tokentracker-cli 0.97.2 → 0.99.0

package/src/lib/rollout.js

@@ -4188,6 +4188,35 @@ function recordProjectMeta(projectState, meta) {
   projectState.projects[projectKey] = next;
 }
 
+function antigravityProjectAssignmentChanged(previous, projectContext) {
+  const projectKey = projectContext?.projectKey || null;
+  const projectRef = projectContext?.projectRef || null;
+  const priorProjectKeys = new Set(
+    Object.values(previous?.contributions || {})
+      .map((contribution) => contribution?.projectKey)
+      .filter((value) => typeof value === "string" && value.length > 0),
+  );
+  if (typeof previous?.projectKey === "string" && previous.projectKey) {
+    priorProjectKeys.add(previous.projectKey);
+  }
+  const priorProjectRefs = new Set(
+    Object.values(previous?.contributions || {})
+      .map((contribution) => contribution?.projectRef)
+      .filter((value) => typeof value === "string" && value.length > 0),
+  );
+  if (typeof previous?.projectRef === "string" && previous.projectRef) {
+    priorProjectRefs.add(previous.projectRef);
+  }
+  return (
+    priorProjectKeys.size !== (projectKey ? 1 : 0) ||
+    (projectKey && !priorProjectKeys.has(projectKey)) ||
+    priorProjectRefs.size !== (projectRef ? 1 : 0) ||
+    (projectRef && !priorProjectRefs.has(projectRef)) ||
+    (typeof previous?.projectStatus === "string" &&
+      previous.projectStatus !== (projectContext?.status || null))
+  );
+}
+
 async function resolveProjectContextForFile({
   filePath,
   projectMetaCache,
@@ -15792,34 +15821,73 @@ function resolveCraftDefaultModel() {
 
 // Reasonix persists content-free cumulative usage beside each session JSONL.
 // Reading only these telemetry sidecars keeps prompts and tool output private.
-function resolveReasonixHome(env = process.env) {
+function reasonixHomeCandidates(env = process.env) {
   if (env.TOKENTRACKER_REASONIX_HOME) {
-    return expandHomePath(env.TOKENTRACKER_REASONIX_HOME, env);
+    return [expandHomePath(env.TOKENTRACKER_REASONIX_HOME, env)];
   }
   if (env.REASONIX_STATE_HOME) {
-    return expandHomePath(env.REASONIX_STATE_HOME, env);
+    return [expandHomePath(env.REASONIX_STATE_HOME, env)];
   }
-  const home = env.HOME || require("node:os").homedir();
-  return path.join(home, ".reasonix");
+  // Windows installs of Git Bash / MSYS / conda export a HOME of their own
+  // (often a POSIX-shaped path), so preferring it silently sends the scan to a
+  // directory that does not exist and Reasonix drops out of `status` entirely
+  // with no "skipped" line to explain it. Match resolveCopilotDbPaths and take
+  // USERPROFILE first on win32.
+  const home =
+    process.platform === "win32"
+      ? env.USERPROFILE || env.HOME || require("node:os").homedir()
+      : env.HOME || require("node:os").homedir();
+  const candidates = [path.join(home, ".reasonix")];
+  // On Windows the dot-directory is not where the data lives: Reasonix keeps
+  // sessions and memory under %APPDATA%\reasonix (no leading dot) and only the
+  // cache under %LOCALAPPDATA%. Taking USERPROFILE over a shell HOME was not
+  // enough for #641 because ~/.reasonix does not exist on Windows at all, so
+  // the provider read as "not installed" and vanished from status without a
+  // skipped line. The cache root is deliberately left out.
+  if (process.platform === "win32" && env.APPDATA) {
+    candidates.push(path.join(env.APPDATA, "reasonix"));
+  }
+  return candidates;
+}
+
+// The home used for "is Reasonix installed" (src/commands/status.js,
+// src/commands/init.js): the first candidate that exists, else the first, so
+// the caller still has a path to report.
+function resolveReasonixHome(env = process.env) {
+  const candidates = reasonixHomeCandidates(env);
+  return candidates.find((dir) => fssync.existsSync(dir)) || candidates[0];
 }
 
-function collectReasonixTelemetryFiles(dir, files) {
+// Depth bound because the scan starts at the Reasonix root rather than at two
+// known subdirectories; it is a guard against a pathological tree, not a layout
+// assumption. The known layout (projects/<p>/sessions) sits at depth 3.
+const REASONIX_SCAN_MAX_DEPTH = 8;
+
+function collectReasonixTelemetryFiles(dir, files, depth = 0) {
+  if (depth > REASONIX_SCAN_MAX_DEPTH) return;
   if (!fssync.existsSync(dir)) return;
   let entries;
   try { entries = fssync.readdirSync(dir, { withFileTypes: true }); } catch { return; }
   for (const entry of entries) {
     const full = path.join(dir, entry.name);
-    if (entry.isDirectory()) collectReasonixTelemetryFiles(full, files);
+    if (entry.isDirectory()) collectReasonixTelemetryFiles(full, files, depth + 1);
     else if (entry.isFile() && entry.name.endsWith(".jsonl.telemetry.json")) files.push(full);
   }
 }
 
 function resolveReasonixTelemetryFiles(env = process.env) {
-  const reasonixHome = resolveReasonixHome(env);
+  // Recurse from each root instead of from a hardcoded projects/ and sessions/.
+  // The #641 reporter supplied a screenshot of the data location, not a
+  // directory listing, so the layout under %APPDATA%\reasonix is unconfirmed —
+  // and a renamed subdirectory would otherwise cost another release to notice.
   const files = [];
-  collectReasonixTelemetryFiles(path.join(reasonixHome, "projects"), files);
-  collectReasonixTelemetryFiles(path.join(reasonixHome, "sessions"), files);
-  return files.sort((a, b) => a.localeCompare(b));
+  const seenHomes = new Set();
+  for (const home of reasonixHomeCandidates(env)) {
+    if (seenHomes.has(home)) continue;
+    seenHomes.add(home);
+    collectReasonixTelemetryFiles(home, files);
+  }
+  return Array.from(new Set(files)).sort((a, b) => a.localeCompare(b));
 }
 
 function readReasonixSnapshot(filePath) {
@@ -19411,27 +19479,236 @@ function resolveAntigravityBrainDirs(geminiHome) {
   ];
 }
 
-async function listAntigravitySessionFiles(brainDir) {
+function antigravityPathWithin(filePath, root) {
+  if (typeof filePath !== "string" || typeof root !== "string") return false;
+  const relative = path.relative(root, filePath);
+  return Boolean(
+    relative &&
+      relative !== ".." &&
+      !relative.startsWith(`..${path.sep}`) &&
+      !path.isAbsolute(relative),
+  );
+}
+
+async function listAntigravitySessionFilesWithStatus(brainDir, knownFiles = null) {
   const out = [];
-  if (!brainDir || typeof brainDir !== "string") return out;
-  const entries = await safeReadDir(brainDir).catch(() => []);
+  const known = knownFiles instanceof Set ? knownFiles : new Set(knownFiles || []);
+  if (!brainDir || typeof brainDir !== "string") {
+    return { files: out, complete: false };
+  }
+
+  let entries;
+  try {
+    entries = await fs.readdir(brainDir, { withFileTypes: true });
+  } catch (err) {
+    // A missing optional install directory is an authoritative empty inventory
+    // only when it has never contained a tracked session. If it has, the
+    // directory may be temporarily unavailable and must not trigger retraction.
+    return {
+      files: out,
+      complete:
+        err?.code === "ENOENT" &&
+        !Array.from(known).some((filePath) => antigravityPathWithin(filePath, brainDir)),
+    };
+  }
+
+  let complete = true;
   for (const entry of entries) {
     if (!entry.isDirectory()) continue;
     const logsDir = path.join(brainDir, entry.name, ".system_generated", "logs");
     const transcriptPath = path.join(logsDir, "transcript.jsonl");
-    const st = await fs.stat(transcriptPath).catch(() => null);
-    if (st && st.isFile()) {
-      out.push(transcriptPath);
+    let st;
+    try {
+      st = await fs.stat(transcriptPath);
+    } catch (err) {
+      // ENOENT means this session directory currently has no transcript (or
+      // the transcript was deleted). Other errors make the scan incomplete.
+      if (err?.code !== "ENOENT") complete = false;
+      continue;
     }
+    if (st.isFile()) out.push(transcriptPath);
   }
+
   out.sort((a, b) => a.localeCompare(b));
-  return out;
+  return { files: out, complete };
 }
 
