perf(server): build the thread list snapshot without decoding it twice

[t3dotgg]

The server decoded the thread list snapshot twice. getShellSnapshot and getArchivedShellSnapshot build it from rows that SqlSchema already decoded. Then they ran a second full OrchestrationShellSnapshot decode over the result. In a local benchmark with 5,191 thread shells, that second decode took about 45 ms and made a full copy of every shell. The snapshot is built for HTTP, the shell socket, and about 8 background callers. Two reactors rebuild it every minute. This came from a slowdown report on T3 Code 0.0.40 with about 5,000 active threads.

Fix

• Both functions now return the built snapshot directly, checked with satisfies OrchestrationShellSnapshot. getCommandReadModel in the same file does the same.
• Removed the unused decodeShellSnapshot const.

What stays the same:

• All decoding defaults are already set in the built object, and satisfies makes sure of that.
• The JSON columns (model selection, title state, PR data, scripts, icons) are already decoded by the row schemas.
• The wire output does not change. HTTP and RPC encode the snapshot against OrchestrationShellSnapshot, and that encode trims and validates.

What changes, only for bad rows:

• Server-side callers (reactors, storage cleanup) now get plain row strings, not trimmed ones. getThreadShellById and getProjectShellById already return shells this way.
• A row that the shell schema rejects no longer fails the whole snapshot inside the server. It fails on the way to the client instead, the same as the per-thread shell updates. Background callers no longer fail on it.

This does not make the snapshot smaller. #13657 (auto-archive settled threads) is the lever for that.

Verification

• vp test run apps/server/src/orchestration/Layers/ProjectionSnapshotQuery.test.ts (33 passed)
• vp test run apps/server/src/server.test.ts -t shell (10 passed)
• vp lint and vp fmt --check on the changed file
• vp run --filter t3 typecheck

Made by Claude Opus 5.5 (1M context) in Claude Code, running in T3 Code.

🤖 Generated with Claude Code

Summary by CodeRabbit

• Refactor
  • No user-facing behavior changes. SQL error handling remains unchanged.

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.5 KiB	13.5 KiB	+24 B (+0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.1 KiB	7.1 KiB	−1 B (−0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.5 KiB	+25 B (+0.4%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	56.2 KiB	56.3 KiB	+44 B (+0.1%)	66.4 KiB	✅
Codex	Live turn messages	9	10	+1 (+11.1%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+12 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.1 KiB	7.1 KiB	0 B (0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.4 KiB	+12 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.0 KiB	57.0 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	9	9	0 (0.0%)	21	✅

Baseline: 6530de0 · PR result: dbb6585 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 114.0 KiB
• Claude decoded thread snapshot: 114.7 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Approved at ce5d5ee

Macroscope's review found this PR approvable — This is a focused server-side performance refactor that removes redundant snapshot decoding and allocation while preserving the existing transport-level contract encoding. It changes one query implementation file and introduces no new capability, product default, schema, or deployment behavior.

No code changes detected at dbb6585. Prior analysis still applies.

^{You can add or adjust custom eligibility rules. Learn more.}

[coderabbitai]

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository: pingdotgg/t3code/.coderabbit.yaml

Review profile: CHILL

Plan: Team

Run ID: 68ce2e41-96c2-4416-9601-f5dcafffe338

📥 Commits

Reviewing files that changed from the base of the PR and between 7f2c22634e83589290da8fa74e53cdf471de4ab4 and dbb6585.

📒 Files selected for processing (1)

• apps/server/src/orchestration/Layers/ProjectionSnapshotQuery.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 5 remain after this review.

---

📝 Walkthrough

Walkthrough

getShellSnapshot and getArchivedShellSnapshot now return assembled snapshots checked against OrchestrationShellSnapshot at compile time. Runtime decoding and decode-error mapping are removed. SQL error handling remains.

Changes

Shell snapshot queries

Layer / File(s)	Summary
Assemble and return shell snapshots apps/server/src/orchestration/Layers/ProjectionSnapshotQuery.ts	Both query methods return assembled snapshots directly and check the result against OrchestrationShellSnapshot at compile time. The runtime decoder and its persistence decode-error paths are removed. The import is now type-only.

Priority: ➖ Normal

Estimated code review effort: 2 (Simple) | ~10 minutes

Change: Refactor

Suggested reviewers: juliusmarminge

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the performance change: avoiding a second decode when building the thread list snapshot.
Description check	✅ Passed	The description clearly explains the change, motivation, behavior differences, verification steps, and non-goals. It does not use the template headings exactly or include the checklist, but it provide…
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1…
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

---

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

[macroscopeapp]

Removing the decode changes the service's returned values for rows whose titles contain surrounding whitespace: row titles use Schema.String, while the shell contract's TrimmedNonEmptyString previously trimmed them (and rejected blank titles). Could you add focused tests for both snapshot methods with such persisted rows and preserve the intended normalization/validation at the appropriate boundary? The existing tests only use already-trimmed titles.

^{Posted via Macroscope — Effect Service Conventions}

[t3dotgg]

Persisted titles are always trimmed and non-empty. The projector only writes title from event payloads (apps/server/src/orchestration/Layers/ProjectionPipeline.ts:506, 529, 616, 825), and those payloads decode it with TrimmedNonEmptyString (packages/contracts/src/orchestration.ts:1733, 1747, 1767, 1856). The wire path still encodes against OrchestrationShellSnapshot, and that encode also trims (packages/contracts/src/baseSchemas.ts:11), so tests for untrimmed rows would cover data the server cannot write.