cctrack 0.3.0__tar.gz → 0.4.1__tar.gz

{cctrack-0.3.0 → cctrack-0.4.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: cctrack
-Version: 0.3.0
+Version: 0.4.1
 Summary: Claude Code cost tracker — scans JSONL logs and reports token usage + estimated cost
 Project-URL: Homepage, https://github.com/rappdw/cctrack
 Project-URL: Repository, https://github.com/rappdw/cctrack
@@ -86,11 +86,14 @@ Daily breakdown:
 
 ## Remote aggregation
 
-With `--remote`, cctrack SSHs to each host (one call per host) and streams back all JSONL content for local parsing. Requires SSH key auth.
+With `--remote`, cctrack SSHs to each host and streams back both the JSONL logs and the host's statusline hook data (`~/.claude/cctrack/`) for local parsing, so authoritative hook costs from every machine are summed per day. Requires SSH key auth.
 
 ```bash
 # Aggregate this machine + DGX server + laptop
 cctrack --remote dgx macbook-air
+
+# For accurate aggregate costs, install the hook on each host too
+cctrack --install-hook --remote dgx macbook-air
 ```
 
 ## Rate card

{cctrack-0.3.0 → cctrack-0.4.1}/README.md

@@ -61,11 +61,14 @@ Daily breakdown:
 
 ## Remote aggregation
 
-With `--remote`, cctrack SSHs to each host (one call per host) and streams back all JSONL content for local parsing. Requires SSH key auth.
+With `--remote`, cctrack SSHs to each host and streams back both the JSONL logs and the host's statusline hook data (`~/.claude/cctrack/`) for local parsing, so authoritative hook costs from every machine are summed per day. Requires SSH key auth.
 
 ```bash
 # Aggregate this machine + DGX server + laptop
 cctrack --remote dgx macbook-air
+
+# For accurate aggregate costs, install the hook on each host too
+cctrack --install-hook --remote dgx macbook-air
 ```
 
 ## Rate card

{cctrack-0.3.0 → cctrack-0.4.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "cctrack"
-version = "0.3.0"
+version = "0.4.1"
 description = "Claude Code cost tracker — scans JSONL logs and reports token usage + estimated cost"
 readme = "README.md"
 license = "MIT"

{cctrack-0.3.0 → cctrack-0.4.1}/src/cctrack/__init__.py

@@ -26,16 +26,18 @@ from pathlib import Path
 
 # Longer/more-specific prefixes MUST come first — "claude-opus-4-6" would
 # otherwise match "claude-opus-4" and get the wrong rate.
+# cache_write is the 5-minute TTL rate (1.25× input); cache_write_1h is the
+# 1-hour TTL rate (2× input). Claude Code hits the 1h tier heavily in practice.
 RATES = [
-    {"family": "claude-opus-4-7",   "input": 5.0,   "output": 25.0,  "cache_read": 0.50,  "cache_write": 6.25},
-    {"family": "claude-opus-4-6",   "input": 5.0,   "output": 25.0,  "cache_read": 0.50,  "cache_write": 6.25},
-    {"family": "claude-opus-4-5",   "input": 5.0,   "output": 25.0,  "cache_read": 0.50,  "cache_write": 6.25},
-    {"family": "claude-opus-4",     "input": 15.0,  "output": 75.0,  "cache_read": 1.50,  "cache_write": 18.75},
-    {"family": "claude-sonnet-4",   "input": 3.0,   "output": 15.0,  "cache_read": 0.30,  "cache_write": 3.75},
-    {"family": "claude-sonnet-3",   "input": 3.0,   "output": 15.0,  "cache_read": 0.30,  "cache_write": 3.75},
-    {"family": "claude-haiku-4",    "input": 1.0,   "output": 5.0,   "cache_read": 0.10,  "cache_write": 1.25},
-    {"family": "claude-haiku-3-5",  "input": 0.80,  "output": 4.0,   "cache_read": 0.08,  "cache_write": 1.00},
-    {"family": "claude-haiku-3",    "input": 0.25,  "output": 1.25,  "cache_read": 0.03,  "cache_write": 0.30},
+    {"family": "claude-opus-4-7",   "input": 5.0,   "output": 25.0,  "cache_read": 0.50,  "cache_write": 6.25,   "cache_write_1h": 10.0},
+    {"family": "claude-opus-4-6",   "input": 5.0,   "output": 25.0,  "cache_read": 0.50,  "cache_write": 6.25,   "cache_write_1h": 10.0},
+    {"family": "claude-opus-4-5",   "input": 5.0,   "output": 25.0,  "cache_read": 0.50,  "cache_write": 6.25,   "cache_write_1h": 10.0},
+    {"family": "claude-opus-4",     "input": 15.0,  "output": 75.0,  "cache_read": 1.50,  "cache_write": 18.75,  "cache_write_1h": 30.0},
+    {"family": "claude-sonnet-4",   "input": 3.0,   "output": 15.0,  "cache_read": 0.30,  "cache_write": 3.75,   "cache_write_1h": 6.0},
+    {"family": "claude-sonnet-3",   "input": 3.0,   "output": 15.0,  "cache_read": 0.30,  "cache_write": 3.75,   "cache_write_1h": 6.0},
+    {"family": "claude-haiku-4",    "input": 1.0,   "output": 5.0,   "cache_read": 0.10,  "cache_write": 1.25,   "cache_write_1h": 2.0},
+    {"family": "claude-haiku-3-5",  "input": 0.80,  "output": 4.0,   "cache_read": 0.08,  "cache_write": 1.00,   "cache_write_1h": 1.60},
+    {"family": "claude-haiku-3",    "input": 0.25,  "output": 1.25,  "cache_read": 0.03,  "cache_write": 0.30,   "cache_write_1h": 0.50},
 ]
 
 SONNET_RATES = RATES[4]  # fallback
@@ -49,13 +51,15 @@ def get_rates(model: str) -> dict:
     return SONNET_RATES
 
 
-def calculate_cost(model: str, input_t: int, output_t: int, cache_read: int, cache_write: int) -> float:
+def calculate_cost(model: str, input_t: int, output_t: int, cache_read: int,
+                   cache_write: int, cache_write_1h: int = 0) -> float:
     r = get_rates(model)
     return (
         input_t / 1_000_000 * r["input"]
         + output_t / 1_000_000 * r["output"]
         + cache_read / 1_000_000 * r["cache_read"]
         + cache_write / 1_000_000 * r["cache_write"]
+        + cache_write_1h / 1_000_000 * r["cache_write_1h"]
     )
 
 
@@ -122,6 +126,84 @@ def parse_remote_events(host: str) -> list[dict]:
     return all_events
 
 
+HOOK_FILE_SEP = "___CCTRACK_HOOK_FILE___"
+MONTHLY_FILE_SEP = "___CCTRACK_MONTHLY_FILE___"
+
+# Hook data locations on remote hosts: the host's own cctrack dir plus any
+# Sandy sandbox equivalents (mirrors discover_hook_data_dirs).
+REMOTE_HOOK_DIRS = "~/.claude/cctrack ~/.sandy/sandboxes"
+
+
+def fetch_remote_hook_data(host: str) -> str:
+    """SSH to host and cat all statusline JSONL + monthly summary files.
+    Returns raw output with file-type separators."""
+    remote_cmd = (
+        f"find {REMOTE_HOOK_DIRS} -name 'statusline*.jsonl' "
+        f"-exec echo '{HOOK_FILE_SEP}' \\; -exec cat {{}} \\; 2>/dev/null; "
+        f"find {REMOTE_HOOK_DIRS} -name 'monthly-*.json' "
+        f"-exec echo '{MONTHLY_FILE_SEP}' \\; -exec cat {{}} \\; 2>/dev/null; "
+        f"true"
+    )
+    cmd = ["ssh", "-o", "ConnectTimeout=10", "-o", "BatchMode=yes", host, remote_cmd]
+    try:
+        result = subprocess.run(cmd, capture_output=True, text=True, timeout=60)
+        if result.returncode != 0 and not result.stdout:
+            print(f"  Warning: SSH to {host} failed: {result.stderr.strip()}", file=sys.stderr)
+            return ""
+        return result.stdout
+    except subprocess.TimeoutExpired:
+        print(f"  Warning: SSH to {host} timed out", file=sys.stderr)
+        return ""
+    except FileNotFoundError:
+        print(f"  Warning: ssh command not found", file=sys.stderr)
+        return ""
+
+
+def parse_remote_hook_data(host: str) -> tuple[list[dict], dict]:
+    """Fetch and parse hook data from a remote host.
+    Returns (hook_events, monthly_summaries) in the same shapes as
+    parse_hook_events and load_monthly_hook_summaries."""
+    raw = fetch_remote_hook_data(host)
+    if not raw:
+        return [], {}
+
+    # Statusline chunks come first, monthly chunks after (find order above).
+    head, _, monthly_raw = raw.partition(MONTHLY_FILE_SEP)
+
+    lines: list[str] = []
+    for chunk in head.split(HOOK_FILE_SEP):
+        lines.extend(chunk.splitlines())
+    events = _hook_events_from_lines(lines)
+
+    monthly: dict[str, dict] = {}
+    for chunk in monthly_raw.split(MONTHLY_FILE_SEP):
+        chunk = chunk.strip()
+        if not chunk:
+            continue
+        try:
+            data = json.loads(chunk)
+        except (json.JSONDecodeError, ValueError):
+            continue
+        for date_str, day_data in data.get("days", {}).items():
+            entry = {
+                "cost": float(day_data.get("cost", 0.0)),
+                "sessions": int(day_data.get("sessions", 0)),
+            }
+            merge_monthly_summaries(monthly, {date_str: entry})
+    return events, monthly
+
+
+def merge_monthly_summaries(target: dict, source: dict) -> None:
+    """Merge per-date {cost, sessions} summaries from source into target,
+    summing where dates overlap (e.g. same day on multiple hosts)."""
+    for date_str, data in source.items():
+        if date_str in target:
+            target[date_str]["cost"] += data["cost"]
+            target[date_str]["sessions"] += data["sessions"]
+        else:
+            target[date_str] = dict(data)
+
+
 # ── Hook infrastructure ───────────────────────────────────────────────
 
 HOOK_SCRIPT = """\
@@ -432,9 +514,19 @@ def parse_lines(lines: list[str]) -> list[dict]:
         input_t = usage.get("input_tokens", 0)
         output_t = usage.get("output_tokens", 0)
         cache_read = usage.get("cache_read_input_tokens", 0)
-        cache_write = usage.get("cache_creation_input_tokens", 0)
 
-        if input_t == 0 and output_t == 0 and cache_read == 0 and cache_write == 0:
+        # Prefer nested cache_creation.{ephemeral_5m,ephemeral_1h}_input_tokens
+        # — 1h cache writes cost 2× input vs 1.25× for 5m, and Claude Code
+        # commonly uses the 1h tier. Fall back to flat cache_creation_input_tokens
+        # as all-5m for older logs that didn't emit the split.
+        cc = usage.get("cache_creation") or {}
+        cache_write = cc.get("ephemeral_5m_input_tokens", 0)
+        cache_write_1h = cc.get("ephemeral_1h_input_tokens", 0)
+        if not cc:
+            cache_write = usage.get("cache_creation_input_tokens", 0)
+
+        if (input_t == 0 and output_t == 0 and cache_read == 0
+                and cache_write == 0 and cache_write_1h == 0):
             continue
 
         entry = {
@@ -444,6 +536,7 @@ def parse_lines(lines: list[str]) -> list[dict]:
             "output": output_t,
             "cache_read": cache_read,
             "cache_write": cache_write,
+            "cache_write_1h": cache_write_1h,
         }
 
         req_id = event.get("requestId", "")
@@ -470,61 +563,47 @@ def parse_events(files: list[str]) -> list[dict]:
 