-async function listAntigravityTranscripts(geminiHome) {
+async function listAntigravitySessionFiles(brainDir) {
+  return (await listAntigravitySessionFilesWithStatus(brainDir)).files;
+}
+
+async function listAntigravityTranscriptsWithStatus(geminiHome, knownFiles = null) {
   const dirs = resolveAntigravityBrainDirs(geminiHome);
-  const lists = await Promise.all(dirs.map((dir) => listAntigravitySessionFiles(dir)));
-  return lists.flat();
+  const results = await Promise.all(
+    dirs.map((dir) => listAntigravitySessionFilesWithStatus(dir, knownFiles)),
+  );
+  return {
+    files: results.flatMap((result) => result.files),
+    complete: results.every((result) => result.complete),
+  };
+}
+
+async function listAntigravityTranscripts(geminiHome) {
+  return (await listAntigravityTranscriptsWithStatus(geminiHome)).files;
+}
+
+const ANTIGRAVITY_CURSOR_VERSION = 2;
+// Bump when extractAntigravityGenInfo starts reading records it used to
+// discard. Kept apart from the cursor version: a cursor-version mismatch routes
+// to the legacy migration, which treats the cursor as having no per-file ledger.
+const ANTIGRAVITY_EXTRACTOR_REVISION = 1;
+const ANTIGRAVITY_MAX_CONTRIBUTIONS = 4096;
+
+function hashAntigravityTranscript(buffer) {
+  if (!Buffer.isBuffer(buffer)) return null;
+  return crypto.createHash("sha256").update(buffer).digest("hex");
+}
+
+function antigravityTranscriptPrefixMatches(buffer, previous) {
+  if (!Buffer.isBuffer(buffer) || !previous || typeof previous.transcriptHash !== "string") {
+    return false;
+  }
+  const previousSize = Number(previous.size);
+  if (!Number.isSafeInteger(previousSize) || previousSize < 0 || buffer.length < previousSize) {
+    return false;
+  }
+  return hashAntigravityTranscript(buffer.subarray(0, previousSize)) === previous.transcriptHash;
+}
+
+async function readAntigravityTranscriptBytes(filePath) {
+  try {
+    return await fs.readFile(filePath);
+  } catch (_error) {
+    return null;
+  }
+}
+
+async function statAntigravityDatabase(dbPath) {
+  const result = {
+    dbExists: false,
+    dbInode: 0,
+    dbMtimeMs: 0,
+    dbCtimeMs: 0,
+    dbSize: 0,
+    dbWalInode: 0,
+    dbWalMtimeMs: 0,
+    dbWalSize: 0,
+    dbJournalInode: 0,
+    dbJournalMtimeMs: 0,
+    dbJournalSize: 0,
+  };
+  if (!dbPath) return result;
+  for (const [suffix, prefix] of [
+    ["", "db"],
+    ["-wal", "dbWal"],
+    ["-journal", "dbJournal"],
+  ]) {
+    const st = await fs.stat(`${dbPath}${suffix}`).catch(() => null);
+    if (!st || !st.isFile()) continue;
+    if (suffix === "") result.dbExists = true;
+    result[`${prefix}MtimeMs`] = Number.isFinite(st.mtimeMs) ? st.mtimeMs : 0;
+    if (suffix === "") {
+      result.dbInode = Number.isFinite(st.ino) ? st.ino : 0;
+      result.dbCtimeMs = Number.isFinite(st.ctimeMs) ? st.ctimeMs : 0;
+    } else {
+      result[`${prefix}Inode`] = Number.isFinite(st.ino) ? st.ino : 0;
+    }
+    result[`${prefix}Size`] = Number.isFinite(st.size) ? st.size : 0;
+  }
+  return result;
+}
+
+// Neither the transcript nor the database changes when the extractor does, so
+// without this stamp a session read before an extractor fix keeps its totals.
+function antigravityExtractorStale(previous, dbPath) {
+  return Boolean(
+    dbPath &&
+      previous &&
+      previous.cursorVersion === ANTIGRAVITY_CURSOR_VERSION &&
+      previous.extractorRevision !== ANTIGRAVITY_EXTRACTOR_REVISION,
+  );
+}
+
+function sameAntigravityDatabase(previous, current) {
+  if (!previous || !current) return false;
+  // SQLite updates the -shm read marks when a reader opens a WAL database.
+  // Those mtime changes do not represent new usage metadata and must not force
+  // a full transcript reconciliation on every sync.
+  return [
+    "dbExists",
+    "dbInode",
+    "dbMtimeMs",
+    "dbCtimeMs",
+    "dbSize",
+    "dbWalInode",
+    "dbWalMtimeMs",
+    "dbWalSize",
+    "dbJournalInode",
+    "dbJournalMtimeMs",
+    "dbJournalSize",
+  ].every((field) => Number(previous[field] || 0) === Number(current[field] || 0));
+}
+
+function capAntigravityContributions(contributions) {
+  if (!contributions || typeof contributions !== "object") {
+    return { contributions: {}, complete: true };
+  }
+  const entries = Object.entries(contributions);
+  if (entries.length <= ANTIGRAVITY_MAX_CONTRIBUTIONS) {
+    return { contributions, complete: true };
+  }
+
+  entries.sort(([, left], [, right]) => {
+    const leftTime = Date.parse(left?.bucketStart || "");
+    const rightTime = Date.parse(right?.bucketStart || "");
+    if (Number.isFinite(leftTime) && Number.isFinite(rightTime) && leftTime !== rightTime) {
+      return rightTime - leftTime;
+    }
+    if (Number.isFinite(rightTime) !== Number.isFinite(leftTime)) {
+      return Number.isFinite(rightTime) ? 1 : -1;
+    }
+    return 0;
+  });
+
+  return {
+    contributions: Object.fromEntries(entries.slice(0, ANTIGRAVITY_MAX_CONTRIBUTIONS)),
+    complete: false,
+  };
+}
+
+function resetAntigravityBucketsForRebuild({
+  hourlyState,
+  projectState,
+  sources,
+  touchedBuckets,
+  projectTouchedBuckets,
+}) {
+  if (!sources || sources.size === 0) return;
+
+  for (const [key, bucket] of Object.entries(hourlyState?.buckets || {})) {
+    if (!bucket || !bucket.totals) continue;
+    const bucketSource = normalizeSourceInput(parseBucketKey(key).source);
+    if (!sources.has(bucketSource)) continue;
+    bucket.totals = initTotals();
+    bucket.queuedKey = null;
+    bucket.retractedUnknownKey = null;
+    touchedBuckets.add(key);
+  }
+
+  for (const key of Object.keys(hourlyState?.groupQueued || {})) {
+    const groupSource = normalizeSourceInput(key.split(BUCKET_SEPARATOR)[0]);
+    if (sources.has(groupSource)) delete hourlyState.groupQueued[key];
+  }
+
+  if (!projectState || !projectTouchedBuckets) return;
+  for (const [key, bucket] of Object.entries(projectState.buckets || {})) {
+    const keyParts = key.split(BUCKET_SEPARATOR);
+    const bucketSource = normalizeSourceInput(bucket?.source || keyParts.at(-2));
+    if (!sources.has(bucketSource) || !bucket?.totals) continue;
+    bucket.totals = initTotals();
+    bucket.queuedKey = null;
+    projectTouchedBuckets.add(key);
+  }
 }
 
