I build libraries for retrieval access boundaries, pre-action policy checks, reliable integration, and voice handoffs. Start with the specific boundary your application needs to enforce.
New to these projects? Read the project guide for a package comparison, installation commands, practical use cases, and how the libraries can work together.
| Need | Repository | First thing to verify |
|---|---|---|
| Keep unauthorized documents out of model context | enterprise-rag-patterns | Retrieval metadata, principal context, and actual store integration |
| Evaluate policy before invoking a tool | regulated-ai-governance | Protected execution path and required audit configuration |
| Retry integrations without duplicate business effects | integration-automation-patterns | Durable deduplication and crash/replay behavior |
| Apply access filters in Haystack | haystack-ferpa-filter | Canonical source for the ferpa-haystack distribution; metadata contracts |
| Preserve context during voice handoffs | voice-ai-governance | Redaction configuration, Redis behavior, and transfer lifecycle |
| Escalate uncertain responses or gate a proposed action | confidence-escalation | Difference between post-response evaluation and pre-action checks |
These are implementation tools and reference patterns. Regulatory labels describe intended use cases, not certification or a guarantee of compliance. Scores are heuristics unless calibrated against representative labeled outcomes. See each repository's README, tests, and changelog for supported behavior and unpublished changes; a source branch can differ from its package release.
Use an isolated environment and start from the selected repository's source instructions. Run its documented examples and tests with synthetic data before adding production integrations. A passing test covers its specific scenario; it does not establish all framework/store combinations or production adoption.
For a combined example, use the governed service demo, which connects retrieval, policy checks, escalation, durable delivery, and a voice handoff using synthetic data.
CONTRIBUTIONS.md separates independently maintained libraries from upstream contribution forks. Fork presence does not imply upstream ownership or merged contribution status; inspect individual pull requests and upstream releases.
ltngoutDemo is a historical static mockup plus a separately configured Lightning Out experiment. heroku is a historical workshop placeholder.
- Every Enterprise AI Framework Has a Compliance Gap — Here's the Architecture That Closes It (dev.to, April 2026)
- FERPA Compliance in RAG Pipelines: Five Rules Your Enterprise System Probably Breaks (dev.to, April 2026)
- Note 01 — Cross-Channel Continuity with Shared Workflow State
- Note 02 — FERPA Boundaries in Retrieval-Augmented Generation
- Note 03 — Context Assembly for Multi-Source Enterprise RAG
- Agentic Security Trends 2026: RAG Retrieval Security in the Age of Autonomous Agents
- Note 01 — Retry-Safe Event Handling and Action Logging
- Note 02 — Idempotency in Enterprise Event Processing
- Note 03 — Sync Boundaries in CRM-ERP Integration
- Note 04 — Observability and Recovery for Enterprise Integration
- Note 05 — CQRS in Enterprise Integration: When to Separate Reads from Writes
- Note 06 — Approval Workflows: When Human Decision Is a First-Class Event
- Note 07 — Backpressure and Retry Storm Prevention
- Note 08 — Schema Evolution: How to Change Event Schemas Without Breaking Consumers
- Note 09 — Process Manager Pattern for Distributed Transaction Coordination
- Note 10 — Temporal Windowing Patterns for Enterprise Event Streams
- Note 11 — Distributed Cache Patterns: When Adding a Cache Makes Things Harder
- Note 12 — API Gateway Patterns: The Six Problems Every Production Gateway Must Solve
- Agentic AI Security Trends 2026: Integration and Tool-Safety Perspective
Silos to Synergy: Practical Strategies for Enterprise AI Integration — Covers compliance-aware AI deployment, multi-system orchestration, and enterprise integration patterns for regulated environments.
- FERPA RAG Compliance Demo — live filtering demonstration on HuggingFace Spaces
Implementation, reproducible checks, and validation limits covers the governed workflow across these libraries.