Dependency Map

This page documents the complete dependency graph for all 24 MATIH Platform services, including both service-to-service dependencies and infrastructure dependencies. Understanding these dependencies is critical for failure analysis, capacity planning, and deployment ordering.

Control Plane Dependencies

Service	Depends On	Protocol
`api-gateway`	`iam-service`, `config-service`	REST
`tenant-service`	`iam-service`, `infrastructure-service`	REST
`tenant-service`	`notification-service`, `audit-service`, `billing-service`	Kafka
`billing-service`	`notification-service`	Kafka
`config-service`	All services (broadcast)	Redis Pub/Sub
`observability-api`	Prometheus, Elasticsearch	REST
`audit-service`	-- (terminal consumer)	--

Data Plane Dependencies

Service	Depends On	Protocol
`ai-service`	`query-engine`, `semantic-layer`, `catalog-service`, `ontology-service`	REST
`bi-service`	`semantic-layer`, `query-engine`, `render-service`	REST
`query-engine`	Trino	JDBC
`semantic-layer`	Trino	JDBC
`pipeline-service`	`query-engine`, Temporal	REST, gRPC
`data-quality-service`	`query-engine`, `catalog-service`	REST
`governance-service`	`catalog-service`, Polaris	REST
`ops-agent-service`	Prometheus, Loki, `ai-service`	REST
`data-plane-agent`	All local services, Control Plane	REST, Kafka

Infrastructure Dependencies

Service Category	PostgreSQL	Redis	Kafka	Elasticsearch
Control Plane (10)	All	Most	5	2
Data Plane Java (7)	All	Most	3	0
Data Plane Python (6)	All	Most	2	0
Data Plane Node.js (1)	0	0	0	0

Failure Impact Analysis

Failed Component	Direct Impact	Cascade Risk
`iam-service`	No authentication	All authenticated operations fail
`api-gateway`	No external access	Total outage for external users
`config-service`	No config updates	Low (cached configs still work)
`query-engine`	No SQL execution	AI, BI, data quality all degraded
`ai-service`	No conversational analytics	Chat features unavailable
`semantic-layer`	No semantic queries	BI and AI lose metric computation
`catalog-service`	No metadata resolution	AI loses schema context
PostgreSQL	No persistent data access	Service-specific failures
Redis	Cache miss, no pub/sub	Performance degradation
Kafka	No async events	Audit, billing, notifications delayed
Trino	No federated queries	Query engine cannot execute SQL

Deployment Order

Services must be deployed in dependency order:

1. Infrastructure: PostgreSQL, Redis, Kafka, Elasticsearch
2. Core: iam-service, config-service
3. Control Plane: tenant-service, api-gateway, audit, billing, notification
4. Data Infrastructure: Trino, MinIO, Qdrant
5. Data Plane: catalog, semantic-layer, query-engine
6. Data Plane: ai-service, bi-service, ml-service, pipeline-service
7. Supporting: render-service, data-quality, governance, ontology, ops-agent
8. Agent: data-plane-agent

Scaling Relationships

When This Scales	Also Scale	Reason
`ai-service` +1	`query-engine` +1	AI generates queries for execution
`bi-service` +1	`render-service` +1	More dashboards need rendering
New tenant	+1 `data-plane-agent`	Each namespace runs its own agent
Kafka throughput	All consumers	Consumer groups must keep pace

Service Registry -- Service discovery and DNS
Communication Patterns -- Protocol selection
Data Flow -- Request tracing through dependencies

Communication Patterns Multi-Tenancy Architecture