Skip to content

README: describe the playback estimate and where per-reply instructions stop - #13

Merged
pmckeetx merged 1 commit into
mainfrom
docs-code-review-2026-10-14
Oct 14, 2026
Merged

pmckeetx merged 1 commit into
mainfrom
docs-code-review-2026-10-14

Conversation

@pmckeetx

Copy link
Copy Markdown
Collaborator

README only, no code change.

  • Known limitations, playback reports. Without reports, Dialog-RSN-1 estimates playback at 15 characters a second, makes the barge-in check for that window, and cuts history to the estimated heard text on barge-in (poly_core dialogue_api/turntaking/playback.py, session.py _estimate_playing and _cut_estimate). The README said any speech after generation finished interrupts playback.
  • generate_reply(instructions=...). The service honors per-response instructions. The plugin is what doesn't send them.

Not changed: the plugin's runtime warning still reads "Dialog-RSN-1 ignores per-reply instructions" (realtime_model.py:396). That's code, so it's left for a separate change.

🤖 Generated with Claude Code

…ns stop

Without playback reports Dialog-RSN-1 estimates playback at 15 characters a
second and cuts history to the estimate on barge-in. Per-reply instructions
are honored by the service; the plugin is what doesn't send them.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@pmckeetx
pmckeetx merged commit f5dd6bc into main Oct 14, 2026
7 checks passed
@pmckeetx
pmckeetx deleted the docs-code-review-2026-10-14 branch October 14, 2026 13:04
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant