@openparachute/agent 0.1.2 → 0.2.0

package/docs/audit/2026-04-30-channel-endpoint-audit.md

@@ -1,36 +0,0 @@
-# Channel endpoint audit (paraclaw#67, Phase 2 prep)
-
-**Date:** 2026-04-30
-**Scope:** what server endpoints exist for channel install/test/wire on `main` vs. what the design doc and the abandoned `feat/setup-wizard-discord` branch assumed.
-
-The Phase 1 design doc flagged `/api/setup/install-channel` and `/api/channels/{adapter}/test` as "unclear-but-noted." This audit resolves them before impl begins so we don't write code that duplicates or contradicts existing surface.
-
-## Findings
-
-### 1. The validate/test handlers do not exist on `main`
-
-`POST /api/channels/discord/test` and `POST /api/channels/telegram/test` are referenced by the wizard's `WireChannelStep` only via the wizard's *idempotent shortcut* (which assumes the adapter is already installed). On main, `src/web/server.ts` does not register either route. The handlers were written on `feat/setup-wizard-discord` under the old `web/server/src/` layout (pre-reorg) and were never ported.
-
-The validator modules themselves (`discord-validate.ts`, `telegram-validate.ts`) and their tests are clean, well-typed, and portable. They will be lifted verbatim into `src/web/` as Phase 2 commit 2.
-
-### 2. The install-channel orchestrator is obsolete
-
-The 341-line `install-channel.ts` orchestrator on `feat/setup-wizard-discord` shells out to `.claude/skills/add-discord` / `add-telegram` to materialize an adapter. As of `main`:
-
-- `src/channels/discord.ts` and `src/channels/telegram.ts` are **trunk-baked** — present at install time, no skill installation needed.
-- `.claude/skills/add-discord` and `.claude/skills/add-telegram` no longer exist.
-- The setup wizard's "install channel" step is a no-op for both, masked by an idempotent shortcut.
-
-Implication for Phase 2: we do **not** need to port install-channel. The new `/channels/new` page can skip the install step entirely for trunk-baked adapters. Slack/WhatsApp/Teams remain "coming soon" and disabled in `ChannelPickStep.tsx`; their install path will be designed when those ship.
-
-### 3. Open question O1 ("reuse wizard install step") doesn't bind
-
-Aaron approved O1 in design review, but on the ground there is nothing to reuse: discord/telegram are already installed when the user lands, and slack/whatsapp aren't in scope. The new `/channels/new` page should hide the install section entirely for the supported adapters and surface a "coming soon" affordance for the rest.
-
-### 4. CLAUDE.md is stale on adapter packaging
-
-The repo-root `CLAUDE.md` states "trunk does not ship any specific channel adapter." That was true pre-reorg; it is not true today. Discord and Telegram are baked into the trunk install. Not fixed in this audit (out of scope), but flagging for a cleanup pass.
-
-## Phase 2 scope (narrowed)
-
-Result of the audit: Phase 2 is **server-light** (validators + 2 routes, ~395 LOC including tests, no DB migration) and **UI-bulky** (new page, descriptor table, picker extraction, ChannelsList re-link, wizard refactor). Single PR, per-feature commits.

