Skip to main content

D-14 API Gateway

Brick D: Tools, skills, plugins, MCP, connectors. French version: D-14-api-gateway

1. Role and responsibilities​

North-south entry point: TLS, authentication, rate limiting, routing to the Control Plane and gateways.

  • TLS and mTLS termination
  • Keycloak JWT / API key validation
  • Rate limit per tenant and key
  • traceparent and X-Request-Id propagation

2. Position in the architecture and interactions​

  • Callers: All external clients.
  • Execution path: Synchronous.
  • Every interaction goes through the Control Plane API v1 (/v1) behind the API Gateway (D-14).

3. Open source component and developed part​

  • Open source: Apache APISIX / Envoy
  • Developed: Declarative configuration and auth plugins.
  • External integration through an adapter eai/integrations/<system>.py with an in-process fallback when the URL is not configured.

4. Data model​

Tables (schema eai, tenant_id RLS):

  • eai.api_keys (lecture)

Common columns: id uuid, tenant_id uuid NOT NULL, created_at, updated_at; RLS policy current_setting('app.tenant_id', true)::uuid.

5. Contract​

  • Applied contract: no manifest; API v1 cross-cutting rules
  • Key fields: Authorization, X-API-Key, X-Project, X-Environment, Idempotency-Key
  • Validation: POST /v1/contracts/{name}/validate; manifest apply: POST /v1/manifests/apply.

6. API endpoints​

  • GET /healthz
  • GET /readyz
  • GET /v1/meta

Request GET /v1/meta:

{}

Response:

{
  "version": "1.0.0",
  "features": [
    "agents",
    "tools",
    "rag"
  ],
  "locales": [
    "en",
    "fr"
  ]
}

7. SDK examples​

Python:

from enterprise_ai import EnterpriseAI

client = EnterpriseAI(base_url="<control-plane-url>", api_key="eai_xxx_secret")
meta = client.system.meta()

TypeScript:

import { EnterpriseAI } from "@enterprise-ai/sdk";

const client = new EnterpriseAI({ baseUrl: "<control-plane-url>", apiKey: process.env.EAI_API_KEY });
const meta = await client.meta.get();

Java:

EnterpriseAIClient client = EnterpriseAIClient.builder()
    .baseUrl("<control-plane-url>")
    .apiKey(System.getenv("EAI_API_KEY"))
    .build();
JsonNode meta = client.meta().get();

Go:

c, _ := eai.NewClient(eai.WithBaseURL("<control-plane-url>"), eai.WithAPIKey(os.Getenv("EAI_API_KEY")))
meta, _ := c.Meta(ctx)

8. Configuration​

  • Environment variables: EAI_DATABASE_URL
  • Helm values keys: global.namespaces

9. Security controls​

  • WAF and size limits
  • Tenant resolved from the credential, never from the body
  • Every request carries the tenant resolved from the credential; PostgreSQL RLS
  • Every mutation and policy decision writes a hash-chained audit event

10. Observability​

  • Prometheus metrics: apisix_http_status, envoy_http_downstream_rq_total
  • Traces: OpenTelemetry spans with GenAI semantic conventions (gen_ai.operation.name, gen_ai.agent.id, gen_ai.tool.name), attributes eai.tenant_id, eai.run_id, traceparent propagation.
  • Audit actions: no dedicated action; APISIX rejections (authentication, rate limit) are logged by APISIX and accepted calls produce the actions of the called domain, for example model.invoke.

11. FinOps units​

  • API requests.
  • Emitted as Usage Event (POST /v1/finops/usage) with tenant, project, agent and run.

12. Failure modes and resilience​

  • Active health checks; retry on 503 for GET only
  • Normalized errors (error envelope, message_i18n en/fr)

13. Standalone adoption (client of a client)​

  • Can front any brick; Helm apiGateway.enabled.
  • SDK: enterprise_ai.tools / mcp / skills / plugins / connectors / channels
  • Minimal deployment: Helm chart with controlPlane.enabled=true and PostgreSQL; disabled external components use their in-process fallback.

14. Acceptance tests​

  1. Without credential -> 401
  2. Rate limit exceeded -> 429
  3. Cross-tenant access refused
  4. Error returned in fr with Accept-Language: fr

Kit L3 proof​

Offline validation: tests/test_apisix_config_schema.py checks deploy/apisix/apisix.yaml against the APISIX 3.x object and plugin schemas (limit-count, ip-restriction, request-id, openid-connect, limit-req) and the ingress policy (authentication, rate limit, request id on every /v1 route, IP restriction on health routes). Starting the APISIX runtime needs root or Docker, unavailable on the test host. Egress: tests/test_security_egress.py, tests/test_security_egress_wiring.py.