 async function parseAntigravityIncremental({
@@ -19442,6 +19719,10 @@ async function parseAntigravityIncremental({
   onProgress,
   source,
   publicRepoResolver,
+  // Source-wide rebuilds may retract sessions absent from sessionFiles. Only
+  // permit that when discovery explicitly completed; direct parser callers
+  // retain the historical complete-inventory default for compatibility.
+  inventoryComplete = true,
 }) {
   await ensureDir(path.dirname(queuePath));
   let filesProcessed = 0;
@@ -19450,17 +19731,147 @@ async function parseAntigravityIncremental({
   const cb = typeof onProgress === "function" ? onProgress : null;
   const files = Array.isArray(sessionFiles) ? sessionFiles : [];
   const totalFiles = files.length;
-  const hourlyState = normalizeHourlyState(cursors?.hourly);
+  // Keep every mutation private until both queue appends succeed. The generic
+  // normalizers intentionally preserve bucket object references for other
+  // parsers, so Antigravity must deep-clone its working state here; otherwise a
+  // queue failure can leave a committed cursor pointing at already-mutated
+  // hourly totals.
+  const hourlyState = structuredClone(normalizeHourlyState(cursors?.hourly));
   const projectEnabled = typeof projectQueuePath === "string" && projectQueuePath.length > 0;
-  const projectState = projectEnabled ? normalizeProjectState(cursors?.projectHourly) : null;
+  const projectState = projectEnabled
+    ? structuredClone(normalizeProjectState(cursors?.projectHourly))
+    : null;
   const projectTouchedBuckets = projectEnabled ? new Set() : null;
   const projectMetaCache = projectEnabled ? new Map() : null;
   const publicRepoCache = projectEnabled ? new Map() : null;
   const touchedBuckets = new Set();
   const defaultSource = normalizeSourceInput(source) || "antigravity";
+  const fileCursors =
+    cursors.files && typeof cursors.files === "object" && !Array.isArray(cursors.files)
+      ? structuredClone(cursors.files)
+      : {};
 
-  if (!cursors.files || typeof cursors.files !== "object") {
-    cursors.files = {};
+  const fileSources = new Set();
+  const fileSourceByPath = new Map();
+  for (const entry of files) {
+    const filePath = typeof entry === "string" ? entry : entry?.path;
+    const fileSource =
+      typeof entry === "string"
+        ? defaultSource
+        : normalizeSourceInput(entry?.source) || defaultSource;
+    if (filePath) fileSourceByPath.set(filePath, fileSource);
+    fileSources.add(fileSource);
+  }
+  if (fileSources.size === 0) fileSources.add(defaultSource);
+  for (const [filePath, prev] of Object.entries(fileCursors)) {
+    if (!resolveAntigravityDbPath(filePath)) continue;
+    const previousSource = normalizeSourceInput(prev?.source);
+    if (previousSource) fileSources.add(previousSource);
+  }
+
+  // Legacy cursors have no per-file contribution ledger. Rebuild when every
+  // legacy file is still present; if discovery is incomplete, preserve the old
+  // aggregate and migrate live files against transcript-only baselines below.
+  const liveFilePaths = new Set(
+    files
+      .map((entry) => (typeof entry === "string" ? entry : entry?.path))
+      .filter(Boolean),
+  );
+  const incompleteContributionPaths = new Set();
+  const deferredIncompletePaths = new Set();
+  let needsFullRebuild = false;
+  for (const [filePath, prev] of Object.entries(fileCursors)) {
+    if (!resolveAntigravityDbPath(filePath) || !prev || Number(prev.lastLine || 0) <= 0) {
+      continue;
+    }
+    const isLegacy =
+      prev.cursorVersion !== ANTIGRAVITY_CURSOR_VERSION ||
+      !prev.contributions ||
+      typeof prev.contributions !== "object";
+    if (!isLegacy && prev.contributionsComplete === false) {
+      incompleteContributionPaths.add(filePath);
+    }
+  }
+
+  // Project reassignment is another full-ledger change: a capped ledger cannot
+  // retract contributions that fell outside its retained window. Resolve
+  // incomplete files against a private project state before deciding whether a
+  // source-wide rebuild is required.
+  const preflightProjectState =
+    projectEnabled && incompleteContributionPaths.size > 0
+      ? structuredClone(projectState)
+      : null;
+  const preflightProjectMetaCache = preflightProjectState ? new Map() : null;
+  const preflightPublicRepoCache = preflightProjectState ? new Map() : null;
+
+  // A capped contribution ledger is safe while its transcript/SQLite identity
+  // is unchanged. If an incomplete ledger needs reconciliation, rebuild the
+  // complete source baseline before applying any per-file deltas. When source
+  // discovery is incomplete, defer the file instead: resetting the source
+  // would erase contributions belonging to sessions omitted by discovery.
+  if (!needsFullRebuild && incompleteContributionPaths.size > 0) {
+    for (const filePath of incompleteContributionPaths) {
+      if (!liveFilePaths.has(filePath)) {
+        if (!inventoryComplete) {
+          deferredIncompletePaths.add(filePath);
+        } else {
+          needsFullRebuild = true;
+        }
+        if (needsFullRebuild) break;
+        continue;
+      }
+      const prev = cursors.files[filePath];
+      const st = await fs.stat(filePath).catch(() => null);
+      const dbPath = resolveAntigravityDbPath(filePath);
+      const dbIdentity = await statAntigravityDatabase(dbPath);
+      const sameInode = st && prev.inode === (st.ino || 0);
+      const transcriptCanAppend =
+        sameInode &&
+        (Number.isFinite(st.size) ? st.size : 0) >= Number(prev.size || 0) &&
+        antigravityTranscriptPrefixMatches(
+          await readAntigravityTranscriptBytes(filePath),
+          prev,
+        );
+      const sameSource =
+        !fileSourceByPath.has(filePath) ||
+        (prev.source == null ? defaultSource : prev.source) === fileSourceByPath.get(filePath);
+      const sameDb = sameAntigravityDatabase(prev, dbIdentity);
+      const staleExtractor = antigravityExtractorStale(prev, dbPath);
+      let projectChanged = false;
+      if (preflightProjectState) {
+        const projectContext = await resolveProjectContextForFile({
+          filePath,
+          projectMetaCache: preflightProjectMetaCache,
+          publicRepoCache: preflightPublicRepoCache,
+          publicRepoResolver,
+          projectState: preflightProjectState,
+        });
+        projectChanged = antigravityProjectAssignmentChanged(prev, projectContext);
+      }
+      if (!transcriptCanAppend || !sameDb || staleExtractor || !sameSource || projectChanged) {
+        if (!inventoryComplete) {
+          deferredIncompletePaths.add(filePath);
+        } else {
+          needsFullRebuild = true;
+        }
+        if (needsFullRebuild) break;
+      }
+    }
+  }
+
+  if (needsFullRebuild) {
+    for (const filePath of Object.keys(fileCursors)) {
+      if (resolveAntigravityDbPath(filePath) && !liveFilePaths.has(filePath)) {
+        delete fileCursors[filePath];
+      }
+    }
+    resetAntigravityBucketsForRebuild({
+      hourlyState,
+      projectState,
+      sources: fileSources,
+      touchedBuckets,
+      projectTouchedBuckets,
+    });
   }
 
   for (let idx = 0; idx < files.length; idx++) {
@@ -19472,17 +19883,140 @@ async function parseAntigravityIncremental({
         ? defaultSource
         : normalizeSourceInput(entry?.source) || defaultSource;
     const st = await fs.stat(filePath).catch(() => null);
-    if (!st || !st.isFile()) continue;
+    if (!st || !st.isFile()) {
+      if (needsFullRebuild) {
+        throw new Error(`Antigravity transcript disappeared during rebuild: ${filePath}`);
+      }
+      continue;
+    }
 
     const key = filePath;
-    const prev = cursors.files[key] || null;
+    if (deferredIncompletePaths.has(key)) continue;
+    const prev = fileCursors[key] || null;
+
     const inode = st.ino || 0;
     const size = Number.isFinite(st.size) ? st.size : 0;
     const mtimeMs = Number.isFinite(st.mtimeMs) ? st.mtimeMs : 0;
+    const ctimeMs = Number.isFinite(st.ctimeMs) ? st.ctimeMs : 0;
 