package/docs/build-and-runtime.md

@@ -1,80 +0,0 @@
-# Build & Runtime
-
-Paraclaw runs a split stack: the host is Bun, the agent container is also Bun. They communicate exclusively through two SQLite files per session — there are no shared modules between them, which is what lets them use different runtimes cleanly.
-
-## Why the split
-
-- **Host stays on Node** because Baileys (WhatsApp) depends on `libsignal-node` native bindings and a long-tested WebSocket/HTTP stack. Bun's Node-API compat has improved, but this isn't where we want risk.
-- **Container runs Bun** because `bun:sqlite` is built-in (no native compile of `better-sqlite3` per image rebuild), source runs directly (no tsc build step at image build or session wake), and `bun install` is ~5-10× faster than `npm install`.
-
-Host and container each have their own package tree:
-
-```
-/                             pnpm + Node 22
-  pnpm-lock.yaml              host deps (channels, Chat SDK, Baileys, better-sqlite3, etc.)
-  pnpm-workspace.yaml         minimumReleaseAge + onlyBuiltDependencies policy
-
-/container/agent-runner/      Bun 1.3+
-  bun.lock                    agent-runner runtime deps (Claude Agent SDK, MCP SDK, zod, etc.)
-  package.json                @types/bun, typescript devDeps for type-checking
-```
-
-The container image also has pnpm + Node inside for global CLIs (`@anthropic-ai/claude-code`, `agent-browser`, `vercel`). Those are Node binaries the agent invokes at runtime, not library deps. Keeping them on pnpm preserves the supply-chain policy for CLI versions.
-
-## Lockfiles
-
-| Tree | Lockfile | Manager | Regenerate after dep change |
-|------|----------|---------|----------------------------|
-| Host | `pnpm-lock.yaml` | pnpm 10 | `pnpm install` |
-| Agent-runner | `container/agent-runner/bun.lock` | Bun 1.3+ | `cd container/agent-runner && bun install` |
-
-Both are committed. CI and the Dockerfile run `--frozen-lockfile` variants — any drift between `package.json` and lockfile fails the build.
-
-## Supply chain
-
-- **Host + global CLIs** (pnpm): `minimumReleaseAge: 4320` (3-day hold on new versions), `onlyBuiltDependencies` allowlist for postinstall scripts. See `pnpm-workspace.yaml` and `docs/SECURITY.md`.
-- **Agent-runner** (Bun): no release-age policy — Bun doesn't have an equivalent today. The defenses are `bun.lock` pinning plus version-pinned CLIs/Bun itself via Dockerfile ARGs. When bumping `@anthropic-ai/claude-agent-sdk` or any runtime dep, review the release date on npm and bump deliberately, not via `bun update`.
-
-## Image build surface
-
-`container/Dockerfile` is a single-stage build on `node:22-slim`:
-
-- **Pinned ARGs** — `BUN_VERSION`, `CLAUDE_CODE_VERSION`, `AGENT_BROWSER_VERSION`, `VERCEL_VERSION`. Bump deliberately in PRs.
-- **CJK fonts** — `ARG INSTALL_CJK_FONTS=false`. `container/build.sh` reads `INSTALL_CJK_FONTS` from `.env` and passes it through. Default build saves ~200MB; opt in when the user works with Chinese/Japanese/Korean content.
-- **BuildKit cache mounts** — `/var/cache/apt`, `/var/lib/apt`, `/root/.bun/install/cache`, `/root/.cache/pnpm`. Rebuilds where `package.json`/`bun.lock` haven't changed are fast. Requires BuildKit (default on Docker 23+, Apple Container-compat).
