feat: add durable Activity history and print compatibility

[lidge-jun]

Activity history now survives reconnects and server restarts through a bounded journal in the existing trace store. Trace admission captures the owning chat and execution scope; replay and raw reads enforce that ownership, including historical backfills and forked pointers.

Print runtimes observe accepted message/tool updates while the existing lifecycle continues to select and deliver final answers. Journal, link, and observer failures preserve successful message writes and existing send/retry behavior. Display-only settings default to Activity and retain explicit legacy mode without execution side effects.

• Fixed-through replay, bounded row/byte budgets, corruption-aware cursor advancement, and whole-prefix retention.
• Latest tool state with stable run/ref identity, safe recovery after RAM eviction, and one omission marker.
• Peer-authored raw drawer session capture integrated without changing the ownership guard.
• Native terminal metadata/budget behavior remains the default; explicit print terminal updates use a separate opt-in.

Native Cursor main now uses the same captured chat/scope ownership at its fifth trace admission. The published parent is integrated by normal merges, including its gateway behavior regression follow-up.

Validation: 792 focused tests across 60 explicitly selected files passed after cascade, plus the parent follow-up checks. Root/frontend typechecks and builds, assets, strict baseline and 487 documentation counts passed. Owned Node Cursor fixtures exercise real factory/protocol/pool/lifecycle -> SQLite journal -> loopback HTTP with multi-session on/off, exact-owner reads, copied-fork denial, fixed-through replay, and storage failure containment. Removing the owner fields makes the regression fail; restoring them passes. Independent Astra review found no remaining blockers.

Current head 43d05fc5bddd1eb941b5b1374effbc96c52eb262 includes parent ef1fb23bf125513fd9a595d46cded24c18107689. Its Tests and Postinstall Platform Checks both passed on this exact head. Final-head regression: 54 tests, typecheck and 487 documentation counts passed; broader 792 tests plus 24 parent follow-up checks observed the identical production tree. Earlier CI remains historical. The existing instance authentication policy remains unchanged.

Stack: this backend layer targets #542 (codex/native-activity-02); dependent Web/TUI consumers reuse the isolated owner patch and fixture commit. No PR merge, release or deployment.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Team

Run ID: 9841fef1-addd-4f6e-967b-accee7a87f71

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

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.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-05T20:58:19.426092Z	2a043b1	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.