-    const unchanged =
-      prev && prev.inode === inode && prev.size === size && prev.mtimeMs === mtimeMs;
-    if (unchanged) {
+    const isLegacy = Boolean(
+      prev &&
+        (prev.cursorVersion !== ANTIGRAVITY_CURSOR_VERSION ||
+          !prev.contributions ||
+          typeof prev.contributions !== "object"),
+    );
+
+    const dbPath = resolveAntigravityDbPath(filePath);
+    const dbIdentity = await statAntigravityDatabase(dbPath);
+
+    const sameInode = prev && prev.inode === inode;
+    const sameTranscriptStats =
+      sameInode &&
+      prev.size === size &&
+      prev.mtimeMs === mtimeMs &&
+      Number(prev.ctimeMs || 0) === ctimeMs;
+    const sameDb = prev && sameAntigravityDatabase(prev, dbIdentity);
+    const sameSource =
+      !prev || (prev.source == null ? defaultSource : prev.source) === fileSource;
+    const currentVersion = !isLegacy && prev?.cursorVersion === ANTIGRAVITY_CURSOR_VERSION;
+    const staleExtractor = !isLegacy && antigravityExtractorStale(prev, dbPath);
+    let transcriptBytes = null;
+    let transcriptHash = typeof prev?.transcriptHash === "string" ? prev.transcriptHash : null;
+    let sameTranscriptContent = sameTranscriptStats;
+    let transcriptCanAppend = false;
+
+    // Stat metadata is only a fast hint. On any apparent file change, compare
+    // the bytes so atomic replacement, truncation, and edits before the cursor
+    // cannot be mistaken for an append. A cursor without this hash is treated
+    // conservatively and gets one reconciliation pass.
+    const transcriptReadRequired =
+      !sameTranscriptStats || !currentVersion || !prev?.transcriptHash || needsFullRebuild;
+    if (transcriptReadRequired) {
+      transcriptBytes = await readAntigravityTranscriptBytes(filePath);
+      const afterReadStat = await fs.stat(filePath).catch(() => null);
+      if (
+        !Buffer.isBuffer(transcriptBytes) ||
+        !afterReadStat ||
+        !afterReadStat.isFile() ||
+        (afterReadStat.ino || 0) !== inode ||
+        transcriptBytes.length !== size ||
+        afterReadStat.size !== size ||
+        afterReadStat.mtimeMs !== mtimeMs ||
+        (Number.isFinite(afterReadStat.ctimeMs) ? afterReadStat.ctimeMs : 0) !== ctimeMs
+      ) {
+        throw new Error(`Antigravity transcript changed while being read: ${filePath}`);
+      }
+      transcriptHash = hashAntigravityTranscript(transcriptBytes);
+      sameTranscriptContent =
+        Boolean(
+          (prev?.transcriptHash && transcriptHash && prev.transcriptHash === transcriptHash && prev.size === size) ||
+            (!prev?.transcriptHash && sameTranscriptStats),
+        );
+      transcriptCanAppend =
+        Boolean(
+          prev &&
+            transcriptBytes &&
+            size >= Number(prev.size || 0) &&
+            (prev.transcriptHash
+              ? antigravityTranscriptPrefixMatches(transcriptBytes, prev)
+              : sameInode),
+        );
+    }
+
+    const projectContext = projectEnabled
+      ? await resolveProjectContextForFile({
+          filePath,
+          projectMetaCache,
+          publicRepoCache,
+          publicRepoResolver,
+          projectState,
+        })
+      : null;
+    const projectRef = projectContext?.projectRef || null;
+    const projectKey = projectContext?.projectKey || null;
+    const projectChanged =
+      projectEnabled && antigravityProjectAssignmentChanged(prev, projectContext);
+    const baselineLine = Number(prev?.baselineLine || (isLegacy ? prev?.lastLine || 0 : 0));
+
+    // Legacy migration: "no retroactive claim". The file's history stays in the
+    // old aggregate untouched. If unchanged, stamp v2 cursor with contributions: {}
+    // and queue no buckets.
+    if (!needsFullRebuild && isLegacy && (sameTranscriptStats || sameTranscriptContent)) {
+      const contributionState = capAntigravityContributions({});
+      fileCursors[key] = {
+        ...prev,
+        inode,
+        size,
+        mtimeMs,
+        ctimeMs,
+        source: fileSource,
+        transcriptHash:
+          transcriptHash ||
+          (transcriptBytes
+            ? hashAntigravityTranscript(transcriptBytes)
+            : prev?.transcriptHash || null),
+        ...dbIdentity,
+        projectKey: projectRef ? projectKey : null,
+        projectRef: projectRef || null,
+        projectStatus: projectContext?.status || null,
+        projectConfigPath: projectContext?.configPath || null,
+        projectConfigMtimeMs: projectContext?.configMtimeMs ?? null,
+        projectConfigSize: projectContext?.configSize ?? null,
+        projectReconciliationDeferred: false,
+        baselineLine,
+        cursorVersion: ANTIGRAVITY_CURSOR_VERSION,
+        extractorRevision: ANTIGRAVITY_EXTRACTOR_REVISION,
+        lastLine: Number(prev.lastLine || 0),
+        contextTokens: Number(prev.contextTokens || 0),
+        previousContextTokens: Number(prev.previousContextTokens || 0),
+        currentModel: prev.currentModel,
+        lastPlannerModel: prev.lastPlannerModel,
+        usageSource: prev.usageSource || (dbIdentity ? "sqlite" : "estimated"),
+        dbReadOk: prev.dbReadOk !== false,
+        contributions: contributionState.contributions,
+        contributionsComplete: contributionState.complete,
+        updatedAt: new Date().toISOString(),
+      };
       filesProcessed += 1;
       if (cb) {
         cb({
@@ -19497,54 +20031,250 @@ async function parseAntigravityIncremental({
       continue;
     }
 
-    const sameFile = prev && prev.inode === inode;
-    const lastLine = sameFile ? Number(prev.lastLine || 0) : 0;
-    const initialContextTokens = sameFile ? Number(prev.contextTokens || 0) : 0;
-    const initialPrevContext = sameFile ? Number(prev.previousContextTokens || 0) : 0;
-    const initialModel = sameFile && typeof prev.currentModel === "string" ? prev.currentModel : null;
-    const initialLastPlannerModel =
-      sameFile && typeof prev.lastPlannerModel === "string" ? prev.lastPlannerModel : null;
-    const initialUsageSource = sameFile && typeof prev.usageSource === "string" ? prev.usageSource : null;
-
-    const projectContext = projectEnabled
-      ? await resolveProjectContextForFile({
+    const metadataRetryNeeded = currentVersion && Boolean(dbPath) && prev?.dbReadOk !== true;
+    if (
+      sameTranscriptStats &&
+      sameTranscriptContent &&
+      sameDb &&
+      sameSource &&
+      currentVersion &&
+      !staleExtractor &&
+      !metadataRetryNeeded &&
+      !projectChanged &&
+      !needsFullRebuild
+    ) {
+      filesProcessed += 1;
+      if (cb) {
+        cb({
+          index: idx + 1,
+          total: totalFiles,
           filePath,
-          projectMetaCache,
-          publicRepoCache,
-          publicRepoResolver,
-          projectState,
-        })
-      : null;
-    const projectRef = projectContext?.projectRef || null;
-    const projectKey = projectContext?.projectKey || null;
+          filesProcessed,
+          eventsAggregated,
+          bucketsQueued: touchedBuckets.size,
+        });
+      }
+      continue;
+    }
 
