@awak-app/simy-cli 0.7.36 → 0.7.38

package/src/session-supervision/claude-sessions.js

@@ -5,7 +5,8 @@ import { lstat, open, readdir, realpath } from 'node:fs/promises';
 import { homedir } from 'node:os';
 import { createHash } from 'node:crypto';
 import { isAbsolute, join, relative, resolve, sep } from 'node:path';
-import { boundConversation, CONVERSATION_LIMITS, textContent } from './conversation-context.js';
+import { boundConversation, CONVERSATION_LIMITS, createConversationCollector, textContent } from './conversation-context.js';
+import { createForegroundAgentContractCollector, isForegroundNativeReviewInput } from './claude-native-agent-contract.js';
 import { sessionTitle } from './session-titles.js';
 import { RECENT_SESSION_LIMIT, recentSessionWindow } from './discovery-window.js';
 import { observeClaudeDesktopLineage, readClaudeDesktopLineage } from './claude-desktop-lineage.js';
@@ -152,7 +153,8 @@ async function inspectTranscript(root, projectName, fileName, observedAt, includ
     const candidate = candidateMetadata(id, metadata, opened.mtimeMs, observedAt);
     signal?.throwIfAborted();
     if (includeStatistics) candidate.statistics = await statisticsFromHandle(handle, opened, 'claude', id, { signal });
-    if (includeContext) candidate.context = await conversationFromHandle(handle, opened, id, metadata.cwd, includeContext === 'user');
+    if (includeContext) candidate.context = await conversationFromHandle(handle, opened, id, metadata.cwd, {
+      userOnly: ['user', 'review_user'].includes(includeContext), reviewUserContext: includeContext === 'review_user', signal });
     if (humanRequest) candidate.humanBoundary = await humanBoundaryFromHandle(handle, opened, path, id, metadata.cwd, humanRequest, signal);
     if (includeBoundary) {
       candidate.transcriptPath = path;
@@ -309,29 +311,47 @@ const humanFileFingerprint = info => humanFileIdentity(info) === null ? null : c
 // cooperative cancellation, exact byte offsets and a digest of every read byte.
 // Whole-history and human-cursor readers retain their separate semantics.
 async function scanTranscriptLines(handle, opened, { maxBytes, maxLineBytes = 4 * 1024 * 1024,
-  maxRows = 250000, signal, onLine, onChunk = () => {} }) {
-  if (!opened.size || opened.size > maxBytes) return null;
+  maxRows = 250000, start = 0, allowIncomplete = false, signal, onLine, onChunk = () => {} }) {
+  if (!opened.size || opened.size - start > maxBytes) return null;
   const hash = createHash('sha256'), buffer = Buffer.alloc(64 * 1024);
-  let offset = 0, pending = Buffer.alloc(0), rows = 0, lineOffset = 0;
+  let offset = start, fragments = [], pendingBytes = 0, rows = 0, lineOffset = start, discard = start > 0, incomplete = false;
   while (offset < opened.size) {
     signal?.throwIfAborted();
     const { bytesRead } = await handle.read(buffer, 0, Math.min(buffer.length, opened.size - offset), offset);
     if (!bytesRead) return null;
-    const chunk = buffer.subarray(0, bytesRead);
+    let chunk = buffer.subarray(0, bytesRead);
     hash.update(chunk); onChunk(chunk, offset); offset += bytesRead;
-    pending = Buffer.concat([pending, chunk]);
-    let newline;
-    while ((newline = pending.indexOf(10)) !== -1) {
+    if (discard) {
+      const end = chunk.indexOf(10);
+      if (end === -1) continue;
+      chunk = chunk.subarray(end + 1); lineOffset = offset - chunk.length; discard = false;
+    }
+    let position = 0;
+    while (position < chunk.length) {
       signal?.throwIfAborted();
-      if (newline > maxLineBytes || ++rows > maxRows) return null;
-      const line = pending.subarray(0, newline), start = lineOffset;
-      pending = pending.subarray(newline + 1); lineOffset += newline + 1;
-      if (line.length && onLine(line, start) === false) return null;
+      const newline = chunk.indexOf(10, position);
+      const part = chunk.subarray(position, newline === -1 ? chunk.length : newline);
+      const length = pendingBytes + part.length;
+      if (length > maxLineBytes) {
+        if (!allowIncomplete) return null;
+        incomplete = true; fragments = []; pendingBytes = 0;
+        if (newline === -1) { discard = true; break; }
+      } else if (newline === -1) {
+        // Copy only fragments crossing reads; concatenate once at the newline.
+        // Repeatedly concatenating a multi-MiB line made scanning quadratic.
+        fragments.push(Buffer.from(part)); pendingBytes = length;
+        break;
+      } else {
+        if (++rows > maxRows) return null;
+        const line = fragments.length ? Buffer.concat([...fragments, part], length) : part;
+        fragments = []; pendingBytes = 0;
+        if (line.length && onLine(line, lineOffset, rows - 1) === false) return null;
+      }
+      position = newline + 1; lineOffset = offset - chunk.length + position;
     }
-    if (pending.length > maxLineBytes) return null;
   }
-  if (pending.length) return null;
-  return { digest: hash.digest('hex'), byteLength: offset, rows };
+  if (pendingBytes) { if (!allowIncomplete) return null; incomplete = true; }
+  return { digest: hash.digest('hex'), byteLength: offset, rows, incomplete };
 }
 
 async function humanBoundaryFromHandle(handle, opened, path, session, cwd, request, signal) {
@@ -410,42 +430,42 @@ export function unchangedClaudeHumanBoundary(proof) {
   } catch { return false; }
 }
 
