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 See each repository's README, tests, and changelog for supported behavior and unpublished changes; a source branch can differ from its package release.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.
- The Agentic Era Is Here: 7 AI Trends Every Developer Must Know in 2026 (dev.to, May 2026)
- Three Security Issues Specific to Multi-Agent AI Systems (OWASP Agentic AI Top 10) (dev.to, May 2026)
- EU AI Act Goes Live in 90 Days: What Developers Building AI Agents Actually Need to Do (dev.to, May 2026)
- MACF: The 6-Component Framework Every Enterprise Multi-Agent AI System Needs (dev.to, April 2026)
- Enterprise AI Governance: Put Policy Checks Before Tool and Data Access (dev.to, April 2026; updated November 2026)
- Google ADK Has a Compliance Gap — Here's How to Close It (dev.to, April 2026)
- FERPA-Aware RAG Pipelines: Five Controls to Review Before Deployment (dev.to, April 2026; updated November 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.
-
AI Made Simple: The Everyday Guide to Using AI Assistants to Save Time, Work Smarter and Live Better (co-author, October 2026)
-
AI for Electric Vehicle Batteries: A Hands-On Python Guide to State of Charge, State of Health, Remaining Life and Smart Charging (co-author, October 2026)
-
TinyML and Edge AI with ESP32 and Raspberry Pi: A Hands-On Guide to Building Intelligent, Offline AI Devices (co-author, September 2026)
-
AI and Deep Learning with Arduino and Raspberry Pi: A Hands-On Guide to Smart IoT Projects (co-author, September 2026)
- FERPA RAG Compliance Demo — live filtering demonstration on HuggingFace Spaces
Implementation, reproducible checks, and validation limits covers the governed workflow across these libraries.