Select current-year writing; point demo and README at the write path - #47
Merged
Merged
Conversation
Co-authored-by: telivity-otaip <telivity-otaip@users.noreply.github.com>
telivity-otaip
marked this pull request as ready for review
July 30, 2026 18:51
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
selectcapped the corpus at a hardcodedDEFAULT_THROUGH_YEAR = 2024, so any corpus written in 2025 or later selected zero pieces.build-style-profilereads that selection, which means the style card driving everywritedraft was measured from a pre-2025 slice — or from nothing.run_build_writer_sftalready bypassed the year gate for exactly this reason;selecthad not caught up. The cap now defaults to the current year, and--through-yearstays for deliberate narrowing. Holdouts remain separate and explicit (--holdout-id).Also:
run_demonow runsselect → index-voice → build-style-profile → write, stubbing only the model call through the existinggenerate_fnseam. No weights download. Retrieval, style card, trim and guards are the real code. The mock train/filter tail stays becausetest_api_healthdepends on it.adapter=none model=stub.selectnow points atindex-voice;trainpoints ateval-write-holdoutwith the keep-only-if-it-beats-RAG rule.selectadded to the quick start —build-style-profilerequires it, so the documented path previously dead-ended.cli-status.pngandcli-demo.pngfrom real output, addedcli-setup.pngandscripts/shot.pyso docs images stay reproducible.ruff clean, 272 tests pass (3 new), sanitize OK.