-- **`tini` as init** — reaps Chromium zombies, forwards signals so in-flight `outbound.db` writes finalize on SIGTERM.
-- **`entrypoint.sh`** (extracted) — `exec bun run /app/src/index.ts` under tini. Readable and diffable.
-- **No compiled `/app/dist`** — Bun runs TS directly. The host also mounts fresh source over `/app/src` at session start, so host edits take effect without rebuilding the image.
-
-## Session wake (two paths)
-
-1. **Base image ENTRYPOINT** — used for stdin-piped test invocations like the sample in `container/build.sh`: `tini --> entrypoint.sh` captures stdin to `/tmp/input.json`, then `exec bun run src/index.ts`.
-2. **Host-spawned session** — `src/container-runner.ts` at line ~301 uses `--entrypoint bash` with `-c 'exec bun run /app/src/index.ts'`. Bypasses tini (Docker's default PID 1 handling applies). Stdin is unused; all IO flows through the mounted session DBs.
-
-Both paths end with Bun running the same source file from `/app/src/index.ts`.
-
-## CI shape
-
-`.github/workflows/ci.yml` installs both Node (with pnpm cache) and Bun, then runs in order:
-
-1. `pnpm install --frozen-lockfile` (host)
-2. `bun install --frozen-lockfile` in `container/agent-runner/` (container)
-3. `pnpm run format:check`
-4. `pnpm exec tsc --noEmit` (host typecheck)
-5. `pnpm exec tsc -p container/agent-runner/tsconfig.json --noEmit` (container typecheck)
-6. `pnpm exec vitest run` (host tests)
-7. `bun test` in `container/agent-runner/` (container tests)
-
-Any failure fails the PR.
-
-## Key invariants
-
-- **Session DBs must use `journal_mode=DELETE`.** WAL's `-shm` memory-map doesn't cross VirtioFS between host and guest. See the doc comment at the top of `container/agent-runner/src/db/connection.ts` and `src/session-manager.ts`.
-- **Named SQL parameters in the container require the prefix in JS object keys.** `bun:sqlite` does not auto-strip `@`/`$`/`:` the way `better-sqlite3` does on the host. Use `$name` in both SQL and keys: `.run({ $id: msg.id })`. Positional `?` params work normally.
-- **Agent-runner tests run under `bun:test`, not vitest.** `vitest.config.ts` excludes the `container/agent-runner/` tree because vitest runs on Node and can't load `bun:sqlite`.
-- **No tsc build step in the container image.** Re-adding one would reintroduce the ~200-500ms per-session-wake cost we removed.
-- **Global container CLIs stay on pnpm, not Bun.** `agent-browser`, `@anthropic-ai/claude-code`, `vercel` and any future Node CLIs the agent invokes should be pinned versions under the Dockerfile's pnpm global-install block. `bun install -g` would bypass the pnpm supply-chain policy.
-
-## Migration history
-
-This structure replaced a uniform npm-on-Node stack across both host and container. The pnpm migration landed first (PR #1771) to bring the host under supply-chain policy, then the container moved to Bun to eliminate native-module compilation and the per-wake tsc step. The split was chosen over going full-Bun because Baileys' native deps are the main risk surface on the host — the container has no such deps, so it benefits from Bun without taking the risk.

