[juliusmarminge]

Problem

Adding six empty methods to the 198-method WsRpcGroup makes the server typecheck fail with TS2589 and TS377030: layer requirements degrade to any. The same probe passes without those methods.

Change

Register the existing handlers through two disjoint RPC groups and merge their contexts. This reduces Effect's handler-requirements type expansion while preserving handler types, middleware, and the complete server RPC group.

Scope and approval

Fixes https://git.cafe/pingdotgg/t3code/issues/22364

Triaged bug via the automated issue flow: #22364.

Verification

• Server typecheck: passes normally and with six (204 total) and 60 (258 total) additional probe methods; before the fix, 204 fails with 22 errors.
• Post-merge 204-method typecheck: exit 0, zero errors (captured summary).
• Existing WebSocket tests: 6 passed. Targeted lint and formatting passed with no new findings.
• Real isolated dev server: authenticated server.getSettings and scheduledTasks.list both returned success on the same WebSocket, which closed cleanly with code 1000. This exercises both handler partitions.
• No render paths changed; visual fixture matrix is empty. No platform-specific surface is unverified.
• Headroom remains finite: the 318-method stress probe still fails. The transcript records that result and distinguishes captured diagnostics from reported compiler exit codes.

Compiler and runtime evidence · compiler probe · runtime probe. Evidence compares 044f07257c with ba593c8ca3; current main was subsequently merged and scoped checks rerun.

Made with Claude Opus 5.5 through Claude Code and GPT-6.1-Sol through Codex in T3 Code.

Verification log

Round 2, 2026-10-22. Reviewed head de44c6e1b7, merge-base 22840ab943. No round-1 verification log was present.

• Reviewer A — Codex gpt-6.1-sol: VERDICT: MERGE. Review transcript.
• Reviewer B — claudeAgent claude-opus-5-5: VERDICT: MERGE. Review transcript. Claude started successfully; no fallback.
• Both reviewed the full committed diff and found no blockers: 198 RPCs are covered exactly once, in 104/94 partitions, with authorization and instrumentation preserved; the full group still drives dispatch. Picks retain precise handler types. Context merging adds bounded linear work during WebSocket layer construction, with no added per-request work. No materially smaller correct shape was identified. The direct count supersedes reviewer B's approximate count; its opening “costs nothing” means no per-request work, not zero construction cost.
• Proof audit — VERDICT: SUFFICIENT. Fresh +6 probe at head: vp run typecheck scoped to apps/server, exit 0, zero errors. Same installed worktree and probe with only merge-base ws.ts: exit 2, 20 TS377030 requirement-channel errors. This rerun emitted no TS2589 diagnostic; the body's 22-error result is historical, not the count observed here. agent-job was absent. Probe files were restored to head.
• Live isolated server at head, seeding disabled (T3CODE_DEV_SEED=0, --no-seed), fresh --home-dir /tmp/ev-23813/home: server.getSettings and scheduledTasks.list both succeeded on one WebSocket. With only orchestration:read, server.updateSettings {patch:{}} was rejected with EnvironmentAuthorizationError, missing settings:write. Clean close 1000. No real T3 database was read, copied, opened or seeded.
• The +60/318-method probes and six WebSocket tests remain historical evidence; this round did not rerun them. Finite headroom remains disclosed. Server port-owner PID was recorded and cwd-checked before shutdown; the owned runner was stopped. Temporary state and the review worktree are removed as the final cleanup step.

Audit transcript · Full sanitized proof bundle. Compiler wrapper exits, full compiler logs, partition counts, runtime probe/results, and both review transcripts are included. No credentials or databases are uploaded.

Model/harness: proof audit and synthesis by GPT-6.1-Sol through Codex in T3 Code; independent reviews by GPT-6.1-Sol through Codex and Claude Opus 5.5 through claudeAgent.

[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	5.0 KiB	5.0 KiB	−7 B (−0.1%)	6.8 KiB	✅
Codex	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.2 KiB	1.2 KiB	−7 B (−0.6%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	20.9 KiB	20.9 KiB	0 B (0.0%)	29.3 KiB	✅
Codex	Live turn messages	2	2	0 (0.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.0 KiB	+7 B (+0.1%)	6.8 KiB	✅
Claude	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.2 KiB	1.2 KiB	+7 B (+0.6%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	21.2 KiB	21.2 KiB	0 B (0.0%)	29.3 KiB	✅
Claude	Live turn messages	2	2	0 (0.0%)	8	✅

Baseline: 22840ab · PR result: de44c6e · 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: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

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

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

• Source mode: Auto
• Repository: pingdotgg/t3code
• Branch: fix/issue-22364
• File: .coderabbit.config.ts
• Commit: de44c6e
• Layers: Repository TS config
• Review profile: CHILL
• Plan: Team
• Run ID: 3df72584-a43a-4b09-a7e9-dff3ed39810e

📥 Commits

Reviewing files that changed from the base of the PR and between 22840ab and de44c6e.