-async function conversationFromHandle(handle, opened, session, cwd, userOnly = false) {
+async function conversationFromHandle(handle, opened, session, cwd, { userOnly = false, reviewUserContext = false, signal } = {}) {
   // Native review receipts can make a short user conversation exceed the usual
   // display tail. Bound the file separately and filter before bounding prose.
   const bytes = userOnly ? 64 * 1024 * 1024 : CONVERSATION_LIMITS.bytes;
   const offset = Math.max(0, opened.size - bytes);
-  const buffer = Buffer.alloc(Math.min(opened.size, bytes));
-  const { bytesRead } = await handle.read(buffer, 0, buffer.length, offset);
-  let text = buffer.subarray(0, bytesRead).toString('utf8');
-  if (offset) text = text.slice(text.indexOf('\n') + 1);
-  const messages = [];
-  let incomplete = !text.endsWith('\n');
-  const lines = text.split('\n');
-  for (const [index, line] of lines.entries()) {
-    if (!line.trim()) continue;
-    let item;
-    try { item = JSON.parse(line); } catch { incomplete = true; continue; }
-    if (unscopedFileHistory(item)) continue;
-    if (item.sessionId !== session || item.isSidechain === true) { incomplete = true; continue; }
-    if (!['user', 'assistant'].includes(item.type) || item.isMeta === true) continue;
-    if (typeof item.cwd !== 'string' || !isAbsolute(item.cwd) || !within(cwd, resolve(item.cwd))) { incomplete = true; continue; }
-    const text = textContent(item.message?.content);
-    // User tool_result envelopes are not a new human request. A missing
-    // human message/attachment-only request cannot silently reuse an old goal.
-    if (item.type === 'user' && !text.trim() && !item.message?.content?.every?.(part => part?.type === 'tool_result')) incomplete = true;
-    if (userOnly && item.type !== 'user') continue;
-    messages.push({ role: item.type, turnRef: typeof item.uuid === 'string' ? item.uuid : `line:${offset}:${index}`, text });
-  }
+  const collector = createConversationCollector({ reviewUserContext });
+  let incomplete = false;
+  const scanned = await scanTranscriptLines(handle, opened, { start: offset, maxBytes: bytes, maxLineBytes: bytes, maxRows: Infinity, allowIncomplete: true, signal,
+    onLine(line, start, index) {
+      const encoded = line.toString('utf8');
+      if (!encoded.trim()) return;
+      let item;
+      try { item = JSON.parse(encoded); } catch { incomplete = true; return; }
+      if (unscopedFileHistory(item)) return;
+      if (item.sessionId !== session || item.isSidechain === true) { incomplete = true; return; }
+      if (!['user', 'assistant'].includes(item.type) || item.isMeta === true) return;
+      if (typeof item.cwd !== 'string' || !isAbsolute(item.cwd) || !within(cwd, resolve(item.cwd))) { incomplete = true; return; }
+      const text = textContent(item.message?.content);
+      // User tool_result envelopes are not a new human request. A missing
+      // human message/attachment-only request cannot silently reuse an old goal.
+      if (item.type === 'user' && !text.trim() && !item.message?.content?.every?.(part => part?.type === 'tool_result')) incomplete = true;
+      if (reviewUserContext && item.type === 'user' && Array.isArray(item.message?.content) &&
+          item.message.content.some(part => !['text', 'input_text', 'output_text', 'tool_result'].includes(part?.type))) incomplete = true;
+      if (userOnly && item.type !== 'user') return;
+      collector.add({ role: item.type, turnRef: typeof item.uuid === 'string' ? item.uuid : `line:${offset}:${index}`, text });
+  } });
   const after = await handle.stat();
-  if (after.size !== opened.size || after.mtimeMs !== opened.mtimeMs) incomplete = true;
-  return boundConversation(messages, { truncated: offset > 0, incomplete });
+  if (!scanned || scanned.incomplete || after.size !== opened.size || after.mtimeMs !== opened.mtimeMs || after.ctimeMs !== opened.ctimeMs) incomplete = true;
+  return collector.finish({ truncated: offset > 0, incomplete });
 }
 
