Services, ports and profiles

Services, ports and profiles

Every container in the CID stack: published ports, compose profile, memory limit and healthcheck.

Type: reference

One row per compose service. Regenerate with make docs-generate.

Profile empty means the service starts with a plain docker compose up -d. Anything else needs --profile <name>.
Ports are as published on the host. 127.0.0.1: means loopback only.
Memory is the declared limit (deploy.resources.limits.memory or mem_limit), not usage.

Services

Service	Ports	Profile	Memory limit	Healthcheck	Purpose
`attack-guard`	`127.0.0.1:8003:8003`	default	`4G`	yes	ONNX Jailbreak + libinjection SQLi/XSS scan
`caddy`	`80:80` `443:443` `443:443/udp`	default	—	yes	—
`cid-inline-proxy`	—	`inline-proxy`	—	no	mitmproxy forward-proxy (opt-in: `--profile inline-proxy`)
`cid-proxy`	`127.0.0.1:8080:8080`	default	`128M`	yes	Go reverse-proxy front (auth + routing)
`cid-userid-agent`	`127.0.0.1:${USERID_AGENT_PORT:-8090}:8090`	`userid-agent`	—	no	AD User-ID attribution (opt-in: `--profile userid-agent`)
`cost-analyzer`	`127.0.0.1:8016:8016`	`routing`	`1G`	yes	—
`deepteam`	`127.0.0.1:8010:8010`	default	`2G`	yes	Red-team testing
`doc-classifier`	`127.0.0.1:8017:8017`	default	`2G`	no	—
`document-parser`	`127.0.0.1:8012:8012`	default	`2G`	yes	PDF/DOCX extraction
`frontend`	`127.0.0.1:${FRONTEND_PORT:-5173}:5173`	default	—	yes	React Dashboard
`hallucination-guard`	`127.0.0.1:${HALLUCINATION_GUARD_HOST_PORT:-8022}:8022`	default	`4G`	yes	RAG hallucination (LettuceDetect EN+TR)
`hap-guard-v2`	`127.0.0.1:8004:8000`	default	`6G`	yes	ONNX Toxicity (100+ langs)
`language-detector`	`127.0.0.1:8008:8008`	default	`2G`	yes	Language detection
`llm-inference`	`127.0.0.1:8021:8000`	`llm`	—	yes	—
`mailpit`	`127.0.0.1:8025:8025`	`mailtest`	—	no	—
`mcp-server`	`127.0.0.1:${MCP_SERVER_HOST_PORT:-8020}:8020`	`llm-analyst`	`1G`	yes	Read-only MCP server over Postgres (opt-in: `--profile llm-analyst`)
`ml-detector`	`127.0.0.1:8001:8000`	default	`6G`	yes	ONNX NER PII
`nestjs-core`	`127.0.0.1:3000:3000` `${ICAP_PUBLISH:-127.0.0.1}:1344:1344`	default	—	yes	API Gateway
`ocr-service`	`127.0.0.1:8009:8009`	default	`4G`	yes	Image OCR
`openldap-test`	`127.0.0.1:389:389` `127.0.0.1:636:636`	`ldap-test`	—	no	—
`pgadmin`	`127.0.0.1:5050:80`	`tools`	—	no	—
`postgres`	`127.0.0.1:5433:5432`	default	—	yes	PostgreSQL 16
`prompt-compressor`	`127.0.0.1:8015:8015`	default	`3G`	yes	—
`reconfigure-agent`	—	default	—	no	—
`redactor`	`127.0.0.1:8013:8013`	default	`2G`	yes	Image/doc redaction
`redis`	`127.0.0.1:6380:6379`	default	—	yes	Redis Cache
`redis-commander`	`127.0.0.1:8081:8081`	`tools`	—	no	—
`report-renderer`	`127.0.0.1:8014:8014`	default	`512M`	yes	Report rendering (PDF/HTML)
`risk-analyst`	`127.0.0.1:${RISK_ANALYST_HOST_PORT:-8019}:8019`	`llm-analyst`	`3G`	yes	Local-LLM risk analyst agent loop (opt-in: `--profile llm-analyst`)

Declared memory in the default profile

19 services start without a profile; their declared limits total 37.6 GB. 6 of them declare no limit at all (caddy, frontend, nestjs-core, postgres, reconfigure-agent, redis), so the total is a floor, not a ceiling.

Startup dependencies

Service	depends_on
`attack-guard`	`redis`
`caddy`	`frontend`, `nestjs-core`, `cid-proxy`
`cid-inline-proxy`	`nestjs-core`
`cid-proxy`	`nestjs-core`, `redis`
`deepteam`	`nestjs-core`
`document-parser`	`ocr-service`
`frontend`	`nestjs-core`
`hap-guard-v2`	`redis`
`mcp-server`	`postgres`
`ml-detector`	`redis`
`nestjs-core`	`postgres`, `redis`, `ml-detector`, `hap-guard-v2`, `attack-guard`, `language-detector`, `hallucination-guard`, `ocr-service`, `redactor`, `document-parser`, `report-renderer`
`pgadmin`	`postgres`
`redis-commander`	`redis`
`risk-analyst`	`mcp-server`