Skip to main content

E-01 Data Classification

Brick E: Data, RAG, context, memory. French version: E-01-data-classification

1. Role and responsibilities​

Assigns a classification level (public, internal, confidential, restricted, secret) to texts and media using Presidio and business rules.

  • Entity detection (PII, financial, health)
  • Per tenant business rules (regex, dictionaries)
  • Return the maximum classification and findings

2. Position in the architecture and interactions​

  • Callers: DLP Gateway, RAG ingestion, Media (F), Context Manager.
  • 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: Microsoft Presidio
  • Developed: Rule engine and level mapping.
  • 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.dlp_rules

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: context-contract
  • JSON Schema: contracts/v1/context-contract.schema.json
  • Key fields: classification_max, segments[].classification
  • Validation: POST /v1/contracts/{name}/validate; manifest apply: POST /v1/manifests/apply.

6. API endpoints​

  • POST /v1/privacy/classify

Request POST /v1/privacy/classify:

{
  "text": "Client Jean Dupont, IBAN FR76 3000 6000 0112 3456 7890 189"
}

Response:

{
  "classification": "confidential",
  "findings": [
    {
      "type": "PERSON",
      "start": 7,
      "end": 18,
      "score": 0.85
    },
    {
      "type": "IBAN_CODE",
      "start": 25,
      "end": 58,
      "score": 1.0
    }
  ]
}

7. SDK examples​

Python:

from enterprise_ai import EnterpriseAI

client = EnterpriseAI(base_url="<control-plane-url>", api_key="eai_xxx_secret")
res = client.privacy.classify(text="...")

TypeScript:

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

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

Java:

EnterpriseAIClient client = EnterpriseAIClient.builder()
    .baseUrl("<control-plane-url>")
    .apiKey(System.getenv("EAI_API_KEY"))
    .build();
JsonNode res = client.privacy().classify(Map.of("text", "Contact jane.doe@example.com", "language", "en"));

Go:

c, _ := eai.NewClient(eai.WithBaseURL("<control-plane-url>"), eai.WithAPIKey(os.Getenv("EAI_API_KEY")))
res, _ := c.Privacy.Classify(ctx, map[string]any{"text": "..."})

8. Configuration​

  • Environment variables: EAI_PRESIDIO_URL, EAI_DATABASE_URL
  • Helm values keys: controlPlane.config.EAI_PRESIDIO_URL

9. Security controls​

  • Analysed text is not logged
  • 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_classify_total{level}, eai_classify_latency_seconds
  • 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:

11. FinOps units​

  • Analysis calls, processed characters.
  • Emitted as Usage Event (POST /v1/finops/usage) with tenant, project, agent and run.

12. Failure modes and resilience​

  • Presidio down: built-in regex fallback, level at least confidential
  • Normalized errors (error envelope, message_i18n en/fr)

13. Standalone adoption (client of a client)​

  • enterprise_ai.privacy + Control Plane (+ Presidio).
  • SDK: enterprise_ai.privacy / dlp / knowledge_bases / memory / retention / lineage
  • Minimal deployment: Helm chart with controlPlane.enabled=true and PostgreSQL; disabled external components use their in-process fallback.

14. Acceptance tests​

  1. IBAN detected -> confidential
  2. Neutral text -> public or internal
  3. Cross-tenant access refused
  4. Error returned in fr with Accept-Language: fr

Governed RAG (kit L4)​

restricted and secret chunks are sealed at rest (Tink envelope) and only opened when the context is built, after the ACL and clearance filters.