-export async function readClaudeConversation(candidate, { home, signal, userOnly = false } = {}) {
+export async function readClaudeConversation(candidate, { home, signal, userOnly = false, reviewUserContext = false } = {}) {
+  if (reviewUserContext && !userOnly) throw new Error('Review context must contain only original user requests.');
   if (signal?.aborted) return boundConversation(null);
   if (!UUID.test(candidate?.session || '') || typeof candidate?.cwd !== 'string') return boundConversation(null);
-  const matches = await collect({ home, session: candidate.session, includeContext: userOnly ? 'user' : true, signal });
+  const matches = await collect({ home, session: candidate.session, includeContext: reviewUserContext ? 'review_user' : userOnly ? 'user' : true, signal });
   const selected = matches.filter(item => item.cwd === candidate.cwd);
   if (signal?.aborted || selected.length !== 1) return boundConversation(null);
   if (Number.isFinite(candidate.lastActivityAt) && selected[0].lastActivityAt !== candidate.lastActivityAt) {
@@ -497,49 +517,90 @@ export async function readClaudeStatistics(candidate, { home, signal } = {}) {
 // Consume only an exact native Agent invocation in the original session. A
 // worker checkpoint, forwarded quote, resumed child or tool-using child is not
 // an independent verdict. Native Claude accounts this child in its own budget.
-export async function readClaudeNativeReview(candidate, request, { home, signal } = {}) {
+async function scanNativeRows(path, visit, { signal, expectedDigest } = {}) {
+  signal?.throwIfAborted();
+  if (await realpath(path) !== path) throw new Error('Native transcript link changed');
+  const handle = await open(path, constants.O_RDONLY | (constants.O_NOFOLLOW || 0));
+  try {
+    const before = await handle.stat();
+    if (!before.isFile()) throw new Error('Native transcript unavailable');
+    const decoder = new TextDecoder('utf-8', { fatal: true });
+    const scanned = await scanTranscriptLines(handle, before, { maxBytes: MAX_COMPLETE_TRANSCRIPT_BYTES, signal,
+      onLine: line => visit(JSON.parse(decoder.decode(line))) });
+    const after = await handle.stat();
+    const pathAfter = await lstat(path);
+    if (await realpath(path) !== path || pathAfter.isSymbolicLink() || pathAfter.ino !== before.ino || pathAfter.dev !== before.dev ||
+        before.size !== after.size || before.mtimeMs !== after.mtimeMs || before.ctimeMs !== after.ctimeMs ||
+        pathAfter.size !== before.size || pathAfter.mtimeMs !== before.mtimeMs || pathAfter.ctimeMs !== before.ctimeMs || !scanned ||
+        expectedDigest && scanned.digest !== expectedDigest) throw new Error('Native transcript changed');
+    return { snapshot: [before.dev, before.ino, before.size, before.mtimeMs, before.ctimeMs].join(':'), digest: scanned.digest };
+  } finally { await handle.close(); }
+}
+
+async function nativeParentTranscript(candidate, { home, signal }, visit) {
   const matches = (await collect({ home, session: candidate.session, includeBoundary: true, signal }))
     .filter(row => row.cwd === candidate.cwd && row.boundary?.complete);
   if (matches.length !== 1) return null;
   const parentPath = matches[0].transcriptPath;
-  const readRows = async path => {
-    signal?.throwIfAborted();
-    if (await realpath(path) !== path) throw new Error('Native transcript link changed');
-    const handle = await open(path, constants.O_RDONLY | (constants.O_NOFOLLOW || 0));
-    try {
-      const before = await handle.stat();
-      if (!before.isFile() || before.size > 64 * 1024 * 1024) throw new Error('Native transcript unavailable');
-      const bytes = await handle.readFile('utf8'), after = await handle.stat();
-      const pathAfter = await lstat(path);
-      if (await realpath(path) !== path || pathAfter.isSymbolicLink() || pathAfter.ino !== before.ino || pathAfter.dev !== before.dev ||
-          before.size !== after.size || before.mtimeMs !== after.mtimeMs || before.ctimeMs !== after.ctimeMs || !bytes.endsWith('\n')) throw new Error('Native transcript changed');
-      return bytes.trim().split('\n').map(line => JSON.parse(line));
-    } finally { await handle.close(); }
-  };
+  const scanned = await scanNativeRows(parentPath, row => {
+    if (row.sessionId !== undefined && row.sessionId !== candidate.session ||
+        ['user', 'assistant'].includes(row.type) && (row.isSidechain === true || row.cwd !== candidate.cwd)) return false;
+    return visit(row);
+  }, { signal, expectedDigest: matches[0].boundary.digest });
+  return { parentPath, ...scanned };
+}
+
+export async function readClaudeNativeAgentContract(candidate, { home, signal } = {}) {
+  try {
+    const collector = createForegroundAgentContractCollector(candidate);
+    const parent = await nativeParentTranscript(candidate, { home, signal }, row => collector.add(row));
+    return parent ? collector.finish() : null;
+  } catch (error) { if (signal?.aborted) throw error; return null; }
+}
+
+export async function readClaudeNativeReview(candidate, request, { home, signal } = {}) {
   try {
-    const rows = await readRows(parentPath);
-    const calls = rows.filter(row => row.sessionId === candidate.session && !row.isSidechain && row.type === 'assistant' && Date.parse(row.timestamp) >= request.requestedAt)
-      .flatMap(row => (row.message?.content || []).filter(part => part.type === 'tool_use' && ['Agent', 'Task'].includes(part.name) && part.input?.prompt === request.prompt));
-    if (calls.length !== 1) return null;
-    const call = calls[0];
-    if (call.input.subagent_type !== request.agentType || call.input.run_in_background !== false ||
-        Object.keys(call.input).some(key => !['description', 'prompt', 'subagent_type', 'run_in_background'].includes(key))) return null;
-    const receipts = rows.filter(row => row.sessionId === candidate.session && !row.isSidechain && row.type === 'user' &&
-      row.message?.content?.some?.(part => part.type === 'tool_result' && part.tool_use_id === call.id) && row.toolUseResult);
+    const calls = [], contracts = createForegroundAgentContractCollector(candidate);
+    const parent = await nativeParentTranscript(candidate, { home, signal }, row => {
+      if (row.sessionId === candidate.session && !row.isSidechain && row.type === 'assistant' && Date.parse(row.timestamp) >= request.requestedAt) {
+        for (const part of row.message?.content || []) {
+          if (part.type === 'tool_use' && ['Agent', 'Task'].includes(part.name) && part.input?.prompt === request.prompt)
+            calls.push({ call: part, contract: contracts.finish() });
+          if (calls.length > 1) return false;
+        }
+      }
+      contracts.add(row);
+    });
+    if (!parent || calls.length !== 1) return null;
+    const { parentPath } = parent, { call, contract } = calls[0];
+    if (call.input.subagent_type !== request.agentType || !isForegroundNativeReviewInput(call.name, call.input, contract?.schema)) return null;
+    const receipts = [];
+    const receiptScan = await scanNativeRows(parentPath, row => {
+      if (row.sessionId === candidate.session && !row.isSidechain && row.type === 'user' &&
+          row.message?.content?.some?.(part => part.type === 'tool_result' && part.tool_use_id === call.id) && row.toolUseResult) receipts.push(row);
+      if (receipts.length > 1) return false;
+    }, { signal, expectedDigest: parent.digest });
+    if (receiptScan.snapshot !== parent.snapshot) return null;
     if (receipts.length !== 1) return null;
     const result = receipts[0].toolUseResult;
     if (result.status !== 'completed' || result.prompt !== request.prompt || result.agentType !== request.agentType ||
         result.totalToolUseCount !== 0 || !/^a[a-f0-9]{8,40}$/.test(result.agentId || '')) return null;
     const childPath = join(parentPath.slice(0, -6), 'subagents', `agent-${result.agentId}.jsonl`);
-    const child = await readRows(childPath);
-    if (child.some(row => row.sessionId !== candidate.session || row.agentId !== result.agentId || ['user', 'assistant', 'attachment'].includes(row.type) && row.isSidechain !== true)) return null;
-    const users = child.filter(row => row.type === 'user');
-    const snapshots = child.filter(row => row.attachment?.type === 'prompt_snapshot').map(row => row.attachment);
-    if (users.length !== 1 || textContent(users[0].message?.content) !== request.prompt ||
-        !snapshots.some(row => Array.isArray(row.tools) && row.tools.length === 0) || snapshots.some(row => row.tools && row.tools.length !== 0 || !row.systemPrompt?.some(text => text.trim() === CLAUDE_NATIVE_REVIEW_PROMPT))) return null;
-    const answers = child.filter(row => row.type === 'assistant');
-    if (!answers.length || answers.some(row => row.message?.content?.some(part => part.type === 'tool_use'))) return null;
-    const answer = textContent(answers.at(-1).message?.content);
+    let users = 0, answer = null, emptyTools = false;
+    await scanNativeRows(childPath, row => {
+      if (row.sessionId !== candidate.session || row.agentId !== result.agentId || ['user', 'assistant', 'attachment'].includes(row.type) && row.isSidechain !== true) return false;
+      if (row.type === 'user' && (++users !== 1 || textContent(row.message?.content) !== request.prompt)) return false;
+      if (row.attachment?.type === 'prompt_snapshot') {
+        const snapshot = row.attachment;
+        if (snapshot.tools && snapshot.tools.length !== 0 || !snapshot.systemPrompt?.some(text => text.trim() === CLAUDE_NATIVE_REVIEW_PROMPT)) return false;
+        emptyTools ||= Array.isArray(snapshot.tools) && snapshot.tools.length === 0;
+      }
+      if (row.type === 'assistant') {
+        if (row.message?.content?.some?.(part => part.type === 'tool_use')) return false;
+        answer = textContent(row.message?.content);
+      }
+    }, { signal });
+    if (users !== 1 || !emptyTools || answer === null) return null;
     if (answer !== textContent(result.content)) return null;
     const output = JSON.parse(answer.trim().replace(/^```(?:json)?\s*\n([\s\S]*?)\n```$/, '$1'));
     if (typeof output.passed !== 'boolean' || typeof output.inspectionAvailable !== 'boolean' || typeof output.summary !== 'string' || output.summary.length > 4000 ||

package/src/session-supervision/codex-app-server.js

@@ -28,7 +28,7 @@ import { CODEX_WRITER_LOCK_MINIMUM_VERSION, compareVersions, inspectBackendExecu
 import { hasPendingUserStop, readRecord, recordPath, rootDirectory, canExecuteAutonomousExternal, assessedCheckpointContract, deliveryProgressContext, manualAcceptanceFeedback, manualAcceptanceFeedbackInstruction, taskDirectory, understandingCheckpointContract } from '../codex-plugins/simy/hooks/delivery-loop.mjs';
 import { boundConversation, boundProviderGoal, textContent } from './conversation-context.js';
 import { JsonDataUrlFilter } from './json-data-url-filter.js';
-import { RECENT_SESSION_LIMIT } from './discovery-window.js';
+import { RECENT_SESSION_LIMIT, liveSessionWindow } from './discovery-window.js';
 
 const REQUEST_TIMEOUT = 15_000;
 const MAX_LINE = 256 * 1024;
@@ -307,7 +307,7 @@ export async function listCodexSessions({ client, all = false, signal, onPage, m
     // reports notLoaded rather than the original writer's actual turn state.
     // Bound this metadata fan-out independently of the analysis worker pool.
     const priorities = new Set(prioritySessions);
-    const ordered = [...sessions].sort((a, b) => Number(priorities.has(b.session)) - Number(priorities.has(a.session)) || (b.lastActivityAt || 0) - (a.lastActivityAt || 0));
+    const ordered = liveSessionWindow(sessions, { priority: item => priorities.has(item.session) });
     let next = 0, unavailable = false;
     for (const metadata of sessions) metadata.goalReadState = 'unavailable';
     await Promise.all(Array.from({ length: Math.min(3, sessions.length) }, async () => {
@@ -355,7 +355,7 @@ export async function readCodexArchiveStates(client, candidates, { budgetMs = 30
   const offset = (archiveProbeOffsets.get(codexHome) || 0) % Math.max(1, missing.length);
   let next = 0;
   await Promise.all(Array.from({ length: Math.min(3, missing.length) }, async () => {
-    while (next < missing.length && Date.now() < deadline) {
+    while (next < Math.min(missing.length, RECENT_SESSION_LIMIT) && Date.now() < deadline) {
       const candidate = missing[(offset + next++) % missing.length];
       try {
         const response = await boundedClient.request('thread/read', { threadId: candidate.session, includeTurns: false });

package/src/session-supervision/codex-children.js

@@ -23,14 +23,14 @@ export async function readCodexChildSessions({ parents = [], includeArchived = f
     db = new DatabaseSync(path, { readOnly: true });
     db.exec('PRAGMA query_only = ON; PRAGMA busy_timeout = 50');
     const rows = db.prepare(`WITH RECURSIVE family(id, depth) AS (
-      SELECT id, 0 FROM threads WHERE id IN (${roots.map(() => '?').join(',')})
+      SELECT id, 0 FROM threads WHERE id IN (SELECT value FROM json_each(?))
       UNION SELECT e.child_thread_id, f.depth + 1 FROM thread_spawn_edges e
         JOIN family f ON f.id = e.parent_thread_id WHERE f.depth < 32
     ) SELECT t.id, t.cwd, t.name, t.source, t.model, t.updated_at, t.archived,
         e.parent_thread_id, MIN(f.depth) AS depth
       FROM family f JOIN threads t ON t.id = f.id
       JOIN thread_spawn_edges e ON e.child_thread_id = t.id
-      WHERE f.depth > 0 GROUP BY t.id ORDER BY depth, t.updated_at DESC LIMIT 2000`).all(...roots);
+      WHERE f.depth > 0 GROUP BY t.id ORDER BY depth, t.updated_at DESC LIMIT 2000`).all(JSON.stringify(roots));
     const candidates = [];
     for (const row of rows) {
       let source;

package/src/session-supervision/continuation-analysis.js

@@ -83,7 +83,7 @@ async function unavailableAnalysis(candidate, error, inspectProvider, signal) {
   if (signal?.aborted) return uncertain('Conversation analysis was cancelled.', 'cancelled');
   if (error?.code === 'provider_sign_in_required' || status?.state === 'signed_out') {
     const label = candidate.provider === 'claude' ? 'Claude Code' : 'Codex';
-    return uncertain(`Sign in to ${label}. SIMY rechecks this conversation automatically when sign-in is detected. You can also retry manually. Unfinished work can continue automatically only when Auto Continue is on.`, 'provider_sign_in_required');
+    return uncertain(`Sign in to ${label}. SIMY rechecks this conversation automatically when sign-in is detected. You can also retry manually. Unfinished work continues only when "Continue automatically" is on.`, 'provider_sign_in_required');
   }
   return uncertain('The local AI app could not check this turn. Check its connection, then retry.', 'analysis_unavailable');
 }

package/src/session-supervision/continuation-command.js

@@ -8,7 +8,7 @@ const UUID = /^[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$/i;
 export async function runContinuationCommand(args, { root, output, signal,
   request = requestOwnerContinuation } = {}) {
   if (args.includes('--help') || args.includes('-h')) {
-    output('simy delivery auto-continue on|off|status --provider codex|claude --session UUID\nUses the running local SIMY app and its existing controls. ON saves delegation; only workState describes execution. Explicit stops, approvals, original goals and history remain in force.');
+    output('simy delivery auto-continue on|off|status --provider codex|claude --session UUID\nContinue this conversation automatically: turns Delivery Loop on or off for one exact conversation. Uses the running local SIMY app and its existing controls. ON saves delegation; only workState describes execution. Explicit stops, approvals, original goals and history remain in force.');
     return 0;
   }
   const [operation, ...flags] = args;
@@ -34,12 +34,12 @@ export async function runContinuationCommand(args, { root, output, signal,
   try {
     const receipt = await request({ root, signal, command: command('session_get_status') });
     output(JSON.stringify({ ...receipt, provider, session, accepted,
-      message: receipt?.message || (operation === 'on' && accepted ? 'Auto Continue delegation saved. Read workState and its nextAction for actual progress; this acknowledgement does not prove a provider turn executed.' : undefined) }));
+      message: receipt?.message || (operation === 'on' && accepted ? 'Delivery Loop delegation saved. Read workState and its nextAction for actual progress; this acknowledgement does not prove a provider turn executed.' : undefined) }));
     return receipt?.ok ? 0 : 2;
   } catch (error) {
     if (!accepted) throw error;
     output(JSON.stringify({ ok: false, provider, session, accepted, code: error.code || 'status_unavailable',
-      message: 'The setting was saved, but execution status could not be read. Check auto-continue status; do not repeat ON to infer execution.' }));
+      message: 'The setting was saved, but execution status could not be read. Check `simy delivery auto-continue status`; do not repeat ON to infer execution.' }));
     return 2;
   }
 }

package/src/session-supervision/continuation-readiness.js

@@ -54,5 +54,5 @@ export function projectContinuationWait(candidate, wait, now, noticeAfterMs = CO
   const needsAttention = wait.reasonCode === 'analysis_retry_exhausted' || now - wait.since >= noticeAfterMs;
   return { ...candidate, continuationWait: { id: wait.id, since: wait.since, reasonCode: wait.reasonCode, needsAttention },
     ...(needsAttention ? { autoContinueStatus: 'needs_user',
-      autoContinueReason: 'Auto Continue is saved, but work has been waiting for execution conditions. Check the original conversation or refresh its connection.' } : {}) };
+      autoContinueReason: 'Delivery Loop is on, but work has been waiting for execution conditions. Check the original conversation or refresh its connection.' } : {}) };
 }

package/src/session-supervision/conversation-context.js

@@ -1,6 +1,8 @@
 // Only human/assistant prose crosses the analysis boundary. Tool payloads,
 // attachments, paths and provider envelopes are deliberately excluded.
 export const CONVERSATION_LIMITS = Object.freeze({ messages: 32, text: 4_000, total: 24_000, bytes: 512 * 1024 });
+// Independent review preserves complete original requests within its own budget.
+export const REVIEW_USER_CONTEXT_LIMITS = Object.freeze({ messages: 32, text: 64 * 1024, total: 128 * 1024 });
 const CREDENTIAL_KEY = 'api[_ -]?key|access[_ -]?token|password|authorization';
 const QUOTED_CREDENTIAL = new RegExp(String.raw`(["'])(${CREDENTIAL_KEY})\1(\s*:\s*)(["'])(?:\\.|(?!\4)[^\\])*\4`, 'gi');
 const PLAIN_CREDENTIAL = new RegExp(String.raw`\b(${CREDENTIAL_KEY})\s*[:=]\s*(?:Bearer\s+)?[^\s,;]+`, 'gi');
@@ -41,32 +43,62 @@ export function boundProviderGoal(goal) {
     ...(goal.textTruncated === true || objective.length > 4000 ? { textTruncated: true } : {}) };
 }
 