package/docs/cross-mount-stress/README.md

@@ -1,112 +0,0 @@
-# Cross-mount SQLite stress harness
-
-This directory exists for one reason: to **prove, with real numbers, that two
-writers contending on a single SQLite file across a container bind-mount is
-unsafe** — even with `journal_mode = DELETE` and a generous `busy_timeout`. It
-is the empirical foundation for paraclaw's load-bearing invariant:
-
-> **Each session has two SQLite files: `inbound.db` (host writes, container
-> reads) and `outbound.db` (container writes, host reads). Exactly one writer
-> per file.** See [`docs/db.md`](../db.md#1-the-three-databases) and
-> [`docs/db-session.md`](../db-session.md).
-
-If a future contributor (human or Claude) is tempted to "simplify" the model to
-a single `session.db` shared by host + container, run this harness first.
-
-## Why it exists
-
-The two-DB split has visible cost — duplicate schema work, two open handles per
-session, slightly more code on every read path. The recurring temptation is
-"can we collapse this to one DB? SQLite has busy_timeout, journal_mode=DELETE
-forces flushes across the mount, what could go wrong?"
-
-This harness was built to answer that question concretely on the actual paraclaw
-runtime stack — Bun on the container side, Node + better-sqlite3 on the host side,
-OrbStack VirtioFS as the bind-mount, on macOS Darwin. The methodology and findings
-are below; commit them so we don't have to rediscover this.
-
-## What's in here
-
-Three pairs of writers, each pair runs simultaneously on host + container:
-
-| Script | Pair | Behavior |
-|---|---|---|
-| `host-writer.mjs` / `container-writer.mjs` | basic | Insert N rows, no retry. Bursty pacing. |
-| `host-writer-slow.mjs` / `container-writer-slow.mjs` | slow | Loop on a duration deadline at ~100 writes/sec. Forces real overlap. |
-| `host-writer-retry.mjs` / `container-writer-retry.mjs` | retry | Same as slow but with 8-attempt exponential backoff per write — the production-shaped variant. |
-
-All three pairs use `journal_mode = DELETE` and `busy_timeout = 5000`, which is
-the strongest configuration the paraclaw runtime can use without resorting to
-WAL (which doesn't survive the bind mount).
-
-## How to run
-
-```bash
-# Pick a working dir — any path the host can read/write that you can also
-# bind-mount into a container.
-WORK=/tmp/paraclaw-xmount-test
-mkdir -p "$WORK"
-cp docs/cross-mount-stress/*.mjs "$WORK/"
-
-# Need better-sqlite3 reachable for the host scripts.
-( cd "$WORK" && pnpm init -y && pnpm add better-sqlite3 )
-
-# Ensure a paraclaw-agent image is built (uses bun).
-./container/build.sh
-
-# Run the slow-overlap variant: 30 seconds, both sides at once.
-( cd "$WORK" && node host-writer-slow.mjs "$WORK/session.db" host 30 ) &
-docker run --rm -v "$WORK:/workspace" \
-  paraclaw-agent:latest \
-  bun /workspace/container-writer-slow.mjs /workspace/session.db container 30
-wait
-```
-
-To run the retry variant, swap the `*-slow.mjs` filenames for `*-retry.mjs`.
-
-## What the data showed (Apr 28 2026)
-
-Run on macOS Darwin 25.3.0 + OrbStack with the paraclaw v2 agent image
-(Bun 1.3.13, bun:sqlite). Session files on a VirtioFS bind mount.
-
-**Basic (no-retry) pair, ~500 inserts each side, 30s overlap window:**
-- Container side: 2–20% of writes failed with `disk I/O error` or `unable to open database file` (the SQLITE_IOERR / SQLITE_CANTOPEN family).
-- Host side: lower error rate but still nonzero. Final rowcount short of expected on both sides.
-- `PRAGMA integrity_check` came back `ok` after the run — *the file isn't corrupted*, but writes are silently dropped.
-
-**Retry pair, 8-attempt exponential backoff, 30s overlap window:**
-- All writes eventually landed (zero permanent failures both sides).
-- Container retry rate: **~21%** of operations needed at least one retry.
-- Host retry rate: ~5–8%.
-- Steady-state throughput dropped roughly 2–3× vs. uncontended.
-- Integrity still `ok`.
-
-In other words: the file isn't corrupted, but a "single shared session.db"
-design needs every caller wrapped in a retry loop, and ~1 in 5 container writes
-will hit the slow path. The current paraclaw codebase has ~14+ writer call
-sites in `container/agent-runner/` alone — adding retry loops to every one of
-them is the cost of "simpler mental model," and the math doesn't work out.
-
-## What this means for the design
-
-- **Two-DB split is the right model.** Host writes `inbound.db`, container
-  writes `outbound.db`. Cross-mount visibility is required for *reading* — and
-  with `journal_mode = DELETE` + open-write-close, reads see committed pages
-  reliably. The dangerous case is two writers contending on the same file,
-  which the split avoids by construction.
-- **No retry wrappers on the writer paths.** Because each file has exactly one
-  writer, there is no contention to retry around.
-- **Don't switch to WAL.** WAL across this mount is a known bug farm — the
-  `-wal` and `-shm` files don't reliably propagate, and the other side can
-  read stale pages. Stick with `journal_mode = DELETE`.
-
-## When to re-run
-
-Re-run the harness if any of the following changes:
-- The container bind-mount technology (e.g., switching off VirtioFS).
-- Bun's `bun:sqlite` storage layer.
-- `better-sqlite3`'s SQLite version.
-- The host platform (different VM, different filesystem).
-
-If the new numbers say two-writer is now safe, *then* it's worth revisiting the
-single-DB design. Until then: two files, one writer each.

package/docs/cross-mount-stress/container-writer-retry.mjs

@@ -1,55 +0,0 @@
-// Container writer with retry-on-error — production-shaped. Used to measure
-// the steady-state retry rate when both sides contend on a single SQLite file
-// across the bind mount.
-import { Database } from 'bun:sqlite';
-
-const dbPath = process.argv[2] || '/workspace/session.db';
-const writerId = process.argv[3] || 'container';
-const durationSec = Number(process.argv[4] || '15');
-
-const db = new Database(dbPath);
-db.exec('PRAGMA journal_mode = DELETE');
-db.exec('PRAGMA busy_timeout = 5000');
-db.exec(`CREATE TABLE IF NOT EXISTS messages (
-  id INTEGER PRIMARY KEY AUTOINCREMENT,
-  writer TEXT NOT NULL,
-  seq INTEGER NOT NULL,
-  ts TEXT NOT NULL DEFAULT (datetime('now'))
-);`);
-
-const stmt = db.prepare('INSERT INTO messages (writer, seq) VALUES (?, ?)');
-const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
-
-async function writeWithRetry(seq, maxAttempts = 8) {
-  for (let attempt = 1; attempt <= maxAttempts; attempt++) {
-    try {
-      stmt.run(writerId, seq);
-      return { ok: true, attempts: attempt };
-    } catch (e) {
-      if (attempt === maxAttempts) return { ok: false, attempts: attempt, err: e.message };
-      await sleep(20 + Math.random() * 80);
-    }
-  }
-}
-
-const start = Date.now();
-const deadline = start + durationSec * 1000;
-let written = 0,
-  perm = 0,
-  retries = 0,
-  i = 0;
-
-while (Date.now() < deadline) {
-  i++;
-  const r = await writeWithRetry(i);
-  if (r.ok) {
-    written++;
-    if (r.attempts > 1) retries += r.attempts - 1;
-  } else perm++;
-  await sleep(5 + Math.random() * 10);
-}
-
-console.log(`[${writerId}] ${written} ok, ${perm} perm-fail, ${retries} retries`);
-console.log(`[${writerId}] integrity:`, JSON.stringify(db.prepare('PRAGMA integrity_check').get()));
-console.log(`[${writerId}] counts:`, db.prepare('SELECT writer, COUNT(*) n FROM messages GROUP BY writer').all());
-db.close();

package/docs/cross-mount-stress/container-writer-slow.mjs

@@ -1,42 +0,0 @@
-// Slowed container writer — duration-bounded loop for true cross-mount overlap.
-import { Database } from 'bun:sqlite';
-
-const dbPath = process.argv[2] || '/workspace/session.db';
-const writerId = process.argv[3] || 'container';
-const durationSec = Number(process.argv[4] || '15');
-
-const db = new Database(dbPath);
-db.exec('PRAGMA journal_mode = DELETE');
-db.exec('PRAGMA busy_timeout = 5000');
-db.exec(`CREATE TABLE IF NOT EXISTS messages (
-  id INTEGER PRIMARY KEY AUTOINCREMENT,
-  writer TEXT NOT NULL,
-  seq INTEGER NOT NULL,
-  ts TEXT NOT NULL DEFAULT (datetime('now'))
-);`);
-
-const stmt = db.prepare('INSERT INTO messages (writer, seq) VALUES (?, ?)');
-const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
-const start = Date.now();
-const deadline = start + durationSec * 1000;
-let written = 0,
-  errors = 0,
-  i = 0;
-
-while (Date.now() < deadline) {
-  i++;
-  try {
-    stmt.run(writerId, i);
-    written++;
-  } catch (e) {
-    errors++;
-    if (errors < 10) console.error(`${writerId} write err #${i}: ${e.message}`);
-  }
-  await sleep(5 + Math.random() * 10);
-}
-
-const elapsed = ((Date.now() - start) / 1000).toFixed(1);
-console.log(`[${writerId}] wrote ${written} in ${elapsed}s, ${errors} errors`);
-console.log(`[${writerId}] integrity:`, JSON.stringify(db.prepare('PRAGMA integrity_check').get()));
-console.log(`[${writerId}] counts:`, db.prepare('SELECT writer, COUNT(*) n FROM messages GROUP BY writer').all());
-db.close();

