[FEATURE]: Independent TTFT toggle in TUI settings #53316

Activity

1. github-actions

   assigned

   kommander

   on Oct 5, 2026
2. github-actions commented on Oct 5, 2026

   

   github-actionsBot

   on Oct 5, 2026 – with GitHub Actions

   Contributor

   More actions

   This issue might overlap with existing issues. Please check:

   • [FEATURE]: Performance tab with TTFT / tok/s / wall time per LLM call in the web app #49381: Requests TTFT and tok/s metrics per LLM call, though scoped to a web-app Performance tab.
   • [FEATURE]: Real-time token generation speed (tok/s) in TUI status bar, like Hermes #47914: Requests configurable live and final tok/s display in the TUI status/footer; related to the performance metric proposed here.
3. opencodejr commented on Oct 27, 2026

   

   opencodejrBot

   on Oct 27, 2026

   More actions

   Thanks for the detailed proposal! You're asking for a separate TTFT option in the TUI /settings, independent of the tokens/second option. When it's on, TTFT would appear in the assistant response footer (e.g. Build · GPT-6.1-Sol · 7.6s · TTFT 0.8s · 13.0 tok/s). It would be measured from sending the model request to the first generated token, reasoning included, and shown only when a real measurement exists.

   The maintainers will look at it and decide on the design. Some notes from reading the current code:

   • The footer only has the tok/s toggle right now (session.tps). Messages don't store a first-token time yet: time.streamed is recorded when the stream ends, and that is intentional (see [Bug]: session time.streamed is never persisted and step.streamed fires after the stream ends, not at first token #57178). So this would need a new timestamp in core as well as the TUI change.
   • This can't be built as a TUI plugin yet, because plugins have no slot for each assistant message. tui: Add a per-assistant-message footer slot to the TUI plugin API #46268 asks for one.

   Related issues:

   • [FEATURE]: Performance tab with TTFT / tok/s / wall time per LLM call in the web app #49381: TTFT, tok/s and wall time for each LLM call, in a web-app Performance tab
   • [FEATURE]: Remember debug bar visibility between restarts (or add a setting for it) #54121: the desktop debug bar, which already shows TPS and TTFT
   • tui: Add a per-assistant-message footer slot to the TUI plugin API #46268: a TUI plugin slot for each assistant message's footer
   • tui: token rate uses the final metadata burst instead of model-call duration #46883 / tui: tok/s excludes reported reasoning tokens in V2 #48705: how the TUI calculates tok/s