📊 Roadmap: surface the live backend in the redesigned UI — wire what's already serving

[JSONbored]

The redesigned metagraphed-ui is merged and solidly wired — a cross-repo wiring audit (2026-06-18: two parallel deep-dives + live verification of every endpoint shape) found all 14 routes call real /api/v1 endpoints with network-scoped query keys, no mocks, no placeholders, nothing broken. AI search (/search/semantic), Ask (/ask) and the agent-catalog are wired.

The gap is the inverse of what you'd expect: very little is missing from the backend — a rich set of already-live endpoints is built and serving but unused by the UI. The redesign's "new data needs" (sparklines, latency heatmaps, growth, pulse trends) map almost 1:1 onto these. So this epic is mostly frontend wiring of data that already exists, plus three small backend enablers.

Tracked centrally here (FE work lands in JSONbored/metagraphed-ui). Labels carry the gittensor emission multiplier: priority 2.5× · feature 1.5× · bug 0.5×.

Where we stand

• Wired & healthy: subnets, providers, surfaces, endpoints, health, freshness, incidents, gaps, review/*, schemas, contracts, fixtures, agent-resources, adapters, AI search/ask. Field shapes verified live (e.g. /rpc/usage.summary matches the UI exactly; /incidents returns grouped surfaces[]).
• Live but unused (0 refs in src/): registry/leaderboards, economics (75 KB!), lineage, subnets/{n}/health/trends, …/health/percentiles, …/health/incidents, …/trajectory, …/uptime, subnets/{n}/overview, surfaces/{id}/verify, events (SSE). All verified returning real data (HTTP 200).

Phase 1 — Wire the marquee data (high impact, zero/low backend work)

• UI: wire the registry leaderboards into a discovery module #1111 Leaderboards discovery module (healthiest/fastest/most-complete/most-enriched/fastest-growing) · feature · seo
• UI: economics panel on subnet detail (validator / emission / alpha) #1112 Economics panel on subnet detail (validator/emission/alpha — 75 KB unused) · feature
• UI: lineage (testnet <-> mainnet) section on subnet detail #1113 Lineage (testnet↔mainnet) section on subnet detail · feature · good first issue

Phase 2 — Subnet-detail depth (live time-series the redesign wants)

• UI: subnet health time-series — trends, latency percentiles, SLA/incidents #1114 Health time-series: trends + latency percentiles + SLA/incidents · feature
• UI: structural growth (trajectory) + long-range uptime on subnet detail #1115 Structural growth (trajectory) + long-range uptime · feature
• UI perf: consolidate subnet detail to the composed /overview call #1116 Consolidate detail page to the composed /overview call (perf) · feature

Phase 3 — Live & interactive

• UI: real-time updates via the /api/v1/events SSE feed #1117 Real-time updates via the /api/v1/events SSE feed · feature
• UI: 'verify now' button — live surface re-probe #1118 "Verify now" surface re-probe button · feature · good first issue
• UI bug: LatencyHeatmap should use real latency_ms (+ percentiles), not last_probed_at #1119 LatencyHeatmap: use real latency_ms (+ percentiles), not last_probed_at · bug

Quality

• UI bug: verify the coverage KPI field mapping (no silent zeros) #1120 Verify the coverage KPI field mapping — no silent zeros · bug

Backend enablers (unblock the above)

• API: bulk health-trends endpoint for the all-subnets matrix #1121 Bulk health-trends endpoint for the all-subnets matrix (unblocks UI: subnet health time-series — trends, latency percentiles, SLA/incidents #1114 matrix sparklines) · feature
• API perf: per-provider counts on /api/v1/providers #1122 Per-provider counts on /api/v1/providers (kills a ~10k-row client aggregation) · feature
• API: intra-window buckets on /api/v1/rpc/usage #1123 Intra-window buckets on /api/v1/rpc/usage (proxy usage heatmap) · feature

Verified evidence (live, 2026-06-18)

• /api/v1/registry/leaderboards → 200, 7.8 KB, 5 boards · /api/v1/economics → 200, 75 KB · /api/v1/lineage → 200, 6.8 KB — all rendered nowhere.
• /api/v1/subnets/64/{health/trends,health/percentiles,trajectory,uptime,overview} → all 200 with real data; 0 references in metagraphed-ui/src.
• /api/v1/endpoints rows already carry latency_ms — the LatencyHeatmap reads last_probed_at instead (UI bug: LatencyHeatmap should use real latency_ms (+ percentiles), not last_probed_at #1119).
• providers.index.tsx fetches up to ~10k surface+endpoint rows in-browser to derive per-provider counts (API perf: per-provider counts on /api/v1/providers #1122).
• /api/v1/coverage exposes chain_subnet_count / application_subnet_count / official_surface_count / …, not the UI's netuids_total / surfaces_total — confirm the mapping (UI bug: verify the coverage KPI field mapping (no silent zeros) #1120).

Complements #740 (growth/DX) and #999 (data integrity). This epic is the frontend-surfacing through-line: make the UI as rich as the API already is.