rbrooks/Quest-Board

Watch

v4.0.0 - Session Pipeline Accuracy

New issue

P0. Make the recording-to-summary pipeline correct and, crucially, VERIFIABLE. Deterministic speaker-to-character resolution in code, structured beat extraction with a pure-code validator so chronology is a sorted() guarantee rather than a model behaviour, transcript segment rows, and the context-budget hygiene set. The accuracy floor must not depend on model size — self-hosted small models are a first-class target. See docs/.internal/session-lifecycle-review-2026-08.md sections 4.1, 4.2 and 5.

No due date

100% Completed

[Backend] The silence guard can delete a quiet player from a full-length session

#425 by claude-bot was closed 2026-08-28 16:48:13 +00:00 0 / 4

[Backend] Summary quality varies 8x run-to-run on identical input

#423 by claude-bot was closed 2026-08-27 21:45:31 +00:00 0 / 4

[Backend] The audio retry endpoint disarms both duration guards

#421 by claude-bot was closed 2026-08-27 23:27:56 +00:00 0 / 4

[Backend] Let GMs set a campaign's spoken language, and pass it to transcription

#419 by claude-bot was closed 2026-08-27 23:27:52 +00:00 0 / 7

[Recording] Guard against re-submitting audio for a session that already has GM-edited content

#397 by claude-bot was closed 2026-08-28 00:22:31 +00:00 0 / 6

[Backend] Add a golden-transcript accuracy eval harness

#349 by claude-bot was closed 2026-08-26 18:28:54 +00:00 0 / 5

[Backend] Drop all-silence tracks before transcription

#348 by claude-bot was closed 2026-08-28 16:48:16 +00:00 0 / 5

[Backend] Fix per-span language detection and lost context on the VAD path

#347 by claude-bot was closed 2026-08-26 03:48:24 +00:00 0 / 5

[Backend] Validate highlights' speaker and timestamp against real segments

#346 by claude-bot was closed 2026-08-27 23:27:51 +00:00 0 / 4

[Backend] Emit start-end ranges and a deterministic tie-break in the merge

#345 by claude-bot was closed 2026-08-26 17:21:15 +00:00 0 / 4

[Backend] Disambiguate speaker labels — nicknames, collisions, and the snowflake fallback

#344 by claude-bot was closed 2026-08-28 19:16:24 +00:00 0 / 5

[Backend] Verify the returned speaker set matches the tracks that were sent

#343 by claude-bot was closed 2026-08-27 23:28:35 +00:00 0 / 4

[Backend] Stop the ASR server's speaker label overriding the known track owner

#342 by claude-bot was closed 2026-08-28 19:16:20 +00:00 0 / 4

[Backend] Compact the transcript prompt — merge same-speaker runs and use block timestamps

#341 by claude-bot was closed 2026-08-27 23:27:23 +00:00 0 / 5

[Backend] Bound the four remaining unbounded LLM contexts

#340 by claude-bot was closed 2026-08-27 23:27:22 +00:00 0 / 5

[Backend] Replace the Anthropic assistant prefill with structured outputs

#339 by claude-bot was closed 2026-08-27 23:28:31 +00:00 0 / 4

[Backend] Extend truncation rejection and token budgeting to the prose paths

#338 by claude-bot was closed 2026-08-28 16:38:01 +00:00 0 / 5

[Backend] Set Ollama num_ctx and verify the prompt was actually consumed

#337 by claude-bot was closed 2026-08-28 15:59:42 +00:00 0 / 5

[Backend] Preflight every prompt against the provider's declared context window

#336 by claude-bot was closed 2026-08-28 16:37:58 +00:00 0 / 5

[Backend] Persist transcript segment rows and render the text format as a view

#335 by claude-bot was closed 2026-08-28 19:16:10 +00:00 0 / 6

1 2