 # ── Hook data parsing ──────────────────────────────────────────────────
 
-def parse_hook_events(cctrack_dir: Path | None = None) -> list[dict]:
-    """Parse statusline-*.jsonl files. Dedup by session_id (last entry by _ts wins).
-    Also reads legacy statusline.jsonl for migration.
-    Returns list of {date_str, cost_usd, input_tokens, output_tokens, cache_read, cache_write, model, session_id}."""
-    if cctrack_dir is None:
-        cctrack_dir = CCTRACK_DIR
-    if not cctrack_dir.exists():
-        return []
+def _hook_events_from_lines(lines) -> list[dict]:
+    """Parse statusline JSONL lines. For each (session_id, date) pair, keeps the
+    entry with the latest _ts — the session's cumulative values as of end-of-day
+    for that date. Returns list of {date_str, session_id, cost_usd, input_tokens,
+    output_tokens, cache_read, cache_write, model}."""
+    by_session_day: dict[tuple[str, str], dict] = {}
 
-    by_session: dict[str, dict] = {}
-
-    files = sorted(cctrack_dir.glob("statusline-*.jsonl"))
-    legacy = cctrack_dir / "statusline.jsonl"
-    if legacy.exists():
-        files.append(legacy)
-
-    for filepath in files:
+    for line in lines:
+        line = line.strip()
+        if not line:
+            continue
         try:
-            with open(filepath, "r", errors="replace") as f:
-                for line in f:
-                    line = line.strip()
-                    if not line:
-                        continue
-                    try:
-                        entry = json.loads(line)
-                    except (json.JSONDecodeError, ValueError):
-                        continue
-
-                    session_id = entry.get("session_id", "")
-                    if not session_id:
-                        continue
-
-                    ts = entry.get("_ts", "")
-                    if not ts:
-                        continue
-
-                    if session_id in by_session:
-                        existing_ts = by_session[session_id].get("_ts", "")
-                        if ts <= existing_ts:
-                            continue
+            entry = json.loads(line)
+        except (json.JSONDecodeError, ValueError):
+            continue
 
-                    by_session[session_id] = entry
-        except (OSError, IOError):
+        session_id = entry.get("session_id", "")
+        if not session_id:
             continue
 
-    results = []
-    for session_id, entry in by_session.items():
         ts = entry.get("_ts", "")
+        if not ts:
+            continue
+
         try:
             dt = datetime.fromisoformat(ts.replace("Z", "+00:00"))
         except (ValueError, AttributeError):
             continue
+        date_str = dt.strftime("%Y-%m-%d")
 
+        key = (session_id, date_str)
+        if key in by_session_day:
+            if ts <= by_session_day[key].get("_ts", ""):
+                continue
+
+        by_session_day[key] = entry
+
+    results = []
+    for (session_id, date_str), entry in by_session_day.items():
         results.append({
-            "date_str": dt.strftime("%Y-%m-%d"),
+            "date_str": date_str,
             "cost_usd": float(entry.get("cost_usd", 0.0)),
             "input_tokens": int(entry.get("input_tokens", 0)),
             "output_tokens": int(entry.get("output_tokens", 0)),
@@ -537,6 +616,75 @@ def parse_hook_events(cctrack_dir: Path | None = None) -> list[dict]:
     return results
 
 
+def parse_hook_events(cctrack_dir: Path | None = None) -> list[dict]:
+    """Parse statusline-*.jsonl files from a local hook data directory.
+    Also reads legacy statusline.jsonl for migration. See _hook_events_from_lines
+    for the per-(session, date) dedup semantics."""
+    if cctrack_dir is None:
+        cctrack_dir = CCTRACK_DIR
+    if not cctrack_dir.exists():
+        return []
+
+    files = sorted(cctrack_dir.glob("statusline-*.jsonl"))
+    legacy = cctrack_dir / "statusline.jsonl"
+    if legacy.exists():
+        files.append(legacy)
+
+    lines: list[str] = []
+    for filepath in files:
+        try:
+            with open(filepath, "r", errors="replace") as f:
+                lines.extend(f.readlines())
+        except (OSError, IOError):
+            continue
+
+    return _hook_events_from_lines(lines)
+
+
+def compute_hook_deltas(events: list[dict]) -> list[dict]:
+    """Convert per-(session, date) cumulative records into per-(session, date)
+    daily deltas. For each session, sorts its records by date and computes
+    day-over-day deltas for cost_usd, input_tokens, output_tokens. The first
+    visible day for a session uses 0 as the prior cumulative. Cache fields are
+    snapshot (not cumulative), so they're passed through unchanged. Records
+    with zero cost delta and zero token delta are dropped."""
+    by_session: dict[str, list[dict]] = defaultdict(list)
+    for e in events:
+        by_session[e["session_id"]].append(e)
+
+    result = []
+    for sid, session_events in by_session.items():
+        session_events.sort(key=lambda e: e["date_str"])
+        prev_cost = 0.0
+        prev_in = 0
+        prev_out = 0
+        for e in session_events:
+            cost = float(e.get("cost_usd", 0.0))
+            in_t = int(e.get("input_tokens", 0))
+            out_t = int(e.get("output_tokens", 0))
+
+            d_cost = max(0.0, cost - prev_cost)
+            d_in = max(0, in_t - prev_in)
+            d_out = max(0, out_t - prev_out)
+
+            prev_cost, prev_in, prev_out = cost, in_t, out_t
+
+            if d_cost == 0.0 and d_in == 0 and d_out == 0:
+                continue
+
+            result.append({
+                "date_str": e["date_str"],
+                "session_id": sid,
+                "cost_usd": d_cost,
+                "input_tokens": d_in,
+                "output_tokens": d_out,
+                "cache_read": int(e.get("cache_read", 0)),
+                "cache_write": int(e.get("cache_write", 0)),
+                "model": e.get("model", ""),
+            })
+    return result
+
+
 def load_monthly_hook_summaries(cctrack_dir: Path | None = None) -> dict:
     """Load pre-aggregated monthly-YYYY-MM.json summary files.
     Returns {date_str: {cost: float, sessions: int}}."""
@@ -561,7 +709,10 @@ def load_monthly_hook_summaries(cctrack_dir: Path | None = None) -> dict:
 
