feat(rebuilder): default trigger_mode threshold (#746)

[robotrocketscience]

Closes #746.

What this PR does

Flips DEFAULT_TRIGGER_MODE in src/aelfrice/context_rebuilder.py from TRIGGER_MODE_MANUAL to TRIGGER_MODE_THRESHOLD. Fresh installs now experience the PreCompact augment-mode rebuild without configuration; opt-out remains available via [rebuilder] trigger_mode = "manual" in .aelfrice.toml.

Why now

Both bench gates referenced by the v1.4.0 ship-default conservatism note have closed:

• feat(eval-harness): wire context-rebuilder eval harness for #587 hot-start scoring #592 (eval-harness commit-3, closed 2026-05-13). Hot-start 100% at trigger_threshold <= 0.6, cold-start 75% across t ∈ {0.5..0.8}.
• eval-harness: Cohen's-κ judge calibration before #592 close #687 (Cohen's-κ multi-run ratification, closed 2026-05-13). Post-bench validation gate.

The original manual default was explicitly framed as "until production telemetry confirms the calibrated threshold" — the telemetry is in, the threshold cleared, the gate exists to be flipped.

Commits (atomic)

1. feat(rebuilder): default trigger_mode threshold (#746) — one-line constant flip + docstring rewrite citing feat(eval-harness): wire context-rebuilder eval harness for #587 hot-start scoring #592/eval-harness: Cohen's-κ judge calibration before #592 close #687 closure.
2. test(rebuilder): default trigger_mode threshold (#746) — renames test_tm1_default_trigger_mode_is_manual → _is_threshold, asserts threshold at both the constant and RebuilderConfig() default; adds new test_tm1_no_config_file_defaults_to_threshold exercising the user-visible promise (no TOML present → threshold); updates obsolete v1.4-vintage comments in test_hook_pre_compact.py that described threshold as opt-in.
3. docs(rebuilder): default trigger_mode is threshold (#746) — docs/context_rebuilder.md roadmap row + TOML example + mode descriptions reordered so threshold leads with gate-clearance history.
4. docs(changelog): note rebuilder default trigger_mode flip (#746) — [Unreleased] > Changed entry.

Acceptance coverage (#746 checklist)

• DEFAULT_TRIGGER_MODE flipped to TRIGGER_MODE_THRESHOLD (context_rebuilder.py:197)
• Docstring at context_rebuilder.py:198-209 updated to reflect gate clearance
• Opt-out via [rebuilder] trigger_mode = "manual" verified working — existing test_tm1_manual_mode_pre_compact_hook_no_ops covers this path; the config-load + hook-dispatch paths already honor the override.
• AELFRICE_NO_REBUILDER env override — checked: no such env var exists in the codebase (grep -rn 'AELFRICE_NO_REBUILDER\|NO_REBUILDER' returns nothing). TOML-based opt-out is the only opt-out and was already the documented surface. The acceptance bullet appears to be aspirational; not adding one in this PR since the issue body itself frames it as "decide whether to add one" and TOML opt-out is sufficient.
• docs/context_rebuilder.md updated. README scan showed no remaining "opt-in via .aelfrice.toml" wording in the v3 revision, so no README change required.
• CHANGELOG entry added under [Unreleased] > Changed.
• Tests updated for new default.

Out of scope (per the issue body)

• Suppress-mode (parked for v2.x).
• Dynamic trigger-mode (parked at v1.4.0).
• Token-budget recalibration (gated on captured production corpus).
• Stop-hook cadence checkpoint (different design space).

Risk

• Per-compaction token spend rises by the rebuild block size (~5-6 KB observed). CHANGELOG entry sets the expectation explicitly. For token-sensitive workflows the opt-out remains a one-line TOML write.
• Calibration drift. threshold_fraction = 0.6 is fixture-bound. Users opting into custom calibration already have the override path ([rebuilder] threshold_fraction = X); this flip doesn't change that.

Test plan

• uv run pytest tests/test_rebuilder_triggers.py tests/test_hook_pre_compact.py → 24 passed.
• uv run pytest -q -x (full suite) → 3821 passed, 59 skipped, 75 xfailed.
• Discretion grep on diff → clean.

[coderabbitai]

Warning

Rate limit exceeded

@robotrocketscience has exceeded the limit for the number of commits that can be reviewed per hour. Please wait 2 minutes and 11 seconds before requesting another review.

You’ve run out of usage credits. Purchase more in the billing tab.

⌛ How to resolve this issue?

After the wait time has elapsed, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout.

Please see our FAQ for further information.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: fbb84fbc-7374-4294-94cf-a3b4401890aa

📥 Commits

Reviewing files that changed from the base of the PR and between bf288f5 and 77af4a4.

⛔ Files ignored due to path filters (1)

• CHANGELOG.md is excluded by !**/CHANGELOG.md

📒 Files selected for processing (4)

• docs/context_rebuilder.md
• src/aelfrice/context_rebuilder.py
• tests/test_hook_pre_compact.py
• tests/test_rebuilder_triggers.py

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/issue-746-rebuilder-default-threshold

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[sourcery-ai]

Sorry @robotrocketscience, you have reached your weekly rate limit of 500000 diff characters.

Please try again later or upgrade to continue using Sourcery

[github-actions]

PR-size soft cap

This PR is over the advisory size threshold:

• 88 changed lines (limit: 200)
• 5 changed files (limit: 3)

Bigger PRs collide with more open work, which under the parallel-session workflow tends to produce repeated attn:merge-conflict cycles (see #602). When practical, split into smaller PRs that each touch a focused surface.

This is advisory only — nothing is blocked. If the size is intentional (large refactor, module removal, generated code), apply the size:override label and this comment will be removed on the next push.

[robotrocketscience]

[claim:review:curie:2026-05-13T21:26:12Z]

[robotrocketscience]

[release:review:curie:2026-05-13T21:27:29Z]