-    const result = await parseAntigravityFile({
-      filePath,
-      lastLine,
-      initialContextTokens,
-      initialPrevContext,
-      initialModel,
-      initialLastPlannerModel,
-      initialUsageSource,
-      hourlyState,
-      touchedBuckets,
-      source: fileSource,
-      projectState,
-      projectTouchedBuckets,
-      projectRef,
-      projectKey,
-    });
+    // Reconcile historical contributions whenever the cursor needs migration or
+    // SQLite changed, even when new transcript lines were appended in the same
+    // sync. Otherwise the incremental branch stamps a migrated cursor without
+    // repairing its old, incomplete totals. Legacy cursors have no retroactive
+    // claim and are not routed through reconciliation.
+    const needsReconcile =
+      !needsFullRebuild &&
+      !isLegacy &&
+      prev &&
+      (metadataRetryNeeded ||
+        !currentVersion ||
+        !sameDb ||
+        staleExtractor ||
+        !sameSource ||
+        projectChanged ||
+        (!sameTranscriptContent && !transcriptCanAppend));
+
+    // Reconciliation reparses the complete transcript. Even when stat metadata
+    // looked unchanged, project/SQLite changes can select this path; obtain a
+    // stable snapshot here rather than letting parseAntigravityFile silently
+    // turn a transient read failure into an empty transcript.
+    if (!transcriptBytes && (needsFullRebuild || needsReconcile)) {
+      transcriptBytes = await readAntigravityTranscriptBytes(filePath);
+      const afterReadStat = await fs.stat(filePath).catch(() => null);
+      if (
+        !Buffer.isBuffer(transcriptBytes) ||
+        !afterReadStat ||
+        !afterReadStat.isFile() ||
+        (afterReadStat.ino || 0) !== inode ||
+        transcriptBytes.length !== size ||
+        afterReadStat.size !== size ||
+        afterReadStat.mtimeMs !== mtimeMs ||
+        (Number.isFinite(afterReadStat.ctimeMs) ? afterReadStat.ctimeMs : 0) !== ctimeMs
+      ) {
+        throw new Error(`Antigravity transcript changed while being read: ${filePath}`);
+      }
+      transcriptHash = hashAntigravityTranscript(transcriptBytes);
+      sameTranscriptContent =
+        Boolean(transcriptHash && prev?.transcriptHash === transcriptHash && prev.size === size);
+      transcriptCanAppend = Boolean(
+        prev &&
+          size >= Number(prev.size || 0) &&
+          antigravityTranscriptPrefixMatches(transcriptBytes, prev),
+      );
+    }
 
-    cursors.files[key] = {
+    const protectExactState = needsReconcile && prev?.usageSource === "sqlite";
+    const hourlyStateBeforeReconcile = protectExactState ? structuredClone(hourlyState) : null;
+    const projectStateBeforeReconcile =
+      protectExactState && projectState ? structuredClone(projectState) : null;
+    const touchedBeforeReconcile = protectExactState ? new Set(touchedBuckets) : null;
+    const projectTouchedBeforeReconcile =
+      protectExactState && projectTouchedBuckets ? new Set(projectTouchedBuckets) : null;
+    let result;
+
+    if (needsFullRebuild || needsReconcile) {
+      if (
+        !needsFullRebuild &&
+        prev &&
+        prev.contributions &&
+        typeof prev.contributions === "object"
+      ) {
+        for (const c of Object.values(prev.contributions)) {
+          if (!c || !c.totals) continue;
+          const oldBucket = getHourlyBucket(hourlyState, c.source, c.model, c.bucketStart);
+          subtractTotals(oldBucket.totals, c.totals);
+          touchedBuckets.add(bucketKey(c.source, c.model, c.bucketStart));
+          if (
+            projectEnabled &&
+            c.projectKey &&
+            projectState &&
+            projectTouchedBuckets
+          ) {
+            const oldProjectBucket = getProjectBucket(
+              projectState,
+              c.projectKey,
+              c.source,
+              c.bucketStart,
+              c.projectRef || null,
+            );
+            subtractTotals(oldProjectBucket.totals, c.totals);
+            projectTouchedBuckets.add(projectBucketKey(c.projectKey, c.source, c.bucketStart));
+          }
+        }
+      }
+      result = await parseAntigravityFile({
+        filePath,
+        ...(transcriptBytes ? { rawBuffer: transcriptBytes } : {}),
+        lastLine: 0,
+        baselineLine: needsFullRebuild ? 0 : baselineLine,
+        watermarkLine: Number(prev?.lastLine || 0),
+        initialUsageSource: prev?.usageSource,
+        hourlyState,
+        touchedBuckets,
+        source: fileSource,
+        projectState,
+        projectTouchedBuckets,
+        projectRef,
+        projectKey,
+      });
+    } else {
+      const canResume = sameTranscriptContent || transcriptCanAppend;
+      const lastLine = canResume ? Number(prev?.lastLine || 0) : 0;
+      const initialContextTokens = canResume ? Number(prev?.contextTokens || 0) : 0;
+      const initialPrevContext = canResume ? Number(prev?.previousContextTokens || 0) : 0;
+      const initialModel =
+        canResume && typeof prev?.currentModel === "string" ? prev.currentModel : null;
+      const initialLastPlannerModel =
+        canResume && typeof prev?.lastPlannerModel === "string"
+          ? prev.lastPlannerModel
+          : null;
+      const initialUsageSource =
+        canResume && typeof prev?.usageSource === "string" ? prev.usageSource : null;
+
+      result = await parseAntigravityFile({
+        filePath,
+        ...(transcriptBytes ? { rawBuffer: transcriptBytes } : {}),
+        lastLine,
+        baselineLine,
+        initialContextTokens,
+        initialPrevContext,
+        initialModel,
+        initialLastPlannerModel,
+        initialUsageSource,
+        hourlyState,
+        touchedBuckets,
+        source: fileSource,
+        projectState,
+        projectTouchedBuckets,
+        projectRef,
+        projectKey,
+      });
+
+      const mergedContributions = { ...(prev?.contributions || {}) };
+      for (const [cKey, cVal] of Object.entries(result.contributions || {})) {
+        if (!mergedContributions[cKey]) {
+          mergedContributions[cKey] = { ...cVal, totals: { ...cVal.totals } };
+        } else {
+          addTotals(mergedContributions[cKey].totals, cVal.totals);
+        }
+      }
+      result.contributions = mergedContributions;
+    }
+
+    if (result.deferred && prev) {
+      if (needsFullRebuild && !hourlyStateBeforeReconcile) {
+        throw new Error(`Cannot rebuild Antigravity transcript while SQLite is unavailable: ${filePath}`);
+      }
+      if (hourlyStateBeforeReconcile) {
+        for (const field of Object.keys(hourlyState)) delete hourlyState[field];
+        Object.assign(hourlyState, hourlyStateBeforeReconcile);
+        if (touchedBeforeReconcile) {
+          touchedBuckets.clear();
+          for (const touched of touchedBeforeReconcile) touchedBuckets.add(touched);
+        }
+      }
+      if (projectState && projectStateBeforeReconcile) {
+        for (const field of Object.keys(projectState)) delete projectState[field];
+        Object.assign(projectState, projectStateBeforeReconcile);
+        if (projectTouchedBeforeReconcile) {
+          projectTouchedBuckets.clear();
+          for (const touched of projectTouchedBeforeReconcile) projectTouchedBuckets.add(touched);
+        }
+      }
+      fileCursors[key] = {
+        ...prev,
+        inode,
+        size,
+        mtimeMs,
+        ctimeMs,
+        source: fileSource,
+        transcriptHash,
+        ...dbIdentity,
+        dbReadOk: false,
+        updatedAt: new Date().toISOString(),
+      };
+      filesProcessed += 1;
+      if (cb) {
+        cb({
+          index: idx + 1,
+          total: totalFiles,
+          filePath,
+          filesProcessed,
+          eventsAggregated,
+          bucketsQueued: touchedBuckets.size,
+        });
+      }
+      continue;
+    }
+
+    const contributionState = capAntigravityContributions(result.contributions);
+    fileCursors[key] = {
       inode,
       size,
       mtimeMs,
+      ctimeMs,
+      source: fileSource,
+      transcriptHash,
+      ...dbIdentity,
+      projectKey: projectRef ? projectKey : null,
+      projectRef: projectRef || null,
+      projectStatus: projectContext?.status || null,
+      projectConfigPath: projectContext?.configPath || null,
+      projectConfigMtimeMs: projectContext?.configMtimeMs ?? null,
+      projectConfigSize: projectContext?.configSize ?? null,
+      projectReconciliationDeferred: false,
+      baselineLine: needsFullRebuild ? 0 : baselineLine,
+      cursorVersion: ANTIGRAVITY_CURSOR_VERSION,
+      extractorRevision: ANTIGRAVITY_EXTRACTOR_REVISION,
       lastLine: result.lastLine,
       contextTokens: result.contextTokens,
       previousContextTokens: result.previousContextTokens,
       currentModel: result.currentModel,
       lastPlannerModel: result.lastPlannerModel,
       usageSource: result.usageSource,
+      dbReadOk: result.dbReadOk !== false,
+      contributions: contributionState.contributions,
+      contributionsComplete: contributionState.complete,
       updatedAt: new Date().toISOString(),
     };
 
