Chat tool telemetry

Find slow, failing, and repetitive agent tools without exposing prompts, arguments, or outputs.

PeriodToolTeam

Detailed production timing active

6101 of 6166 saved tool calls include execution duration.

98.9% timing coverage

Tool calls

New activity vs previous 7d2 saved production turns

p50 latency

329 ms

Timing not instrumented2 timed calls

p95 latency

342 ms

Timing not instrumented2 timed calls

Tool failures

0.0%

No change vs previous 7d0 excluding NoResults

No result

0.0%

No change vs previous 7d0 product outcomes

Calls / turn

1.0

New activity vs previous 7d0.0% of turns at 5+ calls

What changed

Material regressions ranked by severity and affected call volume.

Current 7 days compared with the previous equal window

Baseline pending

A comparison window is not available yet

The page needs activity in the previous equal-length period before it can identify regressions.

Use the current metrics as a starting baseline and revisit after another full period.

Investigate

Regression checks require at least 3 calls in both windows. Latency is flagged at +20%; failure and no-result rates are flagged at +1 and +3 points respectively.

Execution volume

Successful calls, no-results, and failures combined.

Calls

Oct 1Oct 4Oct 7

Payload pressure

How often results exceed the 20 KB inline cap.

5.0%

handle rows / all calls

309 stored handles in this period. This is a payload-size proxy, not a one-to-one call classification.

Tool comparison

Each change is measured against the previous equal window; latency appears only where detailed timing exists.

Sorted by current p95 latency

Tool	Calls	p50	p95	Failures	No result
save_agent	2No prior data	329 msNo prior timing	342 msNo prior timing	0.0%No prior data	0.0%No prior data

Failure codes

Operational failures only; NoResults is excluded.

No operational failures

in the selected view

Interpretation notes

The data has known edges that change what a chart means.

Blocked loops are invisible: Rate-limited and identical-argument calls are rejected before telemetry fires.
NoResults is not downtime: It remains a failed tool outcome, but is separated from runtime reliability here.
Preview volume lives elsewhere: Preview and threadless chats reach PostHog but are intentionally absent from Postgres.
Latency is partially instrumented: Saved tool parts provide production outcomes and volume; duration appears only on calls with detailed telemetry.
Partial turns are retained: They remain in volume counts but are excluded from failures and latency percentiles.
Comparisons use equal windows: Every delta compares the selected period with the immediately preceding period under the same tool and team filters.

Recent sanitized calls

Arguments, outputs, user IDs, and message content are never rendered.

Privacy safe

save_agentTRP 2329 msSep 30, 5:55 PM UTC

Outcome

Success

Thread

a8fe7c6c…

Tool call

call_Vmi…

Turn sequence

1 of 1

Turn state

Completed

save_agentTRP 2342 msSep 30, 3:40 PM UTC