run an event is published to Redis → SSE (live UI refresh) and any subscribed living metric recomputes.

PostgreSQL

identity · structured · vectors

MongoDB

document lake

Neo4j

graph · 0 edges

Full-text scraper

scrape → key highlights → embed, across news articles

Pipelines

Ingestion sources

Scheduled jobs

Celery Beat

JobSchedule (cron)

Agentic workflow runs

DAG executions