@@ -19568,6 +20298,7 @@ async function parseAntigravityIncremental({
     ? await enqueueTouchedProjectBuckets({ projectQueuePath, projectState, projectTouchedBuckets })
     : 0;
   hourlyState.updatedAt = new Date().toISOString();
+  cursors.files = fileCursors;
   cursors.hourly = hourlyState;
   if (projectState) {
     projectState.updatedAt = new Date().toISOString();
@@ -19578,78 +20309,164 @@ async function parseAntigravityIncremental({
 }
 
 function decodeAntigravityVarint(buf, offset) {
+  if (!Buffer.isBuffer(buf) || !Number.isSafeInteger(offset) || offset < 0) {
+    throw new RangeError("invalid Antigravity protobuf offset");
+  }
+  const start = offset;
   let res = 0;
-  let shift = 0;
-  while (offset < buf.length) {
+  for (let count = 0; count < 10; count++) {
+    if (offset >= buf.length) throw new RangeError("truncated Antigravity protobuf varint");
     const b = buf[offset++];
-    res += (b & 0x7f) * 2 ** shift;
-    if (!(b & 0x80)) break;
-    shift += 7;
+    // 64 bits is nine 7-bit groups plus one bit of a tenth byte.
+    if (count === 9 && b > 1) throw new RangeError("Antigravity protobuf varint exceeds 64 bits");
+    res += (b & 0x7f) * 2 ** (count * 7);
+    if (!(b & 0x80)) {
+      if (Number.isSafeInteger(res)) return [res, offset];
+      // Antigravity writes 2^64 - 1 in fields the extractor never reads. The
+      // float sum has already lost precision, so re-read the bytes exactly.
+      let exact = 0n;
+      for (let i = start; i < offset; i++) {
+        exact |= BigInt(buf[i] & 0x7f) << BigInt((i - start) * 7);
+      }
+      return [exact, offset];
+    }
   }
-  return [res, offset];
+  throw new RangeError("overlong Antigravity protobuf varint");
 }
 
 function findAntigravityProtoFields(buf) {
+  if (!Buffer.isBuffer(buf)) throw new RangeError("invalid Antigravity protobuf payload");
   const fields = [];
   let offset = 0;
   while (offset < buf.length) {
     const [tag, next] = decodeAntigravityVarint(buf, offset);
+    if (typeof tag !== "number") throw new RangeError("unsafe Antigravity protobuf field tag");
     offset = next;
-    const fieldNum = tag >> 3;
+    const fieldNum = Math.floor(tag / 8);
     const wireType = tag & 7;
+    if (fieldNum <= 0) throw new RangeError("invalid Antigravity protobuf field number");
     if (wireType === 0) {
+      // May be a BigInt; token reads use Number.isSafeInteger, so it is never billed.
       const [val, vNext] = decodeAntigravityVarint(buf, offset);
       offset = vNext;
       fields.push({ num: fieldNum, val });
     } else if (wireType === 2) {
       const [len, lNext] = decodeAntigravityVarint(buf, offset);
       offset = lNext;
+      if (len > buf.length - offset) {
+        throw new RangeError("truncated Antigravity protobuf field");
+      }
       fields.push({ num: fieldNum, val: buf.subarray(offset, offset + len) });
       offset += len;
     } else if (wireType === 1) {
+      if (buf.length - offset < 8) throw new RangeError("truncated Antigravity protobuf field");
       offset += 8;
     } else if (wireType === 5) {
+      if (buf.length - offset < 4) throw new RangeError("truncated Antigravity protobuf field");
       offset += 4;
     } else {
-      break;
+      throw new RangeError("unsupported Antigravity protobuf wire type");
     }
   }
   return fields;
 }
 
 function extractAntigravityGenInfo(buf) {
-  const root = findAntigravityProtoFields(buf);
-  const f1 = root.find((f) => f.num === 1)?.val;
-  if (!f1) return null;
-
-  const inner = findAntigravityProtoFields(f1);
-  let model = null;
-  const f19 = inner.find((f) => f.num === 19)?.val;
-  if (f19) model = Buffer.from(f19).toString("utf8").trim();
+  try {
+    const root = findAntigravityProtoFields(buf);
+    const f1 = root.find((f) => f.num === 1)?.val;
+    if (!f1) return null;
+
+    const inner = findAntigravityProtoFields(f1);
+    let model = null;
+    const f19 = inner.find((f) => f.num === 19)?.val;
+    if (f19) model = Buffer.from(f19).toString("utf8").trim();
+
+    let contextTokens = 0;
+    const f9 = inner.find((f) => f.num === 9)?.val;
+    if (f9) {
+      const f10 = findAntigravityProtoFields(f9).find((f) => f.num === 10)?.val;
+      if (f10) {
+        const tok = findAntigravityProtoFields(f10).find((f) => f.num === 1)?.val;
+        if (Number.isSafeInteger(tok)) contextTokens = tok;
+      }
+    }
 
-  let contextTokens = 0;
-  const f9 = inner.find((f) => f.num === 9)?.val;
-  if (f9) {
-    const f10 = findAntigravityProtoFields(f9).find((f) => f.num === 10)?.val;
-    if (f10) {
-      const tok = findAntigravityProtoFields(f10).find((f) => f.num === 1)?.val;
-      if (Number.isFinite(tok)) contextTokens = tok;
+    let lastStepIndex = null;
+    for (const f of inner) {
+      if (f.num !== 20 || !f.val) continue;
+      const kv = findAntigravityProtoFields(f.val);
+      const k = kv.find((x) => x.num === 1)?.val;
+      const v = kv.find((x) => x.num === 2)?.val;
+      if (k && Buffer.from(k).toString("utf8") === "last_step_index" && v) {
+        const rawStepIndex = Buffer.from(v).toString("utf8").trim();
+        if (/^\d+$/.test(rawStepIndex)) {
+          const parsed = Number(rawStepIndex);
+          if (Number.isSafeInteger(parsed)) lastStepIndex = parsed;
+        }
+      }
     }
-  }
 