 def rollup_old_hook_files(cctrack_dir: Path | None = None,
                           retention_days: int = HOOK_RETENTION_DAYS) -> None:
-    """Roll up daily JSONL files older than retention_days into monthly summaries, then delete them."""
+    """Roll up daily JSONL files older than retention_days into monthly summaries,
+    then delete them. Computes per-session daily deltas within the rollup pass so
+    a session spanning multiple days contributes correctly to each day (rather
+    than double-counting cumulative values across days)."""
     if cctrack_dir is None:
         cctrack_dir = CCTRACK_DIR
     if not cctrack_dir.exists():
@@ -569,12 +720,21 @@ def rollup_old_hook_files(cctrack_dir: Path | None = None,
 
     cutoff = (datetime.now(timezone.utc) - timedelta(days=retention_days)).strftime("%Y-%m-%d")
 
+    old_paths: list[tuple[str, Path]] = []
     for path in sorted(cctrack_dir.glob("statusline-*.jsonl")):
         date_str = path.stem.replace("statusline-", "")
         if not date_str or date_str >= cutoff:
             continue
+        old_paths.append((date_str, path))
+
+    if not old_paths:
+        return
 
-        by_session: dict[str, dict] = {}
+    # Parse all old files into per-(session, date) latest cumulative snapshots.
+    by_session_day: dict[tuple[str, str], dict] = {}
+    empty_paths: list[Path] = []
+    for date_str, path in old_paths:
+        had_any_row = False
         try:
             with open(path, "r", errors="replace") as f:
                 for line in f:
@@ -589,35 +749,68 @@ def rollup_old_hook_files(cctrack_dir: Path | None = None,
                     ts = entry.get("_ts", "")
                     if not sid or not ts:
                         continue
-                    if sid in by_session and ts <= by_session[sid].get("_ts", ""):
+                    had_any_row = True
+                    key = (sid, date_str)
+                    if key in by_session_day and ts <= by_session_day[key].get("_ts", ""):
                         continue
-                    by_session[sid] = entry
+                    by_session_day[key] = entry
         except (OSError, IOError):
             continue
 
-        if not by_session:
-            try:
-                path.unlink()
-            except OSError:
-                pass
-            continue
-
-        day_cost = sum(float(e.get("cost_usd", 0.0)) for e in by_session.values())
-        day_sessions = len(by_session)
-
+        if not had_any_row:
+            empty_paths.append(path)
+
+    # Compute per-session per-day deltas by walking each session chronologically.
+    by_session: dict[str, list[tuple[str, dict]]] = defaultdict(list)
+    for (sid, date_str), entry in by_session_day.items():
+        by_session[sid].append((date_str, entry))
+
+    day_costs: dict[str, float] = defaultdict(float)
+    day_session_ids: dict[str, set] = defaultdict(set)
+
+    for sid, records in by_session.items():
+        records.sort(key=lambda r: r[0])
+        prev_cost = 0.0
+        for date_str, entry in records:
+            cost = float(entry.get("cost_usd", 0.0))
+            delta = max(0.0, cost - prev_cost)
+            prev_cost = cost
+            day_costs[date_str] += delta
+            day_session_ids[date_str].add(sid)
+
+    # Write per-day deltas into the appropriate monthly file and delete dailies.
+    by_month: dict[str, dict] = {}
+    for date_str in sorted(day_costs.keys()):
         month_str = date_str[:7]
-        monthly_path = cctrack_dir / f"monthly-{month_str}.json"
-        monthly_data: dict = {"month": month_str, "days": {}}
-        if monthly_path.exists():
-            try:
-                monthly_data = json.loads(monthly_path.read_text())
-            except (json.JSONDecodeError, OSError):
-                monthly_data = {"month": month_str, "days": {}}
+        if month_str not in by_month:
+            monthly_path = cctrack_dir / f"monthly-{month_str}.json"
+            if monthly_path.exists():
+                try:
+                    by_month[month_str] = json.loads(monthly_path.read_text())
+                except (json.JSONDecodeError, OSError):
+                    by_month[month_str] = {"month": month_str, "days": {}}
+            else:
+                by_month[month_str] = {"month": month_str, "days": {}}
+            by_month[month_str].setdefault("days", {})
+        by_month[month_str]["days"][date_str] = {
+            "cost": day_costs[date_str],
+            "sessions": len(day_session_ids[date_str]),
+        }
 
-        monthly_data["days"][date_str] = {"cost": day_cost, "sessions": day_sessions}
+    for month_str, data in by_month.items():
+        monthly_path = cctrack_dir / f"monthly-{month_str}.json"
+        try:
+            monthly_path.write_text(json.dumps(data, indent=2) + "\n")
+        except OSError:
+            pass
 
+    for _, path in old_paths:
+        try:
+            path.unlink()
+        except OSError:
+            pass
+    for path in empty_paths:
         try:
-            monthly_path.write_text(json.dumps(monthly_data, indent=2) + "\n")
             path.unlink()
         except OSError:
             pass
@@ -625,21 +818,38 @@ def rollup_old_hook_files(cctrack_dir: Path | None = None,
 
 def aggregate_hook_data(events: list[dict], monthly_summaries: dict | None = None) -> dict:
     """Aggregate hook events by date, merging with pre-aggregated monthly data.
-    Returns {date_str: {cost: float, sessions: int}}."""
+    Returns {date_str: {cost, sessions, input_tokens, output_tokens, cache_read, cache_write, models}}.
+    Note: cache_read/cache_write come from current_usage (snapshot of the last message's
+    context), so aggregating them is not strictly meaningful — treat as indicative only."""
+    def _empty() -> dict:
+        return {"cost": 0.0, "sessions": 0, "input_tokens": 0, "output_tokens": 0,
+                "cache_read": 0, "cache_write": 0, "models": {}}
+
     by_date: dict[str, dict] = {}
 
     if monthly_summaries:
         for date_str, data in monthly_summaries.items():
-            by_date[date_str] = {"cost": data["cost"], "sessions": data["sessions"]}
+            bucket = _empty()
+            bucket["cost"] = data["cost"]
+            bucket["sessions"] = data["sessions"]
+            by_date[date_str] = bucket
 
     recent_dates: set[str] = set()
     for e in events:
         date_str = e["date_str"]
         if date_str not in recent_dates:
-            by_date[date_str] = {"cost": 0.0, "sessions": 0}
+            by_date[date_str] = _empty()
             recent_dates.add(date_str)
-        by_date[date_str]["cost"] += e["cost_usd"]
-        by_date[date_str]["sessions"] += 1
+        bucket = by_date[date_str]
+        bucket["cost"] += e["cost_usd"]
+        bucket["sessions"] += 1
+        bucket["input_tokens"] += e.get("input_tokens", 0)
+        bucket["output_tokens"] += e.get("output_tokens", 0)
+        bucket["cache_read"] += e.get("cache_read", 0)
+        bucket["cache_write"] += e.get("cache_write", 0)
+        model = e.get("model", "")
+        if model:
+            bucket["models"][model] = bucket["models"].get(model, 0) + 1
 
     return by_date
 
@@ -647,14 +857,15 @@ def aggregate_hook_data(events: list[dict], monthly_summaries: dict | None = Non
 # ── Aggregation ─────────────────────────────────────────────────────────
 
 def new_bucket() -> dict:
-    return {"input": 0, "output": 0, "cache_read": 0, "cache_write": 0, "cost": 0.0}
+    return {"input": 0, "output": 0, "cache_read": 0, "cache_write": 0,
+            "cache_write_1h": 0, "cost": 0.0, "models": {}}
 
 
 def aggregate(events: list[dict]) -> tuple[dict, dict]:
     """
     Returns:
-        daily: {date_str: {input, output, cache_read, cache_write, cost}}
-        monthly: {month_str: {input, output, cache_read, cache_write, cost}}
+        daily: {date_str: {input, output, cache_read, cache_write, cache_write_1h, cost, models}}
+        monthly: {month_str: {input, output, cache_read, cache_write, cache_write_1h, cost, models}}
     """
     daily = defaultdict(new_bucket)
     monthly = defaultdict(new_bucket)
@@ -672,14 +883,20 @@ def aggregate(events: list[dict]) -> tuple[dict, dict]:
         date_str = dt.strftime("%Y-%m-%d")
         month_str = dt.strftime("%Y-%m")
 
-        cost = calculate_cost(e["model"], e["input"], e["output"], e["cache_read"], e["cache_write"])
+        cache_1h = e.get("cache_write_1h", 0)
+        cost = calculate_cost(e["model"], e["input"], e["output"],
+                              e["cache_read"], e["cache_write"], cache_1h)
+        model = e.get("model", "")
 
         for bucket in (daily[date_str], monthly[month_str]):
             bucket["input"] += e["input"]
             bucket["output"] += e["output"]
             bucket["cache_read"] += e["cache_read"]
             bucket["cache_write"] += e["cache_write"]
+            bucket["cache_write_1h"] += cache_1h
             bucket["cost"] += cost
+            if model:
+                bucket["models"][model] = bucket["models"].get(model, 0) + 1
 
     return dict(daily), dict(monthly)
 
@@ -690,8 +907,26 @@ def format_tokens(n: int) -> str:
     return f"{n:,}"
 
 
+def format_model(models: dict) -> str:
+    """Format a model counter dict as 'short-name' or 'short-name+N' when
+    multiple models appear. Returns empty string for no models.
+    Strips the 'claude-' prefix and any trailing YYYYMMDD date suffix."""
+    if not models:
+        return ""
+    ranked = sorted(models.items(), key=lambda kv: (-kv[1], kv[0]))
+    primary = ranked[0][0]
+    short = primary[len("claude-"):] if primary.startswith("claude-") else primary
+    parts = short.split("-")
+    if len(parts) > 1 and parts[-1].isdigit() and len(parts[-1]) == 8:
+        short = "-".join(parts[:-1])
+    if len(ranked) > 1:
+        short += f"+{len(ranked) - 1}"
+    return short
+
+
 def total_tokens(b: dict) -> int:
-    return b["input"] + b["output"] + b["cache_read"] + b["cache_write"]
+    return (b["input"] + b["output"] + b["cache_read"]
+            + b["cache_write"] + b.get("cache_write_1h", 0))
 
