API Console — Help
Data source
/home/virtual/vps-2b02c5/b/badb96c1f8/public_html/user1846601-recursion_engine/engine_data/store.jsonFile exists
Yes
File updated
2026-07-13 05:07:45 (server local time)
Now (UTC)
2026-07-28 21:13:14
Total records
0
Latest run
—
Pages
- api_metrics.php — Overview dashboards: totals, cost, error-rate, latency (p50/p90/p95), calls/day (with MA), recent calls/min, per–day & per–model tables, CSV exports, drill-downs.
- api_runs.php — Filterable run list (date range, exact day, mode, model), pagination, link to run detail.
- api_run.php — Single run: key metrics, prompt/system/final (if logged), error blob, raw JSON record.
- api_alerts.php — Alerts dashboard: cap usage (day/hour/min), 24h error-rate spike vs 7d, latency p95 spike.
- api_search.php — Text/regex search over system/prompt/final with filters and links to details.
All timestamps in UI are shown in UTC for consistency.
Query parameters
| Param | Used on | Description | Example |
|---|---|---|---|
since | metrics, runs, search | Lower bound date (inclusive). YYYY-MM-DD. | metrics?since=2025-01-01 |
until | metrics, runs, search | Upper bound date (inclusive). YYYY-MM-DD. | metrics?until=2025-12-31 |
mode | metrics, runs, alerts, search | Filter engine mode (e.g. recursion_api). Empty = all. | alerts?mode=recursion_api |
model | metrics, runs, alerts, search | Filter by model string. | runs?model=gpt-5-thinking |
ma | metrics | Moving average window in days (calls/day chart). | metrics?ma=14 |
mins | metrics | Window in minutes (calls/min sparkline). | metrics?mins=720 |
day | runs | Exact day (YYYY-MM-DD). | runs?day=2026-07-28 |
per | runs, search | Rows per page (25–500). | search?per=200 |
q | search | Search query (substring or regex). | search?q=timeout |
regex | search | Use PCRE regex with delimiters (/.../i). | search?q=/error|fail/i®ex=1 |
CSV exports
- Days — date aggregates: calls, tokens, USD/GBP, OK/Err.
- Runs — per-call rows: timestamps, mode, model, tokens, costs, status, preview, duration.
- Models — per-model aggregates: calls, tokens, costs, cost per 1k tokens.
Rate Guard (overview)
The spend/throughput guard lives only in API-call scripts (e.g. api_form.php, api_bulk.php). It does not run on Metrics/Runs/Alerts/Search pages.
- Config:
/home/virtual/vps-2b02c5/b/badb96c1f8/public_html/user1846601-recursion_engine/engine_data/rate_limits.json - Library:
engine/rate_guard.php - Denies with HTTP
429and JSON{"error":"rate_guard",...}when caps are reached. - Caps supported: daily USD, hourly calls, minute calls, optional allowed_modes.
Troubleshooting
- No data? Ensure
store.jsonexists and your call scripts append records correctly. - 429 on UI? Remove guard includes from read-only pages.
- Times look off? UI shows UTC; log ISO timestamps.
- Store large? Use
tools/store_maint.php(keep days + snapshots).