📒 Files selected for processing: 1

• apps/server/src/ws.ts

Limit details: You’ve used all 10 included reviews currently available.

📝 Walkthrough

The WebSocket server now divides RPC requests into two groups based on tag prefixes. It converts each handler subset with its corresponding RPC group and merges the resulting contexts.

Changes: WebSocket RPC partitioning

Layer / File(s)	Summary
Partition RPC handlers and merge contexts apps/server/src/ws.ts	The server partitions tags with the server, pullRequests, provider, and orchestration prefixes from all other tags. It converts each handler subset through its RPC group and merges the resulting contexts.

Priority: ➖ Normal

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

Change: Bug fix · Severity of issue fixed: Low

Suggested reviewers: t3dotgg

Merge Risk: ⚪ Minimal · up to de44c

No actionable merge-blocking risk was found in the RPC partitioning change.

Pre-merge checks: 3 · 1

Failed checks: 1 warning

Check name	Status	Explanation	Resolution
Description check		The description clearly covers the problem, change, scope, verification results, limitations, evidence, and agent usage. The Scope and approval section links the triaged issue but does not include exp…	Add the maintainer approval comment that confirms the fix direction and scope, or explain why this focused fix qualifies for the obvious-bug exemption without prior approval.

Passed checks: 3 passed

Check name	Status	Explanation
Title check		The title clearly identifies the server fix and its purpose: preserving typed requirements as the RPC group grows.
Linked Issues check		Issue #22364 asks that the server typecheck keep layer requirements typed as WsRpcGroup grows past about 200 methods. The issue lists group splitting as a possible direction, not an acceptance criteri…
Out of Scope Changes check		The whole-PR diff against the merge base changes only apps/server/src/ws.ts, as shown by the per-file diff and the raw summary. The additions are an effect/Context import, an RpcGroup import, the pref…

Full details: Linked Issues check

Explanation

Issue #22364 asks that the server typecheck keep layer requirements typed as WsRpcGroup grows past about 200 methods. The issue lists group splitting as a possible direction, not an acceptance criterion. The whole-PR diff (mergeBase 22840ab to head de44c6e) changes apps/server/src/ws.ts only. It partitions ServerWsRpcGroup.requests into two disjoint sub-groups: tags starting with server, pullRequests, provider, or orchestration, and all other tags. layerWsRpc now builds each subset with toHandlers and merges the resulting contexts with Context.merge, instead of one ServerWsRpcGroup.toLayer call. The two predicates are complementary, so every RPC keeps a handler. The handlers object is type-checked against each Pick, so handler types are kept. ServerWsRpcGroup, which includes the RpcInstrumentation middleware, is still the source of each Rpc. This addresses the reduced type-expansion cost described in the issue. The typecheck results (204 and 258 probe methods passing, 318 still failing) and the runtime WebSocket check are author-reported through external links. I did not run them. No automated test was added to guard group size, but the issue does not require one.

Full details: Out of Scope Changes check

Explanation

The whole-PR diff against the merge base changes only apps/server/src/ws.ts, as shown by the per-file diff and the raw summary. The additions are an effect/Context import, an RpcGroup import, the prefix-based split groups with their type guards, and the replacement of the single toLayer call with the two-group handler merge. Each change supports the typecheck fix for #22364. The PR description's references to render paths, visual fixtures, and tooling notes do not map to code changes in this diff. I did not inspect the full list of changed files beyond the diff for ws.ts. The raw summary lists no other files.

Full details: Description check

Explanation

The description clearly covers the problem, change, scope, verification results, limitations, evidence, and agent usage. The Scope and approval section links the triaged issue but does not include explicit maintainer approval of the direction and scope or explain an exemption.

Resolution

Add the maintainer approval comment that confirms the fix direction and scope, or explain why this focused fix qualifies for the obvious-bug exemption without prior approval.

✨ Finishing Touches

📝 Generate docstrings

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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

[macroscopeapp]

Approvability

Verdict: Approved at de44c6e

Macroscope's review found this PR approvable — This is a narrowly scoped server type-system fix that splits handler-context construction while retaining the complete existing RPC group, middleware, and dispatch behavior. It introduces no product-default changes, diagnostic suppressions, schema changes, or new user-facing capability.

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