 
 def _compute_accuracy_factor(daily: dict, hook_daily: dict, hook_install_date: str | None) -> tuple[float, bool]:
@@ -796,12 +1031,25 @@ def print_report(daily: dict, monthly: dict, days: int | None = None,
         print(f"{current_month_name} — month to date (day {day_of_month}, {days_active} active)")
         print("───────────────────────────────────────")
         if m:
-            print(f"  Input tokens:   {format_tokens(m['input']):>15}")
-            print(f"  Output tokens:  {format_tokens(m['output']):>15}")
-            if m['cache_read'] or m['cache_write']:
+            hook_in_mtd = sum(hook_daily.get(d, {}).get("input_tokens", 0) for d in current_days_all)
+            hook_out_mtd = sum(hook_daily.get(d, {}).get("output_tokens", 0) for d in current_days_all)
+            input_t = hook_in_mtd if hook_in_mtd else m["input"]
+            output_t = hook_out_mtd if hook_out_mtd else m["output"]
+            print(f"  Input tokens:   {format_tokens(input_t):>15}")
+            print(f"  Output tokens:  {format_tokens(output_t):>15}")
+            cache_w_total = m['cache_write'] + m.get('cache_write_1h', 0)
+            if m['cache_read'] or cache_w_total:
                 print(f"  Cache read:     {format_tokens(m['cache_read']):>15}")
-                print(f"  Cache write:    {format_tokens(m['cache_write']):>15}")
-            print(f"  Total tokens:   {format_tokens(total_tokens(m)):>15}")
+                print(f"  Cache write:    {format_tokens(cache_w_total):>15}")
+            total = input_t + output_t + m['cache_read'] + cache_w_total
+            print(f"  Total tokens:   {format_tokens(total):>15}")
+            mtd_model = format_model(m.get("models", {}))
+            if mtd_model:
+                print(f"  Model:          {mtd_model:>15}")
+
+        mtd_sessions = sum(hook_daily.get(d, {}).get("sessions", 0) for d in current_days_all)
+        if mtd_sessions:
+            print(f"  Sessions:       {mtd_sessions:>15}")
 
         if hook_install_date and est_days > 0:
             factor_source = "measured" if factor_is_measured else "estimated ~20-25%"
@@ -829,26 +1077,32 @@ def print_report(daily: dict, monthly: dict, days: int | None = None,
     if current_month_days:
         print("Daily breakdown:")
         if hook_install_date:
-            print(f"  {'Date':<12} {'Input':>12} {'Output':>12} {'Cache R':>12} {'Cache W':>12} {'Cost':>10} {'':>3}")
-            print(f"  {'─'*12} {'─'*12} {'─'*12} {'─'*12} {'─'*12} {'─'*10} {'─'*3}")
+            print(f"  {'Date':<12} {'Input':>11} {'Output':>11} {'Cache R':>11} {'Cache W':>11} {'Sess':>5} {'Model':<12} {'Cost':>9} {'':>2}")
+            print(f"  {'─'*12} {'─'*11} {'─'*11} {'─'*11} {'─'*11} {'─'*5} {'─'*12} {'─'*9} {'─'*2}")
         else:
-            print(f"  {'Date':<12} {'Input':>12} {'Output':>12} {'Cache R':>12} {'Cache W':>12} {'Cost':>10}")
-            print(f"  {'─'*12} {'─'*12} {'─'*12} {'─'*12} {'─'*12} {'─'*10}")
+            print(f"  {'Date':<12} {'Input':>11} {'Output':>11} {'Cache R':>11} {'Cache W':>11} {'Model':<12} {'Cost':>9}")
+            print(f"  {'─'*12} {'─'*11} {'─'*11} {'─'*11} {'─'*11} {'─'*12} {'─'*9}")
         display_days = current_month_days
         if days is not None:
             display_days = display_days[:days]
         for date_str in display_days:
             cost, is_auth = _date_cost(date_str)
             d = daily.get(date_str)
-            input_t = d["input"] if d else 0
-            output_t = d["output"] if d else 0
+            hook = hook_daily.get(date_str, {})
+            hook_in = hook.get("input_tokens", 0)
+            hook_out = hook.get("output_tokens", 0)
+            input_t = hook_in if hook_in else (d["input"] if d else 0)
+            output_t = hook_out if hook_out else (d["output"] if d else 0)
             cache_r = d["cache_read"] if d else 0
-            cache_w = d["cache_write"] if d else 0
+            cache_w = (d["cache_write"] + d.get("cache_write_1h", 0)) if d else 0
+            model_str = format_model(d.get("models", {})) if d else format_model(hook.get("models", {}))
             if hook_install_date:
                 marker = "" if is_auth else "~"
-                print(f"  {date_str:<12} {format_tokens(input_t):>12} {format_tokens(output_t):>12} {format_tokens(cache_r):>12} {format_tokens(cache_w):>12} ${cost:>8.2f} {marker:>3}")
+                sessions = hook.get("sessions", 0)
+                sess_str = str(sessions) if sessions else "-"
+                print(f"  {date_str:<12} {format_tokens(input_t):>11} {format_tokens(output_t):>11} {format_tokens(cache_r):>11} {format_tokens(cache_w):>11} {sess_str:>5} {model_str:<12} ${cost:>7.2f} {marker:>2}")
             else:
-                print(f"  {date_str:<12} {format_tokens(input_t):>12} {format_tokens(output_t):>12} {format_tokens(cache_r):>12} {format_tokens(cache_w):>12} ${cost:>8.2f}")
+                print(f"  {date_str:<12} {format_tokens(input_t):>11} {format_tokens(output_t):>11} {format_tokens(cache_r):>11} {format_tokens(cache_w):>11} {model_str:<12} ${cost:>7.2f}")
 
