Skip to content

Add Unity AI Gateway governance demo - #19

Open
dmatrix wants to merge 2 commits into
databricks-solutions:mainfrom
dmatrix:add-unity-ai-gateway-governance-demo
Open

Add Unity AI Gateway governance demo#19
dmatrix wants to merge 2 commits into
databricks-solutions:mainfrom
dmatrix:add-unity-ai-gateway-governance-demo

Conversation

@dmatrix

@dmatrix dmatrix commented Sep 3, 2026

Copy link
Copy Markdown
Contributor

What

Adds a self-contained demo — demos/unity_ai_gateway_governance/ — showing how to govern coding-agent sprawl (Cursor, Claude Code, Codex CLI, Gemini CLI, Pi) through Unity AI Gateway. Each provider is routed to its own governed model service (catalog.schema.service) carrying guardrails, an inference-table audit trail, usage tracking, and rate limits, all reached through one gateway URL.

The notebook walks through:

  • Act 1 — verify each governed model service and its deployed config
  • Act 2 — simulate the coding-agent swarm (happy path)
  • Act 3 — guardrails in action (PII, jailbreak, unsafe content)
  • Acts 4–5 — audit trail and usage tracking via inference tables + system.ai_gateway.usage
  • Act 6 — per-service QPM/TPM rate limiting
  • Act 7–8 — MLflow tracing and a summary dashboard

Changes

  • New demo project under demos/unity_ai_gateway_governance/ (notebook + supporting modules, images, DAB config, env template)
  • Link the demo from the top-level README.md
  • Default Act 2 happy-path volume to 1 clean request per agent, interleaved and paced, to stay under shared foundation-model backend and guardrail-judge per-minute limits (configurable via get_clean_scenarios(per_agent=...))
  • Add @dmatrix to CODEOWNERS.txt

Secrets are excluded via .gitignore (.env, .databricks/, mlartifacts/); env-template ships with placeholders only.

This pull request and its description were written by Isaac.

dmatrix and others added 2 commits September 3, 2026 15:11
Add a self-contained demo showing how to govern coding-agent sprawl
(Cursor, Claude Code, Codex CLI, Gemini CLI, Pi) through Unity AI Gateway:
per-provider governed model services with guardrails, inference-table audit,
usage tracking, rate limiting, and MLflow tracing.

- Link the demo from the top-level README
- Default Act 2 happy-path volume to 1 clean request per agent, paced, to stay
  under shared backend and guardrail-judge per-minute limits
- Add @dmatrix to CODEOWNERS

Co-authored-by: Isaac <no-reply@databricks.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant