Skip to content

docs(governance): rewrite Token budgets and limits page for ADP GA#6

Merged
micheleRP merged 1 commit intomainfrom
adp-budgets-rewrite
Apr 29, 2026
Merged

docs(governance): rewrite Token budgets and limits page for ADP GA#6
micheleRP merged 1 commit intomainfrom
adp-budgets-rewrite

Conversation

@micheleRP
Copy link
Copy Markdown
Contributor

@micheleRP micheleRP commented Apr 28, 2026

Summary

Replaces the 3-line governance/pages/budgets.adoc placeholder with a 5-section concept page aligned to the Governance V0 PRD's V2 framing for cap-management. Implements workflow #6 (DOC-2114) per the Token Budgets and Limits companion plan.

Preview pages

What this PR does

  • Replaces the 3-line placeholder at modules/governance/pages/budgets.adoc with a 5-section concept page covering:
    • What ADP records automatically — spending event fields (input / output / cached tokens, total_cost_microcents, request count, provider / model / user / org context). Microcent unit explainer (1 cent = 100 microcents, $1 = 10,000 microcents) lifted from RFC 0010 + the SpendingService proto.
    • Where to view your spend — Governance dashboard (xref dashboard/index.adoc), Transcripts (xref observability/transcripts.adoc), Breakdown queries via GetSpendingBreakdown. SpendingFilter dimension combinations.
    • Guardrail evaluator cost — same SpendingService stream as user-facing LLM calls; separate provider attribution; cross-link to Guardrails.
    • Multi-tenant patterns at GA — viewing only — per-org / per-user breakdowns via SpendingFilter.organization_id and user_id. Explicitly notes "see, not cap."
    • Coming in a later release — 5 bullets covering configurable caps, halt-vs-notify, per-agent caps, alert hooks, multi-tenant cap-setting. Per the Governance V0 PRD V2 framing.
  • Topic type: :page-topic-type: overview. Personas: platform_admin, evaluator (with // TODO: flagging the persona-vocabulary mismatch with PRD names — Open Q D2).
  • 3 learning objectives with observable verbs (identify, locate, recognize) — covering awareness/comprehension levels appropriate for an overview page.
  • No nav restructure — flat governance:budgets.adoc entry stays under Trust & Governance per the master plan §2 nav. Sub-folder split layers post-GA when V2 cap-management lands.

Status

  • Pre-beta (May 15): docs commit to GA only — this page is best-effort if it lands by then, not a beta deliverable.
  • GA (2026-06-15): the page covers the read-only spending visibility surface. Cap-management features (per-agent caps, halt-vs-notify, alerts, multi-tenant caps) are V2 / post-GA per the Governance V0 PRD.
  • Plan: https://redpandadata.atlassian.net/wiki/spaces/DOC/pages/1884717057

Open questions / known issues

  • Verdict shifted U → C vs. master plan: the master-plan source page (ai-agents/ai-gateway/admin/setup-guide.adoc) was deleted in cloud-docs#562 (merged 2026-04-28); current adp-docs file is the 3-line placeholder being rewritten. Effective verdict is greenfield create. Reflected in master plan v20.
  • No source spend_limit.proto or ratelimit.proto on cloudv2 main. AI Gateway RFC 0002 Phase 4 (Virtual Keys + budgets) is "Not started"; the partial work that landed was removed in a Kafka-kvstore migration. So at GA there is no user-facing cap-setting feature; the page's "Coming in a later release" section sketches V2 capabilities per the PRD without committing to specifics.
  • 2 transient broken xrefs to governance:dashboard/index.adoc from this PR. The Dashboard pages live in adp-docs PR #5 (also draft); the xrefs resolve once PR docs(governance): add Governance Dashboard pages for ADP GA #5 merges.
  • TODO markers trace to 5 of the 9 Open Questions in the companion plan: persona vocabulary (D2), default-on capture (A1), user_id / organization_id population (A2), guardrail-cost-stream confirmation (A3), public-API exposure of organization_id (B1), and the V2 forward-link section (C1).
  • Master plan §3 docs(governance): rewrite Token budgets and limits page for ADP GA #6 realigned to PRD V2 framing in v20 (2026-04-28): original pre-beta items (per-agent caps, halt vs. notify, alerts, multi-tenant cap-setting) deferred to post-GA. The page reflects this alignment.

Test plan

  • Local build (npm run build && npm run serve) renders the Trust & Governance > Token budgets and limits page; cross-links to Transcripts and Guardrails resolve. (Cross-link to Dashboard resolves once PR docs(governance): add Governance Dashboard pages for ADP GA #5 merges.)
  • On adp-production once spending data flows: walk GetSpendingBreakdown against each dimension (provider / model / user / organization / provider type) and confirm prose matches actual response shapes.
  • Cold-reader walkthrough: a platform admin can answer "where do I see my spend?" "can I set per-agent caps right now?" "how is guardrail cost tracked?" cleanly.
  • Resolve // TODO: markers as eng confirms (default-on, OIDC claim mapping, guardrail-cost stream, public-API organization_id exposure).
  • Resolve persona vocabulary alignment with docs-team-standards.
  • Editorial pass via docs-team-standards:review on the page.
  • Remove adp-la.adoc partial include at GA.

🤖 Generated with Claude Code

Replaces the 3-line placeholder at modules/governance/pages/budgets.adoc
with a 5-section concept page aligned to the Governance V0 PRD V2 framing
for cap-management.

Page structure:

* Intro (overview pattern, ~70 words) — what ADP records automatically
  and what cap-management arrives in a later release.
* What ADP records automatically — spending event fields (input/output/
  cached tokens, total_cost_microcents, request count, provider/model/
  user/org context); microcent unit explainer (1 cent = 100 microcents,
  $1 = 10,000 microcents).
* Where to view your spend — Governance dashboard, Transcripts,
  Breakdown queries via GetSpendingBreakdown; SpendingFilter dimension
  combinations.
* Guardrail evaluator cost — same SpendingService stream as user-facing
  LLM calls, separate provider attribution.
* Multi-tenant patterns at GA — viewing only (not cap-setting); per-org
  per-user breakdowns via SpendingFilter dimensions.
* Coming in a later release — configurable caps, halt-vs-notify,
  per-agent caps, alert hooks, multi-tenant cap-setting.

Verdict shifted from U to C: master plan source page
ai-agents/ai-gateway/admin/setup-guide.adoc was deleted in cloud-docs#562;
current adp-docs file is a placeholder being rewritten.

Master plan §3 #6 was realigned to PRD V2 framing in master plan v20:
original pre-beta items (per-agent caps, halt vs. notify, alerts,
multi-tenant cap-setting) deferred to post-GA. No source spend_limit.proto
or ratelimit.proto on cloudv2 main; AI Gateway RFC 0002 Phase 4 (Virtual
Keys + budgets) is Not started; partial work was removed in Kafka-kvstore
migration. SpendingService (read-only) is the GA surface.

TODO markers track 5 of the 9 open design questions in the companion
plan: persona vocabulary, default-on capture behavior, user_id/org_id
population, guardrail-cost-stream confirmation, public-API exposure of
organization_id, V2 surface forward link.

Cross-links to xref:governance:dashboard/index.adoc point at the planned
post-PR-#5 path; those resolve once Governance Dashboard PR #5 merges.

No nav restructure — flat governance:budgets.adoc entry stays under
Trust & Governance per master plan §2. Sub-folder split layered post-GA
when V2 cap-management lands.

Refs: DOC-2114
RFC: apps/aigw/docs/rfcs/0010-cost-reporting.md (cloudv2 main)
Proto: proto/public/cloud/redpanda/api/adp/v1alpha1/spending_service.proto (cloudv2 main)
Plan: https://redpandadata.atlassian.net/wiki/spaces/DOC/pages/1884717057

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@netlify
Copy link
Copy Markdown

netlify Bot commented Apr 28, 2026

Deploy Preview for redpanda-agentic-data-plane ready!

Name Link
🔨 Latest commit 3acf7b2
🔍 Latest deploy log https://app.netlify.com/projects/redpanda-agentic-data-plane/deploys/69f139808222100008722ed4
😎 Deploy Preview https://deploy-preview-6--redpanda-agentic-data-plane.netlify.app
📱 Preview on mobile
Toggle QR Code...

QR Code

Use your smartphone camera to open QR code link.

To edit notification comments on pull requests, go to your Netlify project configuration.

@micheleRP micheleRP marked this pull request as ready for review April 29, 2026 01:18
@micheleRP micheleRP requested a review from a team as a code owner April 29, 2026 01:18
@micheleRP micheleRP merged commit c7cfc92 into main Apr 29, 2026
4 checks passed
@micheleRP micheleRP deleted the adp-budgets-rewrite branch April 29, 2026 14:45
micheleRP added a commit that referenced this pull request Apr 29, 2026
Beta deliverable (master Week 3, May 4-8): single concept page
covering INPUT/OUTPUT phase model, the three evaluator types
(PII / Toxicity / Custom webhook), violation framing, and
attachment scope. Anchors on RFC 0002 Phase 5; resource-model
details left as TODO markers pending team-ai's post-pivot answer.

GA deliverables (master Weeks 4-6): four pages stubbed with
full outlines and TODO markers keyed to specific Open Questions
in the Confluence companion plan, so live walkthroughs on
adp-production can resolve them in place:

- create-guardrail.adoc — how-to for configuring + attaching
- types-reference.adoc — config schema reference per evaluator
- violations.adoc — read & troubleshoot violations
- cost-tracking.adoc — per-evaluator cost shape + capping

Nav: replaces the flat governance:guardrails.adoc placeholder
under Trust & Governance with a Guardrails parent + 5 children.
The 3-line governance/pages/guardrails.adoc placeholder is
deleted in this commit.

Cross-link to governance:dashboard/index.adoc is transient-broken
until adp-docs PR #5 (Governance Dashboard) merges; same pattern
as Budgets PR #6.

Refs: DOC-2113. Plan:
https://redpandadata.atlassian.net/wiki/spaces/DOC/pages/1881702438

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

# Conflicts:
#	modules/ROOT/nav.adoc
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