Total spend (live)
£0.00
selected window
Financial command
NEXUS / COST INTELLIGENCE
Live operational oversight for agent spend, model routing, task cost, and cheap-lane leakage. This surface is tuned for fast diagnosis, forecasting, and intervention.
Export view0 runs in scope
Powered by Nexus sub-agent execution telemetry. Historical runs without token fields under-report cost; newer runs are more reliable.
Daily trend
+0.0%
vs previous hour
trendWeekly burn rate
£0.00
current run rate
Avg cost / task
£0.00
execution average
Cost anomaly
£0.00
last hour spend
Top spending agent
—
0% of selected spend
Top spending model
—
0% of selected spend
Most expensive task
—
Leakage score 0/10
Leakage overview
£0.00
0 high-risk leakage sources
Cost timeline
Burn curve
00:00
00:00
00:00
00:00
00:00
00:00
00:00
00:00
00:00
00:00
00:00
00:00
Tier classification
Spend by tier
Critical£0.00
0% of spend
Moderate£0.00
0% of spend
Cheap£0.00
0% of spend
Model family
Spend by family
No model-family cost data yet.
Forensics
Agent cost breakdown
| Agent | Tasks | Runtime | Tokens | Cost | Avg Task | Cost / 1k | Actions |
|---|
Forensics
Model cost breakdown
| Model | Family | In | Out | Total | Cost | Cost / 1k |
|---|
Operational leakage
Task cost breakdown
| Task | Agent | Model | Runtime | Tokens | Cost | Leakage | Actions |
|---|
Signal
Alerts
No active anomaly alerts in the selected cost window.
Signal
Operational insights
No instrumented spend in the selected filter window yet.
Signal
Live feed
Document 1 / 3
OBSERVABILITY & DEVOPS
This page now reflects the current Nexus build honestly: what exists in the repo, what is still missing, and which observability upgrades make sense next.
Current BuildGapsRoadmap
EMBER jobs
12
0 active
Research success
100%
12 success
Saved briefs
100%
12 persisted
Avg confidence
97
heuristic confidence
Avg duration
40s
0 escalated
Live now
What this build actually includes
Portal build status
live nowThis Nexus surface is a real Next.js page in the current portal build and ships as part of the existing UI.
Application data
live nowThe repo currently uses Prisma with a local SQLite database for portal state in this build.
Operational scripts
live nowA small amount of operational automation exists today, including milestone forwarding and deadline watching scripts.
Reality check
What is not implemented here
- No GitHub Actions workflows were found in this repo.
- No verified Prometheus, Grafana, Loki, ELK, or Alertmanager setup was found in the codebase.
- No Terraform or Ansible infrastructure-as-code implementation was found here.
- No proven live telemetry dashboards, alert routing, or incident pipeline is represented by this build alone.
- No restore-tested backup and disaster recovery automation is documented as implemented in this repo.
Recommended next layer
Practical sequence
Phase 1
Truthful status surface
Keep this page grounded in what is actually shipped. Separate live capabilities from recommended next steps.
Phase 2
Basic telemetry
Add health checks, structured logs, key counters, and a small internal status board before adopting a full observability stack.
Phase 3
Real alerting
Wire alert thresholds to real failure conditions such as portal downtime, queue stalls, repeated task failures, or deadline misses.
Phase 4
Deployment automation
Introduce CI for lint and build, then controlled deployment automation once release steps are stable.
Phase 5
Advanced stack
Only add Prometheus, Grafana, Loki, or Terraform once the operating model and team response flow are real requirements.
Reference only
Candidate stack for later
Metrics
PrometheusGood future choice if Nexus needs durable scrape-based metrics and alert rules.
Logs
Loki or ELKUseful when logs need central search, retention, and drill-down during incidents.
Dashboards
Grafana or KibanaHelpful once live metrics and logs exist and operators need one review surface.
Alerts
AlertmanagerWorth adding after thresholds, ownership, and escalation routes are clearly defined.
These tools are recommendations, not evidence that the current Nexus deployment already runs them.