Dashboard
The Prysm dashboard gives you a real-time view of your LLM usage. It's organized into four main sections, accessible from the sidebar navigation.
Overview
The main dashboard page shows key metrics at a glance:
| Widget | What It Shows |
|---|---|
| Metric cards | Total requests, total tokens, total cost, average latency (with p50/p95/p99) |
| Request volume chart | Requests over time, grouped by hour or day |
| Latency distribution | Histogram showing latency spread across requests |
| Error rate chart | Error percentage over time |
| Cost accumulation | Cumulative spend over time |
| Model usage breakdown | Pie chart of requests per model |
| Live trace feed | Real-time WebSocket feed of incoming requests |
Request Explorer
A searchable, filterable table of every trace. Click any row to open the detail panel, which shows the full prompt messages, completion, token counts, latency, cost, model, status, tool calls, logprobs, and any metadata you attached.
Filter by model, status (success/error), date range, or search by content.
API Keys Page
Create, view, and revoke API keys. Each key shows its prefix (sk-prysm-xxxx...), creation date, and a quick start code snippet.
Settings
Manage your project configuration (provider, base URL, model, API key), alerts, team members, usage tracking, and custom model pricing.