Analytics

The Analytics drawer turns recorded chat usage into a model-by-model table. It helps users understand which models they have used, how many requests were sent, how long responses took, how much throughput was recorded, and how many tokens were generated.

Analytics are based on saved app usage. They are not synthetic guarantees, hardware promises, or benchmark results.

Analytics Drawer

View Analytics

Administrators can use Enabled only to filter visible model rows using registry state. Overall totals remain unchanged. Missing measurements are not zero-performance results.

Column	What it shows
Model	The model identity recorded with completed chat responses.
Total Sent	Count of recorded prompt/response rows for that model.
Avg Time	Average recorded response time for that model.
Min / Max / Avg TPS	Recorded token-per-second values from saved responses.
TPS Viz	A small bar visualization scaled against the highest average TPS currently in the table.
# Tokens	Total recorded generated tokens for that model, plus an overall token total row.

What Counts

Signed-in user scope

The Analytics drawer reads the signed-in user's saved chat rows. Users see their own aggregate usage, not every user's chat activity.

Generated variants

The drawer includes generated responses from normal sends, regenerate actions, and edit variants when they are recorded in chat history.

Model comparison

The table can help compare how models have behaved during real app usage, especially response time, TPS, and token totals.

Admin Usage Views

User token totals

Admin tools can show token totals by user, along with saved row counts, session counts, and last recorded timestamps.

Chat administration

Admins can export a selected user's chats, export all chats, delete a selected user's chats, or delete all chat history through admin chat tools.

Limits

Recorded data only

Analytics depend on the data stored by completed app interactions. Missing, old, imported, or manually changed data can affect totals. Benchmark results live in the Benchmarks drawer and are tracked separately from chat analytics.

API Analytics

Administrator-only API Analytics shows request counts, errors, durations, backend-reported token usage, and paginated recent request metadata.

Avg Completion Time counts completed chat-completion events with an HTTP status below 400 and a recorded duration. Failed/cancelled completions and model-discovery requests are excluded from that average.

Tracking records operational metadata, not prompts, responses, images, keys, or authorization headers. Other client/proxy logging is a separate operator concern. See API access and troubleshooting.