Schema Design

The PostgreSQL database (zndx_gaius) uses four schemas to organize data by domain.

Public Schema

The default public schema holds core operational tables:

Content Pipeline

• feed_sources – RSS/API feed configurations with fetch intervals
• fetch_jobs – Scheduled and completed fetch job records
• content_items – Raw content items with KB path references
• articles – Curated articles with frontmatter (keywords, news queries)

Agent System

• agent_evaluations – Evaluation scores by agent and evaluator (local/xai)
• evolution_cycles – Training cycle records (success, improvement, duration)
• agent_versions – Version history for agent configurations
• held_out_queries – Reserved evaluation queries not used in training
• routing_decisions – Inference routing analytics (fallback/mismatch tracking)

Health and Observability

• health_incidents – HealthObserver incident records with FMEA scores
• healing_events – Self-healing attempt logs (tier, success, duration)
• fmea_catalog – Failure Mode and Effects Analysis registry
• scheduler_jobs – Async job queue for the inference scheduler

State

• grid_state – Persisted 19x19 grid positions and overlays
• cognition_memory – Self-observation and thought chain storage
• research_state / research_progress – Active research thread tracking

Meta Schema

The meta schema provides materialized analytics tables for Metabase:

• meta.dataset_catalog – Deduplicated dataset registry from lineage events
• meta.job_catalog – Job registry with run/success/failure counts

Populated from OpenLineage events for data provenance dashboards.

Collections Schema

The collections schema manages curated content for the public landing page:

• collections.collections – Named collections with featured flags
• collections.collection_cards – Cards assigned to collections with ordering
• collections.card_summaries – Generated card summary text

Bases Schema

The bases schema implements a feature store registry:

• bases.bases – Feature store definitions (type: feature_group, model, dataset)
• bases.base_versions – Versioned snapshots with Iceberg table references
• bases.entity_history – Entity-level change tracking

Graph Extension

Apache AGE (ag_catalog schema) provides graph query capabilities for lineage traversal using Cypher syntax. The lineage graph connects datasets to jobs via read/write edges.

Source

Full schema dump: db/schema.sql. Migrations: db/migrations/.