-  let lastStepIndex = null;
-  for (const f of inner) {
-    if (f.num !== 20 || !f.val) continue;
-    const kv = findAntigravityProtoFields(f.val);
-    const k = kv.find((x) => x.num === 1)?.val;
-    const v = kv.find((x) => x.num === 2)?.val;
-    if (k && Buffer.from(k).toString("utf8") === "last_step_index" && v) {
-      const parsed = parseInt(Buffer.from(v).toString("utf8"), 10);
-      if (Number.isFinite(parsed)) lastStepIndex = parsed;
+    let uncachedInput;
+    let cachedInput;
+    let outputTokens;
+    let textOutput;
+    let reasoningOutput;
+    let hasUsageMetadata = false;
+
+    const f4 = inner.find((f) => f.num === 4)?.val;
+    if (f4) {
+      // Antigravity internal usage message layout (Field 4):
+      // - f4.1: system / tool instruction prefix tokens (per-model fixed prefix)
+      // - f4.2: user / prompt tokens for this turn
+      // - f4.3: total output tokens checksum (f4.9 text + f4.10 reasoning)
+      // - f4.5: cached prompt tokens (cache read)
+      // - f4.9: text output tokens
+      // - f4.10: reasoning / thought tokens
+      const f4fields = findAntigravityProtoFields(f4);
+      const sTok = f4fields.find((f) => f.num === 1)?.val;
+      const pTok = f4fields.find((f) => f.num === 2)?.val;
+      const cTok = f4fields.find((f) => f.num === 5)?.val;
+      const oTok = f4fields.find((f) => f.num === 3)?.val;
+      const tTok = f4fields.find((f) => f.num === 9)?.val;
+      const rTok = f4fields.find((f) => f.num === 10)?.val;
+      if (
+        Number.isSafeInteger(sTok) ||
+        Number.isSafeInteger(pTok) ||
+        Number.isSafeInteger(cTok) ||
+        Number.isSafeInteger(oTok) ||
+        Number.isSafeInteger(tTok) ||
+        Number.isSafeInteger(rTok)
+      ) {
+        hasUsageMetadata = true;
+        const sysTokens = Number.isSafeInteger(sTok) ? sTok : 0;
+        const promptTokens = Number.isSafeInteger(pTok) ? pTok : 0;
+        uncachedInput = sysTokens + promptTokens;
+        cachedInput = Number.isSafeInteger(cTok) ? cTok : 0;
+        outputTokens = Number.isSafeInteger(oTok) ? oTok : 0;
+        textOutput = Number.isSafeInteger(tTok) ? tTok : 0;
+        reasoningOutput = Number.isSafeInteger(rTok) ? rTok : 0;
+      }
     }
-  }
 
-  return { model, contextTokens, lastStepIndex };
+    return {
+      model,
+      contextTokens,
+      lastStepIndex,
+      ...(hasUsageMetadata
+        ? {
+            uncachedInput,
+            cachedInput,
+            outputTokens,
+            textOutput,
+            reasoningOutput,
+            hasUsageMetadata: true,
+          }
+        : {}),
+    };
+  } catch (_error) {
+    return null;
+  }
 }
 
 function resolveAntigravityDbPath(transcriptPath) {
@@ -19661,32 +20478,67 @@ function resolveAntigravityDbPath(transcriptPath) {
 
 function readAntigravityConversationDb(dbPath) {
   if (!dbPath) return null;
+  let snapshot = null;
+  let effectiveDbPath = dbPath;
+  // Antigravity keeps active generation metadata in SQLite WAL files. Reading
+  // a UNC/WSL database directly is both lock-prone and can miss rows that have
+  // not checkpointed into the main file, so copy the database and sidecars as
+  // one local read view first.
+  if (isUncPath(dbPath)) {
+    try {
+      snapshot = snapshotSqliteDb(dbPath);
+      effectiveDbPath = snapshot.path;
+    } catch (_error) {
+      // Preserve the existing direct-read fallback for transient snapshot or
+      // permission failures.
+    }
+  }
   try {
+    if (!fssync.existsSync(effectiveDbPath)) return null;
     const rows = readSqliteJsonRows(
-      dbPath,
+      effectiveDbPath,
       "SELECT idx, quote(data) as hex FROM gen_metadata ORDER BY idx",
-      { readOnly: true },
+      {
+        readOnly: true,
+        throwOnReadFailure: true,
+        maxBuffer: 64 * 1024 * 1024,
+        timeout: 15_000,
+      },
     );
-    if (!Array.isArray(rows) || rows.length === 0) return null;
+    if (!Array.isArray(rows)) return null;
 
     const stepMap = new Map();
     for (const r of rows) {
       if (!r || typeof r.hex !== "string" || !r.hex.startsWith("X'")) continue;
       const buf = Buffer.from(r.hex.slice(2, -1), "hex");
       const info = extractAntigravityGenInfo(buf);
-      if (info && info.lastStepIndex != null && info.contextTokens > 0) {
+      if (
+        info &&
+        info.lastStepIndex != null &&
+        (info.hasUsageMetadata || info.contextTokens > 0)
+      ) {
         stepMap.set(info.lastStepIndex + 1, info);
       }
     }
-    return stepMap.size > 0 ? stepMap : null;
+    // An empty map means the database was read successfully but has no usable
+    // metadata yet. Keep it distinct from null, which signals a transient read
+    // failure and must be retried on the next sync.
+    return stepMap;
   } catch (_) {
     return null;
+  } finally {
+    if (snapshot) snapshot.cleanup();
   }
 }
 
 async function parseAntigravityFile({
   filePath,
-  lastLine,
+  rawBuffer = null,
+  lastLine = 0,
+  maxLine = null,
+  baselineLine = 0,
+  watermarkLine = 0,
+  applyBuckets = true,
   initialContextTokens,
   initialPrevContext,
   initialModel,
@@ -19700,7 +20552,12 @@ async function parseAntigravityFile({
   projectRef,
   projectKey,
 }) {
-  const raw = await fs.readFile(filePath, "utf8").catch(() => "");
+  const raw = Buffer.isBuffer(rawBuffer)
+    ? rawBuffer.toString("utf8")
+    : await fs.readFile(filePath, "utf8").catch(() => null);
+  if (raw === null) {
+    throw new Error(`Antigravity transcript could not be read: ${filePath}`);
+  }
   if (!raw.trim()) {
     return {
       lastLine: 0,
@@ -19710,14 +20567,22 @@ async function parseAntigravityFile({
       currentModel: null,
       lastPlannerModel: null,
       usageSource: "estimated",
+      dbReadOk: true,
+      contributions: {},
     };
   }
 
-  const lines = raw
+  const allLines = raw
     .split(/\r?\n/)
     .map((line) => line.trim())
     .filter(Boolean);
+  const endLimit =
+    Number.isFinite(maxLine) && maxLine >= 0
+      ? Math.min(allLines.length, maxLine)
+      : allLines.length;
+  const lines = allLines.slice(0, endLimit);
   let eventsAggregated = 0;
+  const contributions = {};
   const dbPath = resolveAntigravityDbPath(filePath);
   const stepMap = dbPath ? readAntigravityConversationDb(dbPath) : null;
   // Resume cached context-token total + model so historical lines (i < lastLine)
@@ -19728,11 +20593,30 @@ async function parseAntigravityFile({
   const cachedTokens = Number.isFinite(initialContextTokens) ? initialContextTokens : 0;
   const cachedPrev = Number.isFinite(initialPrevContext) ? initialPrevContext : 0;
   const cachedModel = typeof initialModel === "string" ? initialModel : null;
+  const dbReadFailed = Boolean(dbPath) && stepMap === null;
+  if (dbReadFailed && initialUsageSource === "sqlite") {
+    return {
+      lastLine: Math.min(Number.isFinite(lastLine) ? lastLine : 0, lines.length),
+      eventsAggregated: 0,
+      contextTokens: cachedTokens,
+      previousContextTokens: cachedPrev,
+      currentModel: cachedModel,
+      lastPlannerModel:
+        typeof initialLastPlannerModel === "string" ? initialLastPlannerModel : cachedModel,
+      usageSource: "sqlite",
+      dbReadOk: false,
+      deferred: true,
+      contributions: {},
+    };
+  }
   const sqliteCursor = initialUsageSource === "sqlite";
+  const hasSqliteMetadata = Boolean(stepMap && stepMap.size > 0);
   const resumed =
-    canResume && (cachedTokens > 0 || cachedModel !== null) && (!stepMap || sqliteCursor);
+    canResume &&
+    (cachedTokens > 0 || cachedModel !== null) &&
+    (!hasSqliteMetadata || sqliteCursor);
   const scanStart = resumed ? lastLine : 0;
-  let currentModel = resumed ? cachedModel : null;
+  let currentModel = resumed ? cachedModel : typeof initialModel === "string" ? initialModel : null;
   if (!currentModel) {
     currentModel = await readAntigravityDefaultModel(filePath);
   }
@@ -19780,7 +20664,12 @@ async function parseAntigravityFile({
 