     if hook_install_date and current_month_days:
         has_estimated = any(not _date_cost(d)[1] for d in current_month_days)
@@ -863,8 +1117,8 @@ def print_report(daily: dict, monthly: dict, days: int | None = None,
     if prev_months:
         print()
         print("Previous months:")
-        print(f"  {'Month':<12} {'Input':>12} {'Output':>12} {'Cache R':>12} {'Cache W':>12} {'Cost':>10}")
-        print(f"  {'─'*12} {'─'*12} {'─'*12} {'─'*12} {'─'*12} {'─'*10}")
+        print(f"  {'Month':<12} {'Input':>11} {'Output':>11} {'Cache R':>11} {'Cache W':>11} {'Model':<12} {'Cost':>9}")
+        print(f"  {'─'*12} {'─'*11} {'─'*11} {'─'*11} {'─'*11} {'─'*12} {'─'*9}")
         for month_str in prev_months:
             m = monthly.get(month_str)
             month_days_list = [d for d in prev_month_days if d.startswith(month_str)]
@@ -890,12 +1144,13 @@ def print_report(daily: dict, monthly: dict, days: int | None = None,
             input_t = m["input"] if m else 0
             output_t = m["output"] if m else 0
             cache_r = m["cache_read"] if m else 0
-            cache_w = m["cache_write"] if m else 0
+            cache_w = (m["cache_write"] + m.get("cache_write_1h", 0)) if m else 0
+            model_str = format_model(m.get("models", {})) if m else ""
 
-            cost_str = f"${month_cost:>8.2f}"
+            cost_str = f"${month_cost:>7.2f}"
             if hook_install_date and month_has_est and not month_has_hook:
                 cost_str += "  ~"
-            print(f"  {month_str + days_str:<12} {format_tokens(input_t):>12} {format_tokens(output_t):>12} {format_tokens(cache_r):>12} {format_tokens(cache_w):>12} {cost_str}")
+            print(f"  {month_str + days_str:<12} {format_tokens(input_t):>11} {format_tokens(output_t):>11} {format_tokens(cache_r):>11} {format_tokens(cache_w):>11} {model_str:<12} {cost_str}")
 
     print()
 
@@ -969,21 +1224,33 @@ def main():
         except (ValueError, AttributeError):
             pass
 
-    hook_dirs = discover_hook_data_dirs()
-    if hook_dirs:
-        all_hook_events: list[dict] = []
-        all_monthly: dict[str, dict] = {}
-        for hdir in hook_dirs:
-            rollup_old_hook_files(hdir)
-            all_hook_events.extend(parse_hook_events(hdir))
-            for date_str, data in load_monthly_hook_summaries(hdir).items():
-                if date_str in all_monthly:
-                    all_monthly[date_str]["cost"] += data["cost"]
-                    all_monthly[date_str]["sessions"] += data["sessions"]
-                else:
-                    all_monthly[date_str] = dict(data)
-        if all_hook_events or all_monthly:
-            hook_daily = aggregate_hook_data(all_hook_events, all_monthly)
+    all_hook_events: list[dict] = []
+    all_monthly: dict[str, dict] = {}
+
+    for hdir in discover_hook_data_dirs():
+        rollup_old_hook_files(hdir)
+        all_hook_events.extend(compute_hook_deltas(parse_hook_events(hdir)))
+        merge_monthly_summaries(all_monthly, load_monthly_hook_summaries(hdir))
+
+    # Remote hook data: each host's deltas are computed independently so a
+    # session's cumulative history never mixes across hosts.
+    if args.remote:
+        for host in args.remote:
+            remote_hook_events, remote_monthly = parse_remote_hook_data(host)
+            all_hook_events.extend(compute_hook_deltas(remote_hook_events))
+            merge_monthly_summaries(all_monthly, remote_monthly)
+
+    if all_hook_events or all_monthly:
+        hook_daily = aggregate_hook_data(all_hook_events, all_monthly)
+
+    # Hook coverage may start earlier than the local install date (e.g. a
+    # remote host installed the hook first). Treat the earliest observed
+    # hook date as the effective install date so those days count as
+    # authoritative rather than being scaled by the accuracy factor.
+    if hook_daily:
+        earliest_hook_date = min(hook_daily.keys())
+        if hook_install_date is None or earliest_hook_date < hook_install_date:
+            hook_install_date = earliest_hook_date
 
     # Suggest hook if not installed
     if not is_hook_installed():

{cctrack-0.3.0 → cctrack-0.4.1}/tests/test_cctrack.py

@@ -172,6 +172,62 @@ def test_calculate_cost_zero_tokens():
     assert cost == 0.0
 
 
+def test_calculate_cost_1h_cache_write():
+    # Opus 4.7: 1M 1h-cache-write tokens at $10/M = $10.00 (2× input rate)
+    cost = cctrack.calculate_cost("claude-opus-4-7", 0, 0, 0, 0, 1_000_000)
+    assert abs(cost - 10.0) < 0.001
+
+
+def test_calculate_cost_1h_vs_5m_cache_write_differ():
+    # Same 1M tokens priced differently depending on TTL.
+    cost_5m = cctrack.calculate_cost("claude-opus-4-7", 0, 0, 0, 1_000_000, 0)
+    cost_1h = cctrack.calculate_cost("claude-opus-4-7", 0, 0, 0, 0, 1_000_000)
+    assert abs(cost_5m - 6.25) < 0.001
+    assert abs(cost_1h - 10.0) < 0.001
+
+
+def test_get_rates_include_1h_cache_write():
+    # Every family should expose a cache_write_1h rate at 2× input.
+    for r in cctrack.RATES:
+        assert "cache_write_1h" in r
+        assert abs(r["cache_write_1h"] - 2.0 * r["input"]) < 0.001
+
+
+def test_parse_lines_extracts_1h_cache_split():
+    # Nested cache_creation.ephemeral_{5m,1h}_input_tokens should be picked up.
+    event = make_event()
+    event["message"]["usage"]["cache_creation_input_tokens"] = 1000
+    event["message"]["usage"]["cache_creation"] = {
+        "ephemeral_5m_input_tokens": 300,
+        "ephemeral_1h_input_tokens": 700,
+    }
+    events = cctrack.parse_lines([json.dumps(event)])
+    assert len(events) == 1
+    assert events[0]["cache_write"] == 300
+    assert events[0]["cache_write_1h"] == 700
+
+
+def test_parse_lines_flat_cache_creation_treated_as_5m():
+    # Legacy logs without the nested object treat the flat field as 5m.
+    event = make_event(cache_write=1000)
+    events = cctrack.parse_lines([json.dumps(event)])
+    assert len(events) == 1
+    assert events[0]["cache_write"] == 1000
+    assert events[0]["cache_write_1h"] == 0
+
+
+def test_aggregate_1h_cache_cost():
+    # 1M 1h-cache-write tokens on opus-4-7 → $10 via aggregation.
+    events = [
+        {"model": "claude-opus-4-7", "timestamp": "2026-03-15T10:00:00Z",
+         "input": 0, "output": 0, "cache_read": 0,
+         "cache_write": 0, "cache_write_1h": 1_000_000},
+    ]
+    daily, monthly = cctrack.aggregate(events)
+    assert abs(daily["2026-03-15"]["cost"] - 10.0) < 0.001
+    assert daily["2026-03-15"]["cache_write_1h"] == 1_000_000
+
+
 # ── File discovery tests ───────────────────────────────────────────────
 
