[SLO] Add AI-powered auto-discover SLOs feature

[drewpost]

Summary

Adds a new Auto-discover creation mode to the SLO creation page that scans the user's cluster data and uses AI to propose user-centric SLOs following SRE best practices.

• Cluster data discovery: Scans APM services, synthetics monitors, log streams, and metric indices to build a summary of available data sources
• AI-powered SLO proposals: Sends the cluster summary to an LLM via the inference plugin with a specialized prompt that prioritizes user-facing reliability (availability > latency > uptime > error logs)
• Bulk SLO creation: One-click creation of all selected SLOs with per-SLO success/failure reporting

New API endpoints

Endpoint	Description
POST /internal/slo/ai/discover	Scans cluster data and proposes SLOs via AI
POST /internal/slo/ai/bulk-create	Creates multiple SLOs at once
POST /internal/slo/ai/generate	Converts NL prompt to structured SLO definition
POST /internal/slo/ai/suggest	Suggests improvements for an SLO definition

User flow

1. Navigate to Create new SLO (/app/slos/create)
2. Click the Auto-discover toggle (alongside Manual and AI-assisted)
3. Click Scan cluster & discover SLOs
4. AI analyzes cluster data and proposes SLOs (Critical/High priority pre-selected)
5. Review proposals — each card shows indicator type, target, time window, service, priority badge, category badge, and AI rationale
6. Toggle selections as needed (select all / deselect all available)
7. Click Create N SLOs to bulk-create all selected
8. Redirected to SLO list with success notification

Files changed

18 files changed (~2,970 lines added)

Server (6 new files):

• server/services/ai/slo_generation_prompt.ts — System prompts and output schemas for generate, suggest, and discover
• server/services/ai/get_default_connector.ts — Default AI connector resolution (UI settings → inference plugin EIS fallback)
• server/routes/slo/ai_discover_slos.ts — Discovery route with cluster scanning + LLM proposals
• server/routes/slo/ai_bulk_create_slos.ts — Bulk create route using createSLOParamsSchema
• server/routes/slo/ai_generate_slo.ts — NL generation route
• server/routes/slo/ai_suggest_slo.ts — Suggestion route

Client (10 new files, 2 modified):

• public/hooks/use_discover_slos.ts — React Query mutation for discovery
• public/hooks/use_bulk_create_slos.ts — React Query mutation for bulk create
• public/hooks/use_generate_slo.ts — React Query mutation for NL generation
• public/hooks/use_suggest_slo.ts — React Query mutation for suggestions
• public/pages/slo_edit/components/nl_slo/creation_mode_toggle.tsx — Three-way toggle
• public/pages/slo_edit/components/nl_slo/slo_discover_form.tsx — Main auto-discover UI
• public/pages/slo_edit/components/nl_slo/discovered_slo_card.tsx — Individual proposal card
• public/pages/slo_edit/components/nl_slo/nl_slo_form.tsx — NL generation form
• public/pages/slo_edit/components/nl_slo/slo_preview.tsx — SLO definition preview
• public/pages/slo_edit/components/nl_slo/slo_suggestions.tsx — AI suggestions display
• public/pages/slo_edit/slo_edit.tsx — Integrated all creation modes
• server/routes/slo/route.ts — Registered all AI routes

Dependencies

• Requires the inference plugin (optional dependency)
• Uses the default AI connector (EIS preferred, falls back to OpenAI/first available)

Test plan

• Navigate to /app/slos/create and verify the three-way toggle appears
• Click "Auto-discover" and verify the scan prompt appears
• Click "Scan cluster & discover SLOs" with APM data in the cluster
• Verify proposals appear with priority badges, category badges, and rationale
• Test select all / deselect all controls
• Click "Create N SLOs" and verify SLOs are created successfully
• Verify navigation to SLO list after creation
• Test with no data in cluster — should show "No SLOs discovered" warning
• Test "AI-assisted" mode — verify NL generation and refinement work
• Verify error handling when no AI connector is configured

Made with Cursor

[cla-checker-service]

💚 CLA has been signed

[elasticmachine]

Pinging @elastic/obs-ux-management-team (Team:obs-ux-management)

[elasticmachine]

💔 Build Failed

• Buildkite Build
• Commit: dedd61e

Failed CI Steps

• Linting
• Check Types

Metrics [docs]

Module Count

Fewer modules leads to a faster build time

id	before	after	diff
slo	1386	1396	+10

Async chunks

Total size of all lazy-loaded chunks that will be downloaded as the user navigates the app

id	before	after	diff
slo	1.1MB	1.1MB	+28.1KB

History

• 💔 Build #395492 failed 90f611d
• 💔 Build #395477 failed 654f8f1
• 💔 Build #395466 failed 592f135
• 💔 Build #395432 failed ce37ce4