     if (!isNewEvent) {
       if (parsed.type === "PLANNER_RESPONSE") {
-        if (dbContextTokens > 0) {
+        if (dbTurn?.hasUsageMetadata) {
+          contextTokens =
+            dbContextTokens > 0
+              ? dbContextTokens
+              : (dbTurn.uncachedInput || 0) + (dbTurn.cachedInput || 0);
+        } else if (dbContextTokens > 0) {
           contextTokens = dbContextTokens;
         }
         previousContextTokens = contextTokens;
@@ -19812,52 +20701,97 @@ async function parseAntigravityFile({
     let billedPlanner = false;
 
     if (parsed.type === "PLANNER_RESPONSE") {
-      const content = typeof parsed.content === "string" ? parsed.content : "";
-      const thinking = typeof parsed.thinking === "string" ? parsed.thinking : "";
+      if (dbTurn?.hasUsageMetadata) {
+        const uncached = dbTurn.uncachedInput || 0;
+        const cached = dbTurn.cachedInput || 0;
+        const reasoning = dbTurn.reasoningOutput || 0;
+        let output = 0;
+        if (Number.isFinite(dbTurn.textOutput) && dbTurn.textOutput > 0) {
+          output = dbTurn.textOutput;
+        } else if (Number.isFinite(dbTurn.outputTokens)) {
+          output = Math.max(0, dbTurn.outputTokens - reasoning);
+        }
 
-      if (dbContextTokens > 0) {
-        contextTokens = dbContextTokens;
-      }
-      if (lastPlannerModel && model !== lastPlannerModel) {
-        previousContextTokens = 0;
-      }
-      const inputDelta = Math.max(0, contextTokens - previousContextTokens);
+        delta.input_tokens = uncached;
+        delta.cached_input_tokens = cached;
+        delta.output_tokens = output;
+        delta.reasoning_output_tokens = reasoning;
+        delta.total_tokens = uncached + cached + output + reasoning;
+        delta.billable_total_tokens = delta.total_tokens;
+        delta.conversation_count = 1;
+        billedPlanner = delta.total_tokens > 0;
 
-      const outputTokens =
-        antigravityValueTokens(content) + antigravityValueTokens(parsed.tool_calls);
-      const reasoningTokens = antigravityValueTokens(thinking);
+        contextTokens = dbContextTokens > 0 ? dbContextTokens : uncached + cached;
+      } else {
+        const content = typeof parsed.content === "string" ? parsed.content : "";
+        const thinking = typeof parsed.thinking === "string" ? parsed.thinking : "";
 
-      delta.input_tokens = inputDelta;
-      delta.output_tokens = outputTokens;
-      delta.reasoning_output_tokens = reasoningTokens;
-      delta.total_tokens = inputDelta + outputTokens + reasoningTokens;
-      delta.billable_total_tokens = delta.total_tokens;
-      delta.conversation_count = 1;
-      billedPlanner = delta.total_tokens > 0;
+        if (dbContextTokens > 0) {
+          contextTokens = dbContextTokens;
+        }
+        if (lastPlannerModel && model !== lastPlannerModel) {
+          previousContextTokens = 0;
+        }
+        const inputDelta = Math.max(0, contextTokens - previousContextTokens);
+
+        const outputTokens =
+          antigravityValueTokens(content) + antigravityValueTokens(parsed.tool_calls);
+        const reasoningTokens = antigravityValueTokens(thinking);
+
+        delta.input_tokens = inputDelta;
+        delta.output_tokens = outputTokens;
+        delta.reasoning_output_tokens = reasoningTokens;
+        delta.total_tokens = inputDelta + outputTokens + reasoningTokens;
+        delta.billable_total_tokens = delta.total_tokens;
+        delta.conversation_count = 1;
+        billedPlanner = delta.total_tokens > 0;
+      }
     }
 
-    if (!billedPlanner) {
+    if (!billedPlanner || i < baselineLine) {
+      if (billedPlanner) {
+        previousContextTokens = contextTokens;
+        lastPlannerModel = model;
+      }
       contextTokens += eventContextTokens;
       lastCompletedLine = i + 1;
       continue;
     }
 
-    const bucket = getHourlyBucket(hourlyState, source, model, bucketStart);
-    addTotals(bucket.totals, delta);
-    touchedBuckets.add(bucketKey(source, model, bucketStart));
-
-    if (projectKey && projectState && projectTouchedBuckets) {
-      const projectBucket = getProjectBucket(
-        projectState,
-        projectKey,
+    const cKey = `${source}|${model}|${bucketStart}|${projectKey || ""}`;
+    if (!contributions[cKey]) {
+      contributions[cKey] = {
         source,
+        model,
         bucketStart,
-        projectRef,
-      );
-      addTotals(projectBucket.totals, delta);
-      projectTouchedBuckets.add(projectBucketKey(projectKey, source, bucketStart));
+        projectKey: projectKey || null,
+        projectRef: projectRef || null,
+        totals: initTotals(),
+      };
+    }
+    addTotals(contributions[cKey].totals, delta);
+
+    if (applyBuckets !== false && hourlyState && touchedBuckets) {
+      const bucket = getHourlyBucket(hourlyState, source, model, bucketStart);
+      addTotals(bucket.totals, delta);
+      touchedBuckets.add(bucketKey(source, model, bucketStart));
+
+      if (projectKey && projectState && projectTouchedBuckets) {
+        const projectBucket = getProjectBucket(
+          projectState,
+          projectKey,
+          source,
+          bucketStart,
+          projectRef,
+        );
+        addTotals(projectBucket.totals, delta);
+        projectTouchedBuckets.add(projectBucketKey(projectKey, source, bucketStart));
+      }
+    }
+    const isNewForCounter = watermarkLine > 0 ? i >= watermarkLine : true;
+    if (isNewForCounter) {
+      eventsAggregated += 1;
     }
-    eventsAggregated += 1;
     // Snapshot the pre-planner context first. The planner's own content+tool_calls
     // (eventContextTokens, added below) become part of the next turn's history,
     // so they MUST be billed as input on the next planner — don't fold them into
@@ -19875,7 +20809,9 @@ async function parseAntigravityFile({
     previousContextTokens,
     currentModel,
     lastPlannerModel,
-    usageSource: stepMap ? "sqlite" : "estimated",
+    usageSource: hasSqliteMetadata ? "sqlite" : "estimated",
+    dbReadOk: !dbPath || stepMap !== null,
+    contributions,
   };
 }
 
@@ -22033,9 +22969,12 @@ module.exports = {
   parseGrokBuildIncremental,
 
   // Antigravity (Google Gemini) - Session logs parser
+  ANTIGRAVITY_CURSOR_VERSION,
+  ANTIGRAVITY_EXTRACTOR_REVISION,
   resolveAntigravityBrainDirs,
   listAntigravitySessionFiles,
   listAntigravityTranscripts,
+  listAntigravityTranscriptsWithStatus,
   parseAntigravityIncremental,
   estimateAntigravityTokens,
   isCjkCodePoint,