+// Streaming readers and array callers share one prose-selection policy. Keep
+// the first goal, latest messages and latest-user identity without retaining
+// an entire transcript or sanitizing/trimming an already clipped message twice.
+export function createConversationCollector({ reviewUserContext = false } = {}) {
+  const limits = reviewUserContext ? REVIEW_USER_CONTEXT_LIMITS : CONVERSATION_LIMITS;
+  const prose = [];
+  let count = 0, clipped = false, latestUser = null, foundRedacted = false;
+  return {
+    add(message) {
+      if (!['user', 'assistant'].includes(message?.role) || reviewUserContext && message.role !== 'user') return;
+      const text = sanitizedText(message.text);
+      if (reviewUserContext) {
+        foundRedacted ||= text.includes('[REDACTED]');
+        clipped ||= message.textTruncated === true;
+      }
+      if (!text) return;
+      if (text.length > limits.text) clipped = true;
+      const normalized = { role: message.role, turnRef: sanitizedText(message.turnRef).slice(0, 160) || `message:${count}`,
+        textTruncated: message.textTruncated === true || text.length > limits.text,
+        text: text.slice(0, limits.text) };
+      count++; prose.push(normalized);
+      if (normalized.role === 'user') latestUser = normalized;
+      if (prose.length > limits.messages + 1) prose.splice(1, 1);
+    },
+    finish({ truncated = false, incomplete = false, redacted = false } = {}) {
+      redacted ||= foundRedacted;
+      incomplete ||= reviewUserContext && redacted;
+      truncated ||= clipped || prose.length > limits.messages;
+      // Preserve goal provenance as well as the latest user/assistant exchange.
+      const selected = prose.length > limits.messages ? [prose[0], ...prose.slice(-(limits.messages - 1))] : prose;
+      let total = 0;
+      const result = [];
+      // Spend the bounded budget from newest to oldest; never trim a latest
+      // request to make room for early goal provenance.
+      for (const message of selected.toReversed()) {
+        if (total + message.text.length > limits.total) { truncated = true; continue; }
+        result.unshift(message);
+        total += message.text.length;
+      }
+      const finalMessage = prose.at(-1);
+      if (!latestUser || !result.includes(latestUser) || latestUser.textTruncated || finalMessage?.textTruncated) incomplete = true;
+      return { messages: result.map(({ textTruncated, ...message }) => ({ ...message, ...(textTruncated ? { textTruncated: true } : {}) })), truncated, incomplete, ...(reviewUserContext ? { redacted } : {}) };
+    },
+  };
+}
+
 export function boundConversation(messages, { truncated = false, incomplete = false } = {}) {
   if (!Array.isArray(messages)) return { messages: [], truncated: false, incomplete: true };
-  const prose = [];
-  for (const message of messages) {
-    if (!['user', 'assistant'].includes(message?.role)) continue;
-    const text = sanitizedText(message.text);
-    if (!text) continue;
-    if (text.length > CONVERSATION_LIMITS.text) truncated = true;
-    prose.push({ role: message.role, turnRef: sanitizedText(message.turnRef).slice(0, 160) || `message:${prose.length}`,
-      textTruncated: message.textTruncated === true || text.length > CONVERSATION_LIMITS.text,
-      text: text.slice(0, CONVERSATION_LIMITS.text) });
-  }
-  if (prose.length > CONVERSATION_LIMITS.messages) truncated = true;
-  // Preserve goal provenance as well as the latest user/assistant exchange.
-  const selected = prose.length > CONVERSATION_LIMITS.messages ? [prose[0], ...prose.slice(-(CONVERSATION_LIMITS.messages - 1))] : prose;
-  let total = 0;
-  const result = [];
-  // Spend the bounded budget from newest to oldest; never trim a latest
-  // request to make room for early goal provenance.
-  for (const message of selected.toReversed()) {
-    if (total + message.text.length > CONVERSATION_LIMITS.total) { truncated = true; continue; }
-    result.unshift(message);
-    total += message.text.length;
-  }
-  const latestUser = prose.findLast(message => message.role === 'user');
-  const finalMessage = prose.at(-1);
-  if (!latestUser || !result.includes(latestUser) || latestUser.textTruncated || finalMessage?.textTruncated) incomplete = true;
-  return { messages: result.map(({ textTruncated, ...message }) => ({ ...message, ...(textTruncated ? { textTruncated: true } : {}) })), truncated, incomplete };
+  const collector = createConversationCollector();
+  for (const message of messages) collector.add(message);
+  return collector.finish({ truncated, incomplete });
+}
+
+export function boundReviewUserContext(messages, { truncated = false, incomplete = false, redacted = false } = {}) {
+  if (!Array.isArray(messages)) return { messages: [], truncated: false, incomplete: true, redacted };
+  const collector = createConversationCollector({ reviewUserContext: true });
+  for (const message of messages) collector.add(message);
+  return collector.finish({ truncated, incomplete, redacted });
 }

package/src/session-supervision/conversation-presentation.js

@@ -82,7 +82,7 @@ export function conversationPresentation(candidate = {}, projectedRun, { rawReco
   const pendingDecision = ['approval', 'question'].includes(work.state) || field('pendingApproval') && !field('pendingApproval').denied ||
     field('understandingQuestion') || managed.userQuestion;
   if (candidate.autoContinueStatus === 'save_failed') {
-    ready = readiness('save_failed', 'danger', reason?.text || 'Could not save Auto Continue. Retry.');
+    ready = readiness('save_failed', 'danger', reason?.text || 'Could not save Delivery Loop. Retry.');
   } else if (candidate.autoContinueStatus === 'saving') {
     ready = readiness('saving', 'accent', 'Saving On…');
   } else if (candidate.autoContinue !== true) {
@@ -116,13 +116,13 @@ export function conversationPresentation(candidate = {}, projectedRun, { rawReco
   } else if (terminal) {
     ready = readiness('finished', 'secondary', state === 'reported_done' ? 'The AI reported completion; verification is pending.' : 'No current work to continue.');
   } else if (candidate.provider === 'claude' && candidate.isRunning === false && !['starting', 'recovering'].includes(candidate.autoContinueStatus)) {
-    ready = readiness('original_turn_required', 'attention', 'Send one message in the original conversation. Auto Continue waits for its next verified hook.', action('open_original', 'Open original conversation'));
+    ready = readiness('original_turn_required', 'attention', 'Send one message in the original conversation. Delivery Loop waits for its next verified hook.', action('open_original', 'Open original conversation'));
   } else if (control && candidate.autoContinueStatus === 'supervising' && ['active', 'verification_pending'].includes(state) &&
       ownership === 'simy_continues' && blocker.state === 'ready' &&
       !field('executionUncertain') && !field('providerWarning') && !managed.nativeWait && !managed.purpose?.nativeBoundary) {
     ready = readiness('ready', 'accent', 'Will continue after this turn');
   } else if (['starting', 'recovering'].includes(candidate.autoContinueStatus)) {
-    ready = readiness(candidate.autoContinueStatus, 'accent', reason?.text || 'Checking Auto Continue readiness.');
+    ready = readiness(candidate.autoContinueStatus, 'accent', reason?.text || 'Checking Delivery Loop readiness.');
   } else {
     ready = readiness('waiting', 'secondary', reason?.text || 'Checking whether the next step can be sent.', action('refresh', 'Refresh status'));
   }

