3 Commits

Include renames

Author	SHA1	Message	Date
4 people	9f5bc3264a	feat(tui): warm pane pool — single-use pre-booted panes, opt-in via OCP_TUI_POOL_SIZE (−41%) (#158) ... * feat(tui): warm pane pool — single-use pre-booted panes, opt-in via OCP_TUI_POOL_SIZE Backlog item #3 of docs/plans/2026-07-13-tui-latency/README.md. Every TUI request currently cold-boots a tmux+claude pane. This adds an OPT-IN pool of pre-booted panes. Recorded as ADR 0008 (docs/adr/0008-tui-warm-pane-pool.md), which extends ADR 0007. MEASURED (this host, Sonnet 4.6, --effort low, through a real OCP instance; a sample counts only if HTTP 200 AND the body carries the demanded marker): pool off (main code) n= 6 p50 10.17s [9164 9499 9760 10572 10774 11281] pool on, warm hits n=12 p50 6.00s [5286 5289 5520 5584 5621 5969 6040 6098 6280 7846 8036 11053] pool on, warm hits (post- n= 6 p50 5.62s [4729 4753 5236 6004 7548 9548] review-fix re-run) -> -4.17s / -41%. 12 hits / 1 miss / 0 bootFailures over 13 requests (and 6/1/0 on the post-fix re-run). Robust to counting the miss: n=13 p50 -> -40.6%. The plan doc predicted only -1.0s (the boot). It is ~4.2s because the cold path also pays ~2.9s INSIDE the first turn beyond claude's own reported turn_duration — post- input-bar init that an idle pane has already finished. Phase decomposition of the cold path (n=6 medians): prep 2ms | tmux spawn 27ms | boot->input-ready 1232ms | paste 8ms | paste-verify 426ms | submit->terminal 8458ms | teardown 8ms = 10162ms total, vs native turn_duration 5539ms => 4490ms of OCP-side overhead, of which the pool recovers ~1.26s of boot and ~2.9s of in-claude cold start. (The 426ms paste-verify is one 400ms poll tick; a real paste lands in ~80ms. Not addressed here — separate item.) DESIGN - SINGLE-USE panes. A pooled pane serves exactly ONE turn, then is killed and replaced in the background. Each carries its OWN fresh --session-id fixed at boot, so one session still holds one exchange. This is what keeps transcript.mjs's extractLatestAssistantText correct; its warning about a future warm pool reusing a session is answered in-place (comment updated) and left standing for anyone who later wants a second turn on a pane — that would be a cross-request TEXT LEAK and needs user-line scoping in the transcript reader first. - Pool keyed by model; --model is fixed at spawn. A miss falls back to the cold path with zero behaviour change. The pool warms the most recently requested model, so the first request after start (and after a model switch) is always a cold miss. - REAPER COEXISTENCE (the crux). An idle warm pane IS ours, and the periodic sweep runs precisely when we are idle. reapStaleTuiSessions() takes a `spare` set of EXACT live session names, and server.mjs DRAINS the pool immediately before the sweep: 1. a live pooled pane is never reaped — INCLUDING one still BOOTING (see below); 2. an orphaned pooled pane IS still reaped — membership is by exact name from a live in-memory registry, never by name shape, so a pane from a dead process generation has nothing claiming it. Omitting `spare` reaps MORE, never less (fail-safe); 3. kill-server is suppressed while any pane is spared — hence the drain, so the sweep still flushes <defunct> claude zombies (the only mechanism that can). - THE POOL TRACKS ITS IN-FLIGHT BOOT BY NAME, NOT AS A COUNT. bootTuiPane creates the tmux session SYNCHRONOUSLY and only then waits up to POOL_BOOT_MS (20s) for the input bar, so a pooled session can be LIVE for ~20s before its boot resolves. Tracking boots as a count meant the pool could not name that session, which caused two real bugs (found in review, reproduced, fixed, and now regression-tested): * the reap sweep KILLED the booting pane (it could not be spared), left the pool empty with nothing scheduled, and logged the exact tui_pool_boot_failed WARN operators are told to alert on — for a completely healthy drain; * graceful shutdown ORPHANED a live authenticated idle `claude`: gracefulShutdown calls process.exit(0) in the SAME TICK as the drain (TUI panes are tmux children, so activeProcesses is empty and the wait-for-children path exits immediately), so cleanup deferred to a .then() never ran. Fix: the