FE-736: JSONL session viability proof (M2)

[lunelson]

No description provided.

[lunelson]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• FE-744: Document Pi command containment evidence #150
• FE-737: Web shell over the same host (M3) #147
• FE-736: JSONL session viability proof (M2) #146 👈 (View in Graphite)
• FE-735: Mode shell and fixture driver (M1) #144
• FE-702: Walking skeleton — brunch binary + TUI over pi (M0) #142
• next

This stack of pull requests is managed by Graphite. Learn more about stacking.

[cursor]

PR Summary

Medium Risk
Changing RPC/projection to active-branch semantics can diverge from full-file reads until D24-L linear guards land; large SPEC/PLAN edits are doc-only but must stay aligned with code behavior.

Overview
M2 JSONL session viability is implemented and reflected in planning memory: new round-trip and M1 fixture replay tests cover raw payloads, brunch.session_binding, Brunch custom entries/messages, continuity/compaction metadata, structured elicitation, and coordinator reload paths. session.elicitationExchanges and related projection now load the Pi session’s active branch via loadActiveBranchTranscriptEntries (not naïve full-file line reads), with branch-shape tests documenting substrate behavior ahead of D24-L linear transcript enforcement.

Canonical docs advance in parallel: memory/SPEC.md / memory/PLAN.md promote M2 completion, set web-shell as active, add linear-session constraints, generative-lens / review-set decisions (D25–D32, I15–I20), and expand M5+ verification notes; new design companions ELICITATION_LENSES.md and REVIEW_SETS.md. Agent workflow adds ln-witness (test evidentiary audit) and updates docs/praxis/ln-skills.md.

^{Reviewed by Cursor Bugbot for commit c6c33ee. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit c6c33ee. Configure here.}

[cursor]

RPC test uses wrong cwd for branched session

Low Severity

createBranchedSessionFile creates the session in a temp directory but doesn't return cwd. The test then passes it to readyState(file) which hardcodes cwd: "/tmp/brunch-project". The production handler in rpc.ts now passes state.cwd to loadActiveBranchTranscriptEntries, so the test exercises the production code path with a mismatched cwd. The test passes because SessionManager.open tolerates a wrong cwd when given an absolute file path, but this means the test doesn't validate that cwd propagation actually works correctly for branch resolution.

Additional Locations (1)

• src/rpc.test.ts#L38-L56

 

^{Reviewed by Cursor Bugbot for commit c6c33ee. Configure here.}