package/src/session-supervision/decision-presentation.js

@@ -78,8 +78,8 @@ export function sessionDecisionPresentation(candidate, run) {
     const acceptancePending = run.state === 'verification_pending';
     const providerName = candidate.provider === 'claude' ? 'Claude' : candidate.provider === 'codex' ? 'Codex' : 'AI';
     const reasons = {
-      session_end: pair(`元の${providerName}会話が終了し、自動継続が止まっています。`, `The original ${providerName} session ended, stopping Auto Continue.`),
-      process_exit: pair('元のAIが終了したため、自動継続できていません。', 'The original AI process exited. Auto Continue cannot proceed.'),
+      session_end: pair(`元の${providerName}会話が終了し、Delivery Loop が止まっています。`, `The original ${providerName} session ended, stopping Delivery Loop.`),
+      process_exit: pair('元のAIが終了したため、Delivery Loop を続けられません。', 'The original AI process exited. Delivery Loop cannot proceed.'),
       provider_error: pair('AIの実行でエラーが発生し、作業が止まっています。', 'An AI execution error stopped the work.'),
       shutdown_cancelled: pair('SIMYの終了時に作業が中断されました。', 'Work was interrupted when SIMY exited.'),
       hook_owner_mismatch: pair('元のAI会話の接続が変わりました。', 'The connection to the original AI session changed.'),
@@ -91,9 +91,9 @@ export function sessionDecisionPresentation(candidate, run) {
     const choices = [];
     if (run.canReplan && frozen) choices.push(choice(pair('AIに状況確認を任せる', 'Ask AI to investigate'), pair('目的・履歴・制限を維持して原因と安全な次の手順を調べます。操作の許可を追加することはありません。', 'Investigate the cause and next steps while keeping the goal, history and limits. No new permission is granted.'), 'delivery_replan', {}, 'primary'));
     const enablesAutoContinue = candidate.provider === 'codex' && run.managed !== true && run.state === 'stopped' && candidate.autoContinue !== true;
-    if (run.canResume && frozen) choices.push(choice(pair(enablesAutoContinue ? '自動継続をオンにして再開' : acceptancePending ? '未完了の作業を続ける' : '再開を依頼する', enablesAutoContinue ? 'Resume with Auto Continue' : acceptancePending ? 'Continue remaining work' : 'Request resume'), pair(enablesAutoContinue ? '自動継続をオンにして元の会話の読み取り専用の状況確認から再開します。元の目的・権限・停止履歴・予算を維持します。' : acceptancePending ? '元の目的・履歴・制限を維持して残作業を確認します。最終確認を済ませたことにはしません。' : '保存した目的と履歴を引き継ぎます。接続や権限が不足する場合は、その内容を案内します。', enablesAutoContinue ? 'Turn on Auto Continue and resume with read-only diagnosis in the original conversation. Preserve its goal, permissions, stop history and budgets.' : acceptancePending ? 'Keep the goal, history and limits while checking what remains. This does not accept the result.' : 'Keep the saved goal and history. Any missing connection or permission will be shown.'), 'delivery_resume', {}, choices.length ? 'secondary' : 'primary'));
+    if (run.canResume && frozen) choices.push(choice(pair(enablesAutoContinue ? 'Delivery Loop をオンにして再開' : acceptancePending ? '未完了の作業を続ける' : '再開を依頼する', enablesAutoContinue ? 'Resume with Delivery Loop' : acceptancePending ? 'Continue remaining work' : 'Request resume'), pair(enablesAutoContinue ? 'Delivery Loop をオンにして元の会話の読み取り専用の状況確認から再開します。元の目的・権限・停止履歴・予算を維持します。' : acceptancePending ? '元の目的・履歴・制限を維持して残作業を確認します。最終確認を済ませたことにはしません。' : '保存した目的と履歴を引き継ぎます。接続や権限が不足する場合は、その内容を案内します。', enablesAutoContinue ? 'Turn on Delivery Loop and resume with a read-only status check in the original conversation. Preserve its goal, permissions, stop history and budgets.' : acceptancePending ? 'Keep the goal, history and limits while checking what remains. This does not accept the result.' : 'Keep the saved goal and history. Any missing connection or permission will be shown.'), 'delivery_resume', {}, choices.length ? 'secondary' : 'primary'));
     choices.push(choice(pair(`元の${providerName}会話を開く`, 'Open original conversation ↗'), pair('接続や権限の確認が必要な場合、元のAI会話で確認できます。会話を開くだけでは作業を再開しません。', 'Check the original conversation. Opening it does not resume work.'), 'open_original', {}, 'link'));
-    if (candidate.autoContinue === true) choices.push(choice(pair('自動継続をオフにする', 'Turn off Auto Continue'), pair('今後の自動継続をオフにします。作業の履歴は残ります。', 'Turn off future automatic continuation and retain the history.'), 'session_set_auto_continue', { ...identity, enabled: false }));
+    if (candidate.autoContinue === true) choices.push(choice(pair('Delivery Loop をオフにする', 'Turn off Delivery Loop'), pair('今後の Delivery Loop をオフにします。作業の履歴は残ります。', 'Turn off Delivery Loop for future work and retain the history.'), 'session_set_auto_continue', { ...identity, enabled: false }));
     const actionable = frozen && (run.canReplan || run.canResume);
     return model('recovery', {
       title: pair(acceptancePending ? '最終確認を待ちながら、残作業を続けられます' : '作業が止まっています。対応を選んでください', acceptancePending ? 'Continue work while acceptance is pending' : 'Work stopped. Choose how to proceed'),
@@ -105,12 +105,12 @@ export function sessionDecisionPresentation(candidate, run) {
       candidate.capabilities?.canAutoContinue === true && candidate.autoContinueAvailability?.canEnable !== false) return model('continue', {
     title: pair('続けますか？', 'Continue?'),
     question: pair(clean(candidate.analysis.reason) || '許可済みの残作業があります。', clean(candidate.analysis.reason) || 'Authorized work remains.'),
-    context: pair('自動継続はオフです。オンにすると、元の目的と権限の範囲で残作業を確認します。結果を受け入れる操作ではありません。', 'Auto Continue is off. Turning it on checks remaining work within the original goal and permissions. It does not accept the result.'),
-  }, { choices: [choice(pair('自動継続をオンにする', 'Turn on Auto Continue'), pair('自動継続の設定を保存します。作業の開始や完了を証明するものではありません。', 'Save consent for Auto Continue. This does not confirm that work started or finished.'), 'session_set_auto_continue', { ...identity, enabled: true }, 'primary'),
+    context: pair('Delivery Loop はオフです。オンにすると、元の目的と権限の範囲で残作業を確認します。結果を受け入れる操作ではありません。', 'Delivery Loop is off. Turning it on checks remaining work within the original goal and permissions. It does not accept the result.'),
+  }, { choices: [choice(pair('Delivery Loop をオンにする', 'Turn on Delivery Loop'), pair('Delivery Loop の設定を保存します。作業の開始や完了を証明するものではありません。', 'Save consent for Delivery Loop. This does not confirm that work started or finished.'), 'session_set_auto_continue', { ...identity, enabled: true }, 'primary'),
     choice(pair('元の会話を開く', 'Open original conversation ↗'), pair('会話を開くだけでは作業を再開しません。', 'Opening the conversation does not resume work.'), 'open_original', {}, 'link')] });
   if (candidate.continuationWait?.needsAttention === true && candidate.autoContinue === true) return model('continuation_wait', {
-    title: pair('自動継続の条件を確認してください', 'Check Auto Continue conditions'),
-    question: pair('自動継続はオンのまま、実行条件の確認を待っています。元の会話や接続状態を確認してください。', 'Auto Continue is still on, waiting for execution conditions. Check the original conversation or its connection.'),
+    title: pair('Delivery Loop の条件を確認してください', 'Check Delivery Loop conditions'),
+    question: pair('Delivery Loop はオンのまま、実行条件の確認を待っています。元の会話や接続状態を確認してください。', 'Delivery Loop is still on, waiting for execution conditions. Check the original conversation or its connection.'),
     context: pair('設定・目的・停止指示・承認待ちは維持しています。この通知で作業を再開したり、権限を追加したりしません。', 'The setting, goal, stops and pending approvals are preserved. This notice does not resume work or grant permission.'),
   }, { choices: [choice(pair('状況を再確認する', 'Refresh status'), pair('条件を再確認します。', 'Recheck conditions.'), 'refresh', identity),
     choice(pair('元の会話を開く', 'Open original conversation ↗'), pair('元の会話を確認します。', 'Check the original conversation.'), 'open_original', {}, 'link')] }, 'helper');

