Skip to content

perf(index): substantially accelerate text catch-up - #721

Merged
ScriptedAlchemy merged 71 commits into
codex/tracedecay-total-redesign-plan-reopenedfrom
codex/code-index-catchup-pipeline
Aug 25, 2026
Merged

perf(index): substantially accelerate text catch-up#721
ScriptedAlchemy merged 71 commits into
codex/tracedecay-total-redesign-plan-reopenedfrom
codex/code-index-catchup-pipeline

Conversation

@ScriptedAlchemy

@ScriptedAlchemy ScriptedAlchemy commented Aug 25, 2026

Copy link
Copy Markdown
Owner

Summary\n\n- batch authenticated lexical source pages with exact cursor rollback and atomic bulk commits\n- parallelize bounded relational preparation and publish real-time Code/Observatory progress\n- add feature-gated, low-cardinality Hotpath instrumentation plus repository/Codex MCP configuration\n- continue the production optimization until the cold text artifact is Ready within 300 seconds and below 8 GiB RSS\n\n## Current verified evidence\n\n- source batch tests: 6/6\n- disk artifact tests: 27/27 feature-off and 27/27 feature-on\n- sealed-source integration: 2/2\n- Hotpath reader lock regression: 1/1 feature-on\n- microbenchmark: 240 commits reduced to 15; 7.46x ingestion and 5.96x end-to-end; byte-identical digests\n- Rust library check, generated-contract check, dashboard typecheck/build, and focused Code/Observatory DOM suites green at the latest checkpoint\n\n## Truthful production RED baseline\n\nThe preserved external fixture contains 4,851 tracked files (4,399 admitted sealed files); the historical “10,592 files” count included 5,741 .git internals and is not the indexing denominator.\n\nFirst feature-off production run:\n- no Ready receipt after 414.821s\n- 92 pages / 11,324 chunks / 88 of 4,399 files committed\n- 26.96 MiB of 1.029 GiB lexical bytes complete; ETA ~1,565s\n- peak RSS 9,235,228 KiB (8.807 GiB)\n- 20.55 MiB source payload produced 27.59 GiB userspace writes and 3.59 GiB physical writes\n\nConfirmed remaining causes:\n- mutable worktree wakes cancel/restart immutable 1.47 GiB sealed-source verification\n- full active-generation decode retains multiple corpus-sized wire/object representations\n- rejected batches are restaged and one refusal keeps later batches artificially narrow\n- 33 per-row SQLite epoch triggers, online secondary indexes, and row-at-a-time derived stats roughly double five million mutations and amplify writes\n\n## Ongoing acceptance\n\n- generation-owned text cancellation with reconcile fairness\n- accepted-prefix batching without replay\n- zero-copy/streaming generation decode and a lightweight text owner\n- authenticated SQLite bulk-load followed by set-based stats and deferred indexes\n- feature-off and feature-on cold journeys under 300s / 8 GiB with identical generation/artifact receipts\n- live Chrome verification of Code and Observatory one-second progress through scan, prepare, commit, finalization, and Ready\n

@changeset-bot

changeset-bot Bot commented Aug 25, 2026

Copy link
Copy Markdown

⚠️ No Changeset found

Latest commit: e09665a

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

…plan-reopened' into codex/code-index-catchup-pipeline
…plan-reopened' into codex/code-index-catchup-pipeline
`fix(semantic): build activation candidate in daemon` removed
`wait_for_evaluated_semantic_profile_current` from this module but left its
only call site in `advanced_workflow_journey_test.rs` untouched, so the
auto-discovered `daemon_suite` target no longer compiles:

    error[E0425]: cannot find function
    `wait_for_evaluated_semantic_profile_current` in module `task_session`

The helper is restored verbatim from the last revision that carried it, which
also puts the module's now-unused `ApplicationConfigurationGet` import back to
work. It proves the evaluated profile is selected through the public
configuration authority and ready through the mounted runtime authority, and
that the readiness receipt authorizes the exact configuration the SDK selected.
@ScriptedAlchemy
ScriptedAlchemy marked this pull request as ready for review August 25, 2026 23:55
@ScriptedAlchemy
ScriptedAlchemy merged commit 8b1afa2 into codex/tracedecay-total-redesign-plan-reopened Aug 25, 2026
5 checks passed
ScriptedAlchemy added a commit that referenced this pull request Aug 26, 2026
…707-20260826-a1

perf(index): integrate post-#721 catch-up delta into #707
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