package/docs/cross-mount-stress/container-writer.mjs

@@ -1,47 +0,0 @@
-// Container-side stress writer — uses bun:sqlite, what the agent-runner runs.
-// Mirror of host-writer.mjs for the bun runtime. Run inside a container that
-// bind-mounts the same session.db the host writer is hitting.
-import { Database } from 'bun:sqlite';
-
-const dbPath = process.argv[2] || '/workspace/session.db';
-const writerId = process.argv[3] || 'container';
-const count = Number(process.argv[4] || '500');
-
-const db = new Database(dbPath);
-db.exec('PRAGMA journal_mode = DELETE');
-db.exec('PRAGMA busy_timeout = 5000');
-db.exec(`CREATE TABLE IF NOT EXISTS messages (
-  id INTEGER PRIMARY KEY AUTOINCREMENT,
-  writer TEXT NOT NULL,
-  seq INTEGER NOT NULL,
-  ts TEXT NOT NULL DEFAULT (datetime('now'))
-);`);
-
-const stmt = db.prepare('INSERT INTO messages (writer, seq) VALUES (?, ?)');
-const start = Date.now();
-let written = 0;
-let errors = 0;
-
-const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
-
-for (let i = 1; i <= count; i++) {
-  try {
-    stmt.run(writerId, i);
-    written++;
-  } catch (e) {
-    errors++;
-    if (errors < 5) console.error(`${writerId} write error #${i}: ${e.message}`);
-  }
-  if (i % 50 === 0) await sleep(2 + Math.random() * 8);
-  else if (Math.random() < 0.1) await sleep(Math.random() * 3);
-}
-
-const elapsed = ((Date.now() - start) / 1000).toFixed(1);
-console.log(`[${writerId}] wrote ${written}/${count} rows in ${elapsed}s, ${errors} errors`);
-
-const integrity = db.prepare('PRAGMA integrity_check').get();
-console.log(`[${writerId}] integrity:`, JSON.stringify(integrity));
-
-const counts = db.prepare('SELECT writer, COUNT(*) AS n FROM messages GROUP BY writer ORDER BY writer').all();
-console.log(`[${writerId}] row counts:`, counts);
-db.close();

package/docs/cross-mount-stress/host-writer-retry.mjs

@@ -1,55 +0,0 @@
-// Host writer with retry-on-busy/IO-error — the production-shaped variant.
-// Used to measure the steady-state retry rate when both sides contend on a
-// single bind-mounted SQLite file under VirtioFS.
-import Database from 'better-sqlite3';
-import { setTimeout as sleep } from 'node:timers/promises';
-
-const dbPath = process.argv[2] || '/tmp/paraclaw-xmount-test/session.db';
-const writerId = process.argv[3] || 'host';
-const durationSec = Number(process.argv[4] || '15');
-
-const db = new Database(dbPath);
-db.pragma('journal_mode = DELETE');
-db.pragma('busy_timeout = 5000');
-db.exec(`CREATE TABLE IF NOT EXISTS messages (
-  id INTEGER PRIMARY KEY AUTOINCREMENT,
-  writer TEXT NOT NULL,
-  seq INTEGER NOT NULL,
-  ts TEXT NOT NULL DEFAULT (datetime('now'))
-);`);
-
-const stmt = db.prepare('INSERT INTO messages (writer, seq) VALUES (?, ?)');
-
-async function writeWithRetry(seq, maxAttempts = 8) {
-  for (let attempt = 1; attempt <= maxAttempts; attempt++) {
-    try {
-      stmt.run(writerId, seq);
-      return { ok: true, attempts: attempt };
-    } catch (e) {
-      if (attempt === maxAttempts) return { ok: false, attempts: attempt, err: e.message };
-      await sleep(20 + Math.random() * 80);
-    }
-  }
-}
-
-const start = Date.now();
-const deadline = start + durationSec * 1000;
-let written = 0,
-  perm = 0,
-  retries = 0,
-  i = 0;
-
-while (Date.now() < deadline) {
-  i++;
-  const r = await writeWithRetry(i);
-  if (r.ok) {
-    written++;
-    if (r.attempts > 1) retries += r.attempts - 1;
-  } else perm++;
-  await sleep(5 + Math.random() * 10);
-}
-
-console.log(`[${writerId}] ${written} ok, ${perm} perm-fail, ${retries} retries`);
-console.log(`[${writerId}] integrity:`, JSON.stringify(db.prepare('PRAGMA integrity_check').get()));
-console.log(`[${writerId}] counts:`, db.prepare('SELECT writer, COUNT(*) n FROM messages GROUP BY writer').all());
-db.close();