pool mints each pane's identity up front ({sessionId, name}) and holds it in _bootingPane. liveNames() includes it; drain() kills it SYNCHRONOUSLY. A generation counter distinguishes "cancelled by us" from "genuinely failed", so a drain never inflates bootFailures and resume() reliably starts a fresh boot. Deriving the name from the session-id also makes `tmux ls` correlate to the transcript file. - SLOT ACCOUNTING. Refill boots take NO TuiSemaphore slot (those bound real turns and would be starved); they cannot leak one either, since they never hold one. Refills are SERIALIZED, one boot at a time — live at size=2, two cold boots racing an in-flight turn overran the readiness cap and a refill was discarded. A genuinely failed boot does not re-kick the chain (backoff; a broken claude must not respawn forever). Background boots get a more generous readiness cap (POOL_BOOT_MS = 5x BOOT_MS): BOOT_MS is tight because a client is blocked on it, which is not true of a pre-boot. - BOUNDED COST. A warm pane is a LIVE idle claude process held whether or not a request arrives. Peak processes = pool size + OCP_TUI_MAX_CONCURRENT + 1 booting replacement. Size clamped to POOL_MAX_SIZE=4; garbage values disable rather than guess. Panes have a 10-min TTL and a health check at hand-out (dead/degraded pane => miss, never a hang). Missing collaborators throw at CONSTRUCTION, not on a live request (refill() is called synchronously from the request path). DEFAULT OFF (OCP_TUI_POOL_SIZE=0). This is a stable production path and the pool holds standing processes, so the operator opts in. With the pool off, runTuiTurn takes the IDENTICAL code path as before (the `pool ? pool.acquire() : null` branch yields null, and tuiPool is null so no observer is attached and no new log line is emitted) — that is what establishes the default path is unchanged. A pool-off control run (n=6, p50 9.40s) is consistent with the 10.17s baseline but had 2/6 samples >12s, so it is corroboration, NOT proof: n=6 cannot establish "unregressed" on its own. The code-path equivalence can. BANNER: NO SPAWN ARGUMENT CHANGED. buildTuiCmd is byte-identical to main (verified by extracting the function body from both revisions and comparing). Live banner captured from two real POOLED panes anyway: "Sonnet 4.6 with low effort · Claude Max" — the subscription pool, never "API Usage Billing". /health: `tui.pool` added (null when off), incl. `cancelled` (boots WE killed — not a fault; do not alert on it). The tui block is ADR-0007-owned and post-dates ADR 0006's v3.16.4 grandfather snapshot; the addition is purely additive — every pre-existing key keeps a byte-identical value. Authorization recorded in ADR 0008. ALIGNMENT: Class B / ADR 0007 + ADR 0008 (OCP-owned TUI spawn machinery). cli.js does NOT perform this operation — there is no cli.js citation and none is required: this is not an Anthropic API surface, it is OCP's own process management around the claude CLI, exactly as the existing tmux session lifecycle and reaper already are (ALIGNMENT.md Rule 2). TESTS: 294 passed / 0 failed (was 267). +27 covering acquire/hit/miss, single-use (a pane is never handed out twice), bounded + serialized refill, TTL + health-check drops, model retarget, drain/resume, boot-failure backoff, identity linkage, all three reaper invariants incl. post-drain kill-server restoration, and — the coverage gap that let both bugs ship — FIVE mid-boot tests: the booting pane is nameable/spareable, the sweep's drain kills it and resume starts a fresh boot with no bogus WARN, shutdown kills it synchronously (asserted WITHOUT awaiting, since process.exit runs in the same tick), a stale settle cannot clear a newer boot's slot, and a model switch cancels an in-flight boot for the old model. Live verification (temporary 20s reap interval, reverted): sweep drained both panes -> reaped -> refilled with NEW panes; a foreign tmux session survived untouched; with no foreign session kill-server fired and the pool still recovered and served the next request. Both review bugs reproduced against a PRIVATE tmux server (-L pr3repro, so the reaper's internal kill-server could not touch the host) before and after the fix. