Skip to content

[pull] main from cdeust:main - #119

Merged
pull[bot] merged 3 commits into
asleekgeek:mainfrom
cdeust:main
Sep 16, 2026
Merged

pull[bot] merged 3 commits into
asleekgeek:mainfrom
cdeust:main

Conversation

@pull

@pull pull Bot commented Sep 16, 2026 •

Copy link
Copy Markdown

See Commits and Changes for more details.


Created by pull[bot] (v2.0.0-alpha.4)

Can you help keep this open source service alive? 💖 Please sponsor : )

cdeust and others added 3 commits September 15, 2026 23:19
…ratch-and-commit install (#573) (#576)

Both installers ran `pip install --target` straight into the persistent
deps directory. pip skipped every package directory that already existed
but still wrote the new *.dist-info, so the 4.21.0 -> 4.22.0 upgrade kept
sentence-transformers 5.6.1 code under 6.0.1 metadata and verification
failed on tokenizers.

- launcher_deps.install_requirements (also `python3 launcher_deps.py -r
  FILE DEPS_DIR`) installs a generated closure into scratch under the
  deps lock and commits it through _commit_resolved_entries.
- setup.sh step 3 moves to scripts/lib/install_python_deps.sh and calls
  it; setup.py calls it as a subprocess. No direct --target remains.
- dist_info_versions omits a distribution with several *.dist-info, so a
  directory an earlier upgrade left mixed is repaired whatever order the
  filesystem lists it in.

Decision: ADR-1063. Follow-up on guard churn: #575.

Co-authored-by: Claude Opus 5 (1M context) <[email protected]>
…575) (#577)

* fix(launcher): read deps entry owners from the scratch RECORD files (#575)

The idempotence guard derived an entry's distribution from its name, so
every entry named differently from its distribution (yaml, sklearn,
google, bin, the shared __pycache__, ...) was replaced on every install.
It now reads each entry's owners from the scratch install's RECORD files
and skips the entry only when every owner is already at the resolved
version and the entry is still present (ADR-1064).

Measured on an APFS clone of a live 4.22.0 deps directory: 27 of 213
entries replaced before, 0 after, tree byte-identical to the original.

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>

* refactor(launcher): move the post-commit pruning out of _commit_resolved_entries (#575)

The owners lookup pushed _commit_resolved_entries to 43 AST lines,
over the craftsmanship gate's 40-line method limit when run with
--base origin/main. The dist-info and foreign-ABI pruning now lives in
_prune_after_full_commit, called at the same point, behaviour unchanged.

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>

---------

Co-authored-by: Claude Opus 5 (1M context) <[email protected]>
#578) (#580)

handler_refine wrote a fixed 0.85 into every refined draft and its
refined_llm memo. The synthesizer sets a draft's confidence to the mean
of its source claims, and the curator approves only from 0.6, so a
prose rewrite put any draft past the approve threshold. The handler no
longer passes a confidence to update_draft, and the memo records the
draft's own confidence (ADR-1065).

The handler is not registered as an MCP tool; the maintainer's store
holds one refined_llm memo, from a 2026-04-14 test runner. The three
new tests fail on the previous handler and pass on this one.

Co-authored-by: Claude Opus 5 (1M context) <[email protected]>
@pull pull Bot locked and limited conversation to collaborators Sep 16, 2026
@pull pull Bot added the ⤵️ pull label Sep 16, 2026
@pull
pull Bot merged commit c848d51 into asleekgeek:main Sep 16, 2026
4 checks passed
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant