Skip to content

Security: invarcore/knowledge-fabric

Security

SECURITY.md

Security Policy

Supported Versions

We release patches and security fixes for the active minor release branch.

Version Supported
0.1.x ✅

Reporting a Vulnerability

Security is paramount to Knowledge Fabric, especially given its role in multi-tenant data isolation and enterprise evidence retrieval.

If you believe you have discovered a security vulnerability in Knowledge Fabric:

  1. Do not open a public GitHub issue.
  2. Please submit a confidential report via GitHub Private Vulnerability Reporting or email the maintainers at security@knowledge-fabric.dev.
  3. Provide a clear description, proof of concept, and reproduction steps.

We will acknowledge receipt within 48 hours and work with you on a coordinated disclosure timeline.


Deployment & Security Configuration Flags

Knowledge Fabric provides configurable security guardrails that adapt to different enterprise workloads without disrupting high-throughput pipelines or large document processing:

Environment Variable Default Purpose & Flexibility
KNOWLEDGE_MAX_QUERY_LENGTH 4000 Maximum length in characters for incoming query strings. Can be increased for deep analytical queries or error stack traces.
KNOWLEDGE_MAX_TOP_K 100 Maximum top_k results returned by lexical/vector search. Prevents unbounded payload sizes and out-of-memory errors.
KNOWLEDGE_MAX_RERANK_CANDIDATES 100 Maximum candidate pool evaluated by cross-encoder rerankers before returning top hits. Protects CPU/GPU from compute spikes.
KNOWLEDGE_STRICT_TENANT_CHECK true Enforces safe tenant namespace syntax (^[a-zA-Z0-9_.:@-]{1,128}$). Set to false if custom internal naming schemes are required.
KNOWLEDGE_ENFORCE_NON_SUPERUSER true Restricts database connection roles to dedicated least-privilege non-superusers (e.g. kf_app) to enforce PostgreSQL Row-Level Security (KF-01). Set to false to permit superusers for migrations/admin tooling.

There aren't any published security advisories