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(tui): kill a cancelled boot's pane when it settles + make async tests actually count Folds in the independent review's remaining nit — and, in proving the nit's fix, uncovers two defects in the test suite itself. ## The nit (latent M1b, second costume) `_cancelBooting` kills BY NAME, but the tmux session only EXISTS once `bootPane` has run — and `bootPane` is queued on a microtask. So a caller doing `refill()` then `drain()` in the SAME synchronous block leaves `_cancelBooting` with nothing to kill (a no-op); it bumps the generation, and the boot microtask then CREATES the session, succeeds, and — under the old bare `return` on a stale generation — walked away from a LIVE authenticated `claude` that nothing owns. Reproduced: reverted: drain() kills nothing (no session yet) -> boot creates it -> ORPHAN: ['p1'] fixed : drain() kills nothing (no session yet) -> boot creates it -> boot kills it -> [] Not reachable from any current call site, so this is defense-in-depth — but ADR 0008 and the reap-tick comment in server.mjs BOTH explicitly contemplate a boot-time pre-warm, which is exactly the shape that reaches it. Killing an already-dead session is a harmless no-op, so the fix is idempotent whichever way the race lands. ## Defect 1 in the suite: async tests were never awaited (44 of them) Writing the regression guard exposed this. `test()` called `fn()`, got a promise back, and IMMEDIATELY printed ✓ and incremented `passed` — without awaiting it. For all 44 tests written as `test("...", async () => {...})`: - ✓ meant "did not throw SYNCHRONOUSLY", not "passed"; - a failed assertion escaped as an unhandled rejection, crashing the process (CI stays red on the non-zero exit) but never being COUNTED — so the summary could print "0 failed" and be wrong. The suite's headline number was therefore not evidence for ANY async test, including this PR's own M1a/M1b guards. `test()` now settles an async body before counting it, and the summary awaits them. ## Defect 2, exposed the instant defect 1 was fixed: a false guard `"a boot that resolves AFTER a drain kills its own pane ... no orphan process left behind"` asserted `killed.length === 1` — i.e. that kill was CALLED once. But `_cancelBooting`'s kill-by-name on a not-yet-existent session is a NO-OP that still increments that counter. So "kill was called once" and "a live session is orphaned" were both true at the same time: a test named for the absence of an orphan was passing while the orphan was present. Now asserts LIVENESS (`live.size === 0`) — the only honest question. ## Evidence fix present : 295 passed, 0 failed, exit 0 fix reverted: 293 passed, 2 failed <- BOTH liveness guards fire (the old kill-count guard did not) Also: `dropped`'s doc comment now lists `cancelled` (a cancelled in-flight boot lands there via _drop). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VqgWJcjxrjjL9L9SkpZyXR --------- Co-authored-by: dtzp555 <dtzp555@gmail.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-07-13 16:51:12 +10:00
4 people	d96da46fa0	fix: honor runtime-lowered concurrency limit + cancel queued waiters on disconnect (#149) ... * fix: honor runtime-lowered concurrency limit + cancel queued waiters on disconnect Fixes three findings from an independent concurrency audit of the -p/stream-json wait-queue (lib/tui/semaphore.mjs, reused by server.mjs as `claudeSemaphore`) and its acquireClaudeSlot()/callClaudeTui() callers in server.mjs: F1 (MEDIUM) — release() handed a freed slot straight to the next queued waiter without re-checking `this.limit`, so a PATCH /settings maxConcurrent decrease was silently ignored until every already-inflight task happened to finish on its own. release() now only re-grants when post-decrement inflight is still under the current limit, and a new setLimit() wakes queued waiters immediately when the limit is raised instead of only on the next incidental release(). F2 (MEDIUM) — a request queued behind the concurrency limit had no link to its HTTP connection, so a client that disconnected while still queued would still get a claude process spawned for it once a slot freed — burning subscription quota for a dead socket. acquire() now