E-03 Tokenization Service
Brick E: Data, RAG, context, memory. French version: E-03-tokenization-service
1. Role and responsibilities
Reversible pseudonymization: replaces a sensitive value with a stable token, reversible only by an authorized principal.
- Tokenize (deterministic per tenant)
- Detokenize under policy and audit
- Vault encryption key in OpenBao
2. Position in the architecture and interactions
- Callers: DLP Gateway, business applications.
- 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: OpenBao (transit)
- Developed: Token vault, detokenization policy.
- External integration through an adapter
eai/integrations/<system>.pywith an in-process fallback when the URL is not configured.
4. Data model
Tables (schema eai, tenant_id RLS):
eai.token_vault
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: policy-contract
- JSON Schema:
contracts/v1/policy-contract.schema.json - Key fields:
action=detokenize, purpose - Validation:
POST /v1/contracts/{name}/validate; manifest apply:POST /v1/manifests/apply.
6. API endpoints
POST /v1/tokenization/tokenizePOST /v1/tokenization/detokenize
Request POST /v1/tokenization/tokenize:
{
"values": [
"FR7630006000011234567890189"
],
"type": "IBAN"
}
Response:
{
"tokens": [
"tok_iban_8f3a1c"
]
}
7. SDK examples
Python:
from enterprise_ai import EnterpriseAI
client = EnterpriseAI(base_url="<control-plane-url>", api_key="eai_xxx_secret")
t = client.tokenization.tokenize(values=[{"type": "IBAN", "value": "FR76..."}])
v = client.tokenization.detokenize(tokens=t.tokens, purpose="remboursement")
TypeScript:
import { EnterpriseAI } from "@enterprise-ai/sdk";
const client = new EnterpriseAI({ baseUrl: "<control-plane-url>", apiKey: process.env.EAI_API_KEY });
const t = await client.tokenization.tokenize({ values: ["FR76..."] });
Java:
EnterpriseAIClient client = EnterpriseAIClient.builder()
.baseUrl("<control-plane-url>")
.apiKey(System.getenv("EAI_API_KEY"))
.build();
Tokens t = client.tokenization().tokenize(List.of("FR76..."), "IBAN");
Go:
c, _ := eai.NewClient(eai.WithBaseURL("<control-plane-url>"), eai.WithAPIKey(os.Getenv("EAI_API_KEY")))
t, _ := c.Tokenization.Tokenize(ctx, []string{"FR76..."}, "IBAN")
8. Configuration
- Environment variables:
EAI_OPENBAO_URL,EAI_OPENBAO_TRANSIT_MOUNT,EAI_DATABASE_URL - Helm values keys:
global.openbao
9. Security controls
- Detokenization: role + purpose + OPA policy, never for the agent role by default
- 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:
eai_tokenize_total,eai_detokenize_total{decision} - Traces: OpenTelemetry spans with GenAI semantic conventions (
gen_ai.operation.name,gen_ai.agent.id,gen_ai.tool.name), attributeseai.tenant_id,eai.run_id,traceparentpropagation. - Audit actions:
tokenization.tokenize,tokenization.detokenize
11. FinOps units
- Tokenization operations.
- Emitted as Usage Event (
POST /v1/finops/usage) with tenant, project, agent and run.
12. Failure modes and resilience
- Deterministic token: retry without duplicate
- Normalized errors (error envelope, message_i18n en/fr)
13. Standalone adoption (client of a client)
- enterprise_ai.dlp + OpenBao.
- SDK:
enterprise_ai.privacy / dlp / knowledge_bases / memory / retention / lineage - Minimal deployment: Helm chart with
controlPlane.enabled=trueand PostgreSQL; disabled external components use their in-process fallback.
14. Acceptance tests
- Same value -> same token
- Detokenization by agent -> 403
- Cross-tenant access refused
- Error returned in fr with Accept-Language: fr
Governed RAG (kit L4)
The RAG sensitive cache never holds a clear value: every entry is Tink encrypted and bound to its tenant key.