Enterprise Modules¶

Three modules ship with the open-source package (AGPL-3.0) and are also included in commercial tiers.

TenantRouter¶

Isolates scorer instances per tenant. Each tenant gets its own CoherenceScorer with independent thresholds, knowledge bases, and caches.

from director_ai.core import TenantRouter
from director_ai import CoherenceScorer

router = TenantRouter()
router.register("acme", CoherenceScorer(threshold=0.7, use_nli=True))
router.register("beta", CoherenceScorer(threshold=0.5))

scorer = router.get("acme")
approved, score = scorer.review(prompt, response)

Policy¶

Declarative rules evaluated before scoring. Block prompts by pattern, enforce minimum thresholds per domain, tag violations.

from director_ai.core import Policy

policy = Policy(
    name="medical",
    min_threshold=0.30,
    blocked_patterns=[r"(?i)prescri(be|ption)"],
    require_nli=True,
)
violations = policy.check(prompt, response, score)

AuditLogger¶

SQLite-backed audit trail. Logs every review with prompt hash, score, approved/rejected, violations, and timestamp.

from director_ai.core import AuditLogger

logger = AuditLogger(db_path="audit.db")
logger.log(prompt, response, score, approved, violations=[])

Lazy Loading¶

Enterprise modules are lazy-loaded (v2.2.0+). import director_ai does not pull in tenant/policy/audit until you access them. No performance cost for users who don't need multi-tenancy.

import director_ai  # fast — enterprise modules not loaded

# Only loaded on first access:
router = director_ai.TenantRouter()

When to Use Commercial vs AGPL¶

Scenario	License
Internal tool, source stays private	Commercial (Indie+)
Open-source project	AGPL-3.0 (free)
SaaS with source disclosure	AGPL-3.0 (free)
SaaS without source disclosure	Commercial (Pro+)
Research / academic	AGPL-3.0 (free)
Multi-tenant production with SLA	Commercial (Enterprise)

See Licensing for pricing.