Performance Insights
Rockxy tracks timing and size data for every captured request and computes per-endpoint statistics in real time. Use performance insights to find slow endpoints, oversized payloads, blocking dependency chains, and regressions between sessions.
Entry Points
| Action | How to Access |
|---|---|
| Performance tab | Performance tab in the main sidebar |
| Timeline view | Timeline tab in the main sidebar |
| Per-request timing | Select a request > Timing tab in the inspector |
| Threshold settings | Settings > Performance |
Latency Analysis
Rockxy computes latency percentiles for every endpoint that has received at least 5 requests in the current session. The latency table shows:| Endpoint | Requests | p50 | p95 | p99 | Avg Size | Error Rate |
|---|---|---|---|---|---|---|
GET /api/v2/users | 142 | 45 ms | 120 ms | 340 ms | 2.1 KB | 0.7% |
POST /api/v2/orders | 38 | 180 ms | 450 ms | 890 ms | 8.4 KB | 5.3% |
GET /api/v2/products | 271 | 32 ms | 85 ms | 210 ms | 14.2 KB | 0.0% |
- p50 — the median latency. Half of all requests to this endpoint completed faster.
- p95 — the 95th percentile. Only 5% of requests were slower.
- p99 — the 99th percentile. Captures worst-case latency excluding extreme outliers.
Slow Request Detection
Requests that exceed the p95 latency threshold for their endpoint are flagged as slow. Slow requests are:- Highlighted in the traffic list with a warning indicator.
- Listed in the Slow Requests section of the Performance tab.
- Included in trend tracking for regression detection.
Thresholds are calculated per endpoint, not globally. A 200 ms request to a fast endpoint (p95: 100 ms) is flagged, while a 200 ms request to a slow endpoint (p95: 500 ms) is not. This prevents alert fatigue from endpoints with legitimately different latency profiles.
Payload Size Analysis
Rockxy tracks request and response body sizes for every endpoint and surfaces size outliers:- Average size per endpoint, tracked over time.
- Oversized responses — responses significantly larger than the endpoint’s average are flagged.
- Size breakdown — request size vs. response size, with compression ratio when Content-Encoding is present.
Request Timeline
The timeline view shows a visual waterfall chart of all requests in the current session, ordered by start time. Each request bar is broken into timing phases:| Phase | Description |
|---|---|
| DNS | Domain name resolution |
| Connect | TCP connection establishment |
| TLS | TLS handshake (HTTPS only) |
| TTFB | Time to first byte — waiting for the server to respond |
| Transfer | Response body transfer |
Dependency Detection
Rockxy automatically detects sequential request chains — cases where request B starts only after request A completes, suggesting a data dependency or waterfall pattern. Detected chains are highlighted in the timeline view with a connecting line between dependent requests. The Performance tab lists the longest chains with their total duration and individual request breakdown. Use dependency detection to find opportunities for:- Parallelizing independent requests.
- Prefetching data that is needed later.
- Reducing chain depth by combining API calls.
Trend Tracking
Rockxy compares current session metrics against historical baselines from previous sessions. The trend view shows:- Latency trends — is the p95 for an endpoint increasing or decreasing compared to previous sessions?
- Error rate trends — are errors becoming more or less frequent?
- Payload size trends — are responses growing over time?
Baselines are calculated from the most recent 10 completed sessions stored in SQLite. If fewer than 3 historical sessions exist for an endpoint, trend data is not shown for that endpoint.
Next Steps
Error Analysis
Automatic error detection and pattern grouping across network and logs.
Traffic Capture
How Rockxy captures HTTP, HTTPS, and WebSocket traffic.
Log Intelligence
Capture and correlate OSLog streams with network requests.