accepts an optional AbortSignal; server.mjs derives one from the client's res "close" event (closeSignalFor) and passes it into claudeSemaphore.acquire() / tuiSemaphore.acquire() while queued. On abort the waiter is spliced out of the queue (not just flagged), so `queued` accounting stays exact; the same "close" signal is wired into acquireClaudeSlot() (-p path, non-streaming + streaming + singleflight-wrapped) and callClaudeTui() (TUI path). If the response is already destroyed by the time we try to queue, we reject immediately without ever entering the queue. F8 (cosmetic) — acquireClaudeSlot() set `stats.queued = claudeSemaphore.queued + 1` BEFORE calling acquire(), over-reporting /health's queued count by 1 whenever a slot was granted immediately (the common, non-queued case). acquire() already updates its internal queue synchronously before returning a Promise, so reading claudeSemaphore.queued right AFTER calling it (instead of guessing "+1" before) is exact. No /health field was added, removed, or renamed. ALIGNMENT.md: this PR touches request-handler code (callClaude, callClaudeStreaming, callClaudeTui, acquireClaudeSlot) but is local concurrency-control/queue-accounting infrastructure with no cli.js wire analogue — it does not add, rename, or change any endpoint, header, request field, or response field, and does not touch the /v1/messages forwarding path or the OAuth bearer machinery (the two Class A surfaces this repo governs). The /health response shape is unchanged (same field set, same nesting; only the *value* of the pre-existing `stats.queued` field is corrected). Per CLAUDE.md hard-requirement #1, a cli.js citation is therefore declared ABSENT: there is no corresponding cli.js operation to cite because this is not a cli.js-mirror (Class A) change and not a Class B endpoint-contract change either. Tests: added 6 unit tests to test-features.mjs against the shared TuiSemaphore (lowering the limit mid-load does not over-admit; raising the limit wakes queued waiters up to the new headroom, FIFO; a queued waiter cancelled via AbortSignal is spliced out and never later acquires; an already-aborted signal never touches the queue; cancelling one of several queued waiters preserves FIFO for the rest). 238 pre-existing tests remain green; suite is now 244/244. Verification: `node --check server.mjs && node --check lib/tui/semaphore.mjs && npm test` — 244 passed, 0 failed. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: singleflight follower retry on leader disconnect + quiet disconnect handling (review M1/L1/L2) Addresses the independent reviewer's APPROVE-WITH-CHANGES findings on PR #149: M1 (MEDIUM, F2 regression) — when a singleflight LEADER disconnected while queued, its RequestDisconnectedError rejected the SHARED promise, so live followers fell into respondUpstreamError's generic branch and got a spurious 500 on a healthy socket. Fix: keys.mjs singleflight() gains an optional follower-side `retryIf` predicate. When a follower joins an existing flight and the shared promise rejects with an error retryIf() accepts, the follower does NOT inherit the rejection — it re-enters singleflight with its OWN fn (the map entry is guaranteed already deleted: the delete-finally is attached upstream of the promise followers await), becoming the new leader or joining a retrying sibling's fresh flight. The leader's own rejection is never retried (it IS that client's disconnect). server.mjs passes retryIf = (err) => err instanceof RequestDisconnectedError && !res.destroyed, so a follower whose own client is also gone still propagates quietly. Callers without retryIf keep byte-for-byte pre-existing share-everything semantics (pinned by the existing failure-fan-out test). L1 (LOW) — a disconnect-while-queued on the non-streaming paths was recorded as a usage FAILURE row and logged as a [proxy] error: metric noise for a non-error. Both non-streaming catch blocks now early-return on RequestDisconnectedError without recordUsage(success:false) and without console.error — mirroring the streaming path, which returns silently. The disconnect remains observable at info level (concurrency_wait_cancelled, now also emitted with