package/src/session-supervision/discovery-window.js

@@ -4,14 +4,16 @@ export const RECENT_SESSION_LIMIT = 200;
 
 // Explicit tracking and unsettled work remain visible outside recent history.
 // Completed, untracked records are history, not an unlimited extra catalog.
-export const retainedSession = record => Boolean(record?.listPinned || record?.autoContinue || record?.monitor || record?.tracked ||
+export const supervisedSession = record => Boolean(record?.autoContinue || record?.monitor || record?.tracked ||
   record?.run && !TERMINAL.has(record.run.state));
+export const retainedSession = record => Boolean(record?.listPinned || supervisedSession(record));
 
-export function recentSessionWindow(items, { limit = RECENT_SESSION_LIMIT, priority = () => false } = {}) {
+export function recentSessionWindow(items, { limit = RECENT_SESSION_LIMIT, priority = () => false, includeAncestors = true } = {}) {
   const recent = [...items].sort((a, b) => (b.lastActivityAt || 0) - (a.lastActivityAt || 0) || a.session.localeCompare(b.session));
   const retained = [], history = [];
   for (const item of recent) (priority(item) ? retained : history).push(item);
   const selected = [...retained, ...history.slice(0, Math.max(0, limit - retained.length))];
+  if (!includeAncestors) return selected;
   const byKey = new Map(items.map(item => [`${item.provider}:${item.session}`, item]));
   const included = new Set(selected.map(item => `${item.provider}:${item.session}`));
   // Keep observed ancestors so a recent child does not silently become a root.
@@ -25,3 +27,7 @@ export function recentSessionWindow(items, { limit = RECENT_SESSION_LIMIT, prior
   }
   return selected;
 }
+
+// Tree visibility and navigation pins do not subscribe all ancestors to live
+// work. Explicit supervision and owned jobs are supplied by each caller.
+export const liveSessionWindow = (items, options = {}) => recentSessionWindow(items, { ...options, includeAncestors: false });