Skip to content

#140 chore: catch up co-core 0.19.6 -> 0.36.0 - #141

Merged
gregoryfoster merged 5 commits into
mainfrom
140-co-core-0.36
Oct 17, 2026
Merged

gregoryfoster merged 5 commits into
mainfrom
140-co-core-0.36

Conversation

@gregoryfoster

@gregoryfoster gregoryfoster commented Oct 17, 2026 •

Copy link
Copy Markdown
Contributor

Refs #140. Draft until shipping.

co-core catch-up

  • Pins co-core[extract], co-core-aio[bus,gcs] and co-core-sync[gcs] at >=0.36.0,<0.37. uv.lock moves only those three.
  • No behaviour change. Every module in Replicator's co-core import closure is either identical bar docstrings (compared by AST) or byte-identical. The exceptions are additive and not called here: the pointer methods (cannobserv#496) and XML extraction (cannobserv#523). The evidence is in the pin comment and on Catch up co-core 0.19.6 → 0.36.0 ahead of the generation release #140.
  • The pin text is updated in AGENTS.md, docs/DEPLOYMENT.md and the SocratiCode manifest.

CR (blanket "address the findings")

Item Action Result
1 Fixed 0a34455: the closure-diff lesson trimmed to the rule and its case
2 Stet No red test: there's no behaviour to make fail. The evidence is the unchanged suite plus the integration and gcs runs
3 GH #142: drop the unused [extract] extra

Context curation (curating-context, exact counts)

File Before After Budget
AGENTS.md 5955 5991 6000
docs/DEPLOYMENT.md 10889 7570 10000
docs/MEMORY-PROTECTION.md (new) — 3778 10000

Dispositions (DEPLOYMENT.md):

  • "Memory protection" is class B: 3449 tokens, 31% of the doc. Every other subsection is class A and stays.
  • AGENTS.md was classified last week (99d8a22) and is unchanged apart from its index.

Relocated, verbatim:

Rewritten (retarget warrants in .skills/context-loss-ok):

  • DEPLOYMENT.md's two preamble lines.
  • "from the table above", which now points at DEPLOYMENT.md#server-lifecycle.
  • The AGENTS.md DEPLOYMENT.md index line. "The host's memory tunables" moves to a new MEMORY-PROTECTION.md line.

Repointed:

  • docs/COMMANDS.md
  • a docstring in tests/test_deploy.py
  • .skills/doc-sections
  • the manifest: the DEPLOYMENT entry is split, and a memory-protection-reference entry is added

Fixed (FALSE): the manifest said StartLimitBurst=6. The unit has had 8 since #134.

Deleted: nothing.

Gates:

  • prove-no-loss: lost 0, claims_dropped 0.
  • Seams: 0 open, and 5 newly judged legitimate.
  • Counts: 20, the same as at the branch point.
  • Links: none dead, no dead anchors, no orphans, nothing unindexed.
  • Telemetry: rows are in .skills/context-metrics.jsonl. Baselines were taken at f638266 and the after-rows at 2b1fd5b.

Accepted cost: AGENTS.md has 9 tokens of headroom, down from 45, which squeezes #137's charter link. See #140.

Local runs: full suite 1408 passed and 1 skipped (matches the baseline); -m integration 55 passed; -m gcs 16 passed. After the merge, run codebase_update from the main checkout, because the manifest descriptions changed.

🤖 Generated with Claude Code

gregoryfoster and others added 5 commits October 17, 2026 03:17
Seventeen-minor catch-up ahead of the cannobserv#568 release that #136
needs. No behaviour change: every module in Replicator's co-core import
closure is AST-identical bar docstrings or byte-identical, apart from
additive pointer verbs (cannobserv#496) and XML extraction
(cannobserv#523), neither called here. The pin comment records the diff.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
AGENTS.md, DEPLOYMENT.md and the SocratiCode manifest's pyproject entry
name >=0.36.0,<0.37. DEPLOYMENT.md adds that the wheel diff must cover the
import closure, compared by AST: changes.py pulls in pure/extract, which
no import line here names.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
DEPLOYMENT.md 10889 -> 7570 tokens (budget 10000); AGENTS.md 5955 -> 5991
(budget 6000). Exact counts.

Relocated, verbatim (prove-no-loss: lost 0, claims_dropped 0):
- DEPLOYMENT.md "### Memory protection — the worker outranks the dev session
  that shares this VM", body (3449 tokens, 31% of the doc, class B) ->
  docs/MEMORY-PROTECTION.md. The heading stays in DEPLOYMENT.md as a
  pointer subsection, the #99 FAILURE-NOTIFICATION.md shape.

Rewritten (retarget, judged in .skills/context-loss-ok):
- DEPLOYMENT.md preamble: memory tunables now point at the new doc.
- "from the table above" -> DEPLOYMENT.md#server-lifecycle's table.
- AGENTS.md index: "the host's memory tunables" moves from the
  DEPLOYMENT.md line to a new MEMORY-PROTECTION.md line.

Repointed: docs/COMMANDS.md (OOMScoreAdjust= cross-reference),
tests/test_deploy.py docstring, .skills/doc-sections (new lead line),
manifest (DEPLOYMENT entry split; new memory-protection-reference entry).

Fixed (FALSE): manifest's DEPLOYMENT description said StartLimitBurst=6;
the unit has 8 since #134.

Deleted: nothing. AGENTS.md otherwise unchanged: every section was
classified last week (99d8a22), and the one duplicate-shaped candidate
(the Overview's ARCHITECTURE.md pointer) is a CR decision (a19d77e).

Seams: 5 judged legitimate (acked), 0 open. Counts: 20 rot-prone in the
two docs, identical to the branch point.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…MENT.md 10889 -> 7570)

Baselines measured at the branch point f638266; after-rows at 2b1fd5b.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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