path:"tui" from callClaudeTui for parity with acquireClaudeSlot's -p log). L2 (LOW, test gap) — added a unit test for the abort-after-grant race: a waiter granted its slot whose signal aborts afterward must see no rejection, no queue corruption, and its slot released exactly once via the normal path (the semaphore detaches the abort listener at grant; the onAbort idx===-1 guard is the in-dispatch backstop). Tests: +3 (2× M1 in the singleflight section, 1× L2 in the F2 section) — suite is now 247/247 green. ALIGNMENT.md: unchanged declaration — still local concurrency/dedup infrastructure with no cli.js wire analogue; no endpoint, header, request field, or response field added or changed; /health shape untouched. cli.js citation declared ABSENT per CLAUDE.md hard-requirement #1 (not a Class A mirror change, not a Class B contract change). Verification: node --check server.mjs lib/tui/semaphore.mjs keys.mjs && npm test — 247 passed, 0 failed. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: dtzp555 <dtzp555@gmail.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-07-07 23:04:13 +10:00
4 people	3322d7bdae	feat(tui): per-path concurrency limit + /health observability (PR-B) (#139) ... Two P1 audit fixes for the TUI subscription-pool bridge (ADR 0007). The default path (CLAUDE_TUI_MODE unset) is unchanged except the additive /health `tui` block (enabled:false when off). C-4 — TUI path had NO concurrency bound. The global MAX_CONCURRENT gate lives in spawnClaudeProcess (the -p/stream-json path); callClaudeTui never calls it — it calls runTuiTurn, which cold-boots a full interactive claude in tmux. So N concurrent TUI requests spawned N simultaneous cold boots (a family burst of 5 on a Pi 4 = OOM risk + subscription rate-limit pressure). Adds an independent in-process limiter (lib/tui/semaphore.mjs, TuiSemaphore) gating callClaudeTui: OCP_TUI_MAX_CONCURRENT (default 2 — a TUI turn is heavy: per-request cold-boot + up to 120s wallclock). Queues rather than rejects (mirrors MAX_CONCURRENT intent), with a bounded wait queue (default 32x the limit) → tui_queue_full (503) on overflow rather than unbounded growth. The slot releases in a finally, so PR-A's honesty-gate throws / timeouts / paste failures never leak a slot. C-5 — no operator-visible TUI drift surface. The tui_entrypoint_mismatch warning only reached journald; after the 6/15 flip a silent sdk-cli drift (the documented top risk) would drain metered credits invisibly. Adds an additive `tui` block to /health: { enabled, entrypointMode, lastEntrypoint, entrypointMismatches, inflight, queued, maxConcurrent }. lastEntrypoint / entrypointMismatches are recorded in callClaudeTui (same mismatch branch the journald warning covers); inflight/queued come from the C-4 semaphore. ALIGNMENT (Class B): cli.js does NOT perform this operation — both the TUI path and /health are OCP-owned, so no cli.js citation applies. /health is a grandfathered B.2 endpoint (ADR 0006, frozen at v3.16.4). The response-shape change is authorized by the ADR 0007 PR-B amendment added in this commit and is behaviour-preserving: the `tui` block is NEW fields only — no existing /health field is changed, renamed, removed, or re-typed, and no existing semantics change, so existing consumers (dashboard, ocp-connect, monitoring) are unaffected. Per ALIGNMENT.md's grandfather provision, an additive behaviour-preserving change to a grandfathered B.2 endpoint is authorized by an ADR; ADR 0007 is the authority for the TUI observability surface. No Class A forwarding path, no alignment.yml, no models.json touched — alignment blacklist is unaffected (zero new network tokens). Tests: 11 new (lib/tui/semaphore.mjs is importable, so the semaphore + the two pure /health helpers are tested directly): limit=1 serializes two overlapping calls; limit=2 runs two + queues the third; slot released on throw; bounded queue → tui_queue_full; mismatch counter increments on cli→other drift; auto mode never counts a mismatch; /health tui block shape + live inflight/queued. npm test: 235 passed, 0 failed (was 224). Co-authored-by: dtzp555 <dtzp555@gmail.com> Co-authored-by: Claude <claude-opus> <noreply@anthropic.com>	2026-06-10 21:54:07 +10:00