For AI agents: the complete documentation index is available at https://docs.flezise.com/llms.txt, the full documentation bundle is available at https://docs.flezise.com/llms-full.txt, and this page is available as Markdown at https://docs.flezise.com/administration/request-tracing.md.
Docs

Request tracing

Request Tracing is the gateway request-log explorer. It shows stats, a volume histogram, and a searchable table of every request the AI Gateway has handled. Drill into any row to read the raw message payloads and tool-call details.

::::note Read this if you administer the platform AI Gateway. Logs are written by the gateway's log_store component; if the table is empty, verify that component is running. ::::

Request Tracing showing five stat cards, a Request Volume histogram, a filters row with time range, provider, model, and status selectors, and the request table below

Read the stats and histogram

Open Settings → Request Tracing ("Monitor and inspect AI gateway requests, response times, token usage, and costs.").

Five stat cards span the top of the page:

CardWhat it shows
Total RequestsCount of requests in the selected time range
Success RatePercentage of requests that completed without error
Avg LatencyMean response time across all requests
Total TokensAggregate token consumption
Total CostAggregate cost in the configured currency

A Request Volume histogram sits below the stat cards and plots request counts over time.

Filter requests

The filters row lets you narrow the request table by time range, All Providers, All Models, All Status, or free-text using Search content.... Combine filters to isolate a specific provider, model, or failed request type.

If the table stays empty after filtering, the page shows: "Ensure AI Gateway is running with log_store enabled."

Inspect a request

The request table columns are Time, Type, Message, Provider, Model, Latency, Tokens, and Cost. Select a row to open the detail sheet. The sheet shows the raw message payloads via ChatMessageView and any tool calls via ToolCallView.

Next steps

  • Token Saving: reduce token consumption upstream of this log
  • Guardrails: screen traffic before it appears in the log
  • Run History: org-wide generation run history across all projects