package/docs/cross-mount-stress/host-writer-slow.mjs

@@ -1,43 +0,0 @@
-// Slowed host writer — loops on a duration deadline so it overlaps with the
-// container writer. ~100 writes/sec, 5–15ms between writes.
-import Database from 'better-sqlite3';
-import { setTimeout as sleep } from 'node:timers/promises';
-
-const dbPath = process.argv[2] || '/tmp/paraclaw-xmount-test/session.db';
-const writerId = process.argv[3] || 'host';
-const durationSec = Number(process.argv[4] || '15');
-
-const db = new Database(dbPath);
-db.pragma('journal_mode = DELETE');
-db.pragma('busy_timeout = 5000');
-db.exec(`CREATE TABLE IF NOT EXISTS messages (
-  id INTEGER PRIMARY KEY AUTOINCREMENT,
-  writer TEXT NOT NULL,
-  seq INTEGER NOT NULL,
-  ts TEXT NOT NULL DEFAULT (datetime('now'))
-);`);
-
-const stmt = db.prepare('INSERT INTO messages (writer, seq) VALUES (?, ?)');
-const start = Date.now();
-const deadline = start + durationSec * 1000;
-let written = 0;
-let errors = 0;
-let i = 0;
-
-while (Date.now() < deadline) {
-  i++;
-  try {
-    stmt.run(writerId, i);
-    written++;
-  } catch (e) {
-    errors++;
-    if (errors < 10) console.error(`${writerId} write err #${i}: ${e.message}`);
-  }
-  await sleep(5 + Math.random() * 10);
-}
-
-const elapsed = ((Date.now() - start) / 1000).toFixed(1);
-console.log(`[${writerId}] wrote ${written} in ${elapsed}s, ${errors} errors`);
-console.log(`[${writerId}] integrity:`, JSON.stringify(db.prepare('PRAGMA integrity_check').get()));
-console.log(`[${writerId}] counts:`, db.prepare('SELECT writer, COUNT(*) n FROM messages GROUP BY writer').all());
-db.close();

package/docs/cross-mount-stress/host-writer.mjs

@@ -1,47 +0,0 @@
-// Host-side stress writer — uses better-sqlite3 (Node) like the paraclaw host.
-// Inserts N rows into session.db with random sleeps. Pair with container-writer.mjs
-// running inside a container that bind-mounts the same file to demonstrate
-// cross-mount lock contention.
-import Database from 'better-sqlite3';
-import { setTimeout as sleep } from 'node:timers/promises';
-
-const dbPath = process.argv[2] || '/tmp/paraclaw-xmount-test/session.db';
-const writerId = process.argv[3] || 'host';
-const count = Number(process.argv[4] || '500');
-
-const db = new Database(dbPath);
-db.pragma('journal_mode = DELETE');
-db.pragma('busy_timeout = 5000');
-db.exec(`CREATE TABLE IF NOT EXISTS messages (
-  id INTEGER PRIMARY KEY AUTOINCREMENT,
-  writer TEXT NOT NULL,
-  seq INTEGER NOT NULL,
-  ts TEXT NOT NULL DEFAULT (datetime('now'))
-);`);
-
-const stmt = db.prepare('INSERT INTO messages (writer, seq) VALUES (?, ?)');
-const start = Date.now();
-let written = 0;
-let errors = 0;
-
-for (let i = 1; i <= count; i++) {
-  try {
-    stmt.run(writerId, i);
-    written++;
-  } catch (e) {
-    errors++;
-    if (errors < 5) console.error(`${writerId} write error #${i}: ${e.message}`);
-  }
-  if (i % 50 === 0) await sleep(2 + Math.random() * 8);
-  else if (Math.random() < 0.1) await sleep(Math.random() * 3);
-}
-
-const elapsed = ((Date.now() - start) / 1000).toFixed(1);
-console.log(`[${writerId}] wrote ${written}/${count} rows in ${elapsed}s, ${errors} errors`);
-
-const integrity = db.prepare('PRAGMA integrity_check').get();
-console.log(`[${writerId}] integrity:`, JSON.stringify(integrity));
-
-const counts = db.prepare('SELECT writer, COUNT(*) AS n FROM messages GROUP BY writer ORDER BY writer').all();
-console.log(`[${writerId}] row counts:`, counts);
-db.close();