 def test_discover_files_finds_jsonl():

{cctrack-0.3.0 → cctrack-0.4.1}/tests/test_hook_reporting.py

@@ -159,6 +159,207 @@ def test_parse_hook_events_reads_legacy_file():
         assert events[0]["session_id"] == "legacy_1"
 
 
+def test_parse_hook_events_session_spans_multiple_days():
+    """A session active on multiple days returns one record per (session, day)
+    with that day's maximum cumulative values."""
+    with tempfile.TemporaryDirectory() as d:
+        write_hook_jsonl(d, [
+            make_hook_entry(session_id="sA", ts="2026-04-14T10:00:00+00:00", cost_usd=2.00, input_tokens=100, output_tokens=50),
+            make_hook_entry(session_id="sA", ts="2026-04-14T18:00:00+00:00", cost_usd=5.00, input_tokens=300, output_tokens=150),
+        ], date="2026-04-14")
+        write_hook_jsonl(d, [
+            make_hook_entry(session_id="sA", ts="2026-04-15T09:00:00+00:00", cost_usd=8.00, input_tokens=500, output_tokens=250),
+            make_hook_entry(session_id="sA", ts="2026-04-15T16:00:00+00:00", cost_usd=12.00, input_tokens=800, output_tokens=400),
+        ], date="2026-04-15")
+        events = cctrack.parse_hook_events(cctrack_dir=Path(d))
+        by_date = {e["date_str"]: e for e in events}
+        assert len(events) == 2
+        assert by_date["2026-04-14"]["cost_usd"] == 5.00
+        assert by_date["2026-04-14"]["input_tokens"] == 300
+        assert by_date["2026-04-15"]["cost_usd"] == 12.00
+        assert by_date["2026-04-15"]["input_tokens"] == 800
+
+
+# ── compute_hook_deltas tests ──────────────────────────────────────────
+
+def test_compute_hook_deltas_single_day_session():
+    """Session visible on one day produces one delta equal to its cumulative."""
+    events = [
+        {"date_str": "2026-04-15", "session_id": "s1", "cost_usd": 3.0,
+         "input_tokens": 100, "output_tokens": 50, "cache_read": 0, "cache_write": 0, "model": "m"},
+    ]
+    deltas = cctrack.compute_hook_deltas(events)
+    assert len(deltas) == 1
+    assert deltas[0]["cost_usd"] == 3.0
+    assert deltas[0]["input_tokens"] == 100
+    assert deltas[0]["output_tokens"] == 50
+
+
+def test_compute_hook_deltas_session_spans_days():
+    """A session spanning days produces one delta per day, each being the
+    day-over-day increment rather than the cumulative value."""
+    events = [
+        {"date_str": "2026-04-14", "session_id": "sA", "cost_usd": 5.00,
+         "input_tokens": 300, "output_tokens": 150, "cache_read": 0, "cache_write": 0, "model": "m"},
+        {"date_str": "2026-04-15", "session_id": "sA", "cost_usd": 12.00,
+         "input_tokens": 800, "output_tokens": 400, "cache_read": 0, "cache_write": 0, "model": "m"},
+    ]
+    deltas = cctrack.compute_hook_deltas(events)
+    by_date = {d["date_str"]: d for d in deltas}
+    assert abs(by_date["2026-04-14"]["cost_usd"] - 5.00) < 0.001
+    assert abs(by_date["2026-04-15"]["cost_usd"] - 7.00) < 0.001
+    assert by_date["2026-04-14"]["input_tokens"] == 300
+    assert by_date["2026-04-15"]["input_tokens"] == 500
+    assert by_date["2026-04-14"]["output_tokens"] == 150
+    assert by_date["2026-04-15"]["output_tokens"] == 250
+
+
+def test_compute_hook_deltas_drops_zero_records():
+    """A record whose cumulative didn't advance from the prior day is dropped."""
+    events = [
+        {"date_str": "2026-04-14", "session_id": "sA", "cost_usd": 5.0,
+         "input_tokens": 100, "output_tokens": 50, "cache_read": 0, "cache_write": 0, "model": "m"},
+        {"date_str": "2026-04-15", "session_id": "sA", "cost_usd": 5.0,
+         "input_tokens": 100, "output_tokens": 50, "cache_read": 0, "cache_write": 0, "model": "m"},
+    ]
+    deltas = cctrack.compute_hook_deltas(events)
+    assert len(deltas) == 1
+    assert deltas[0]["date_str"] == "2026-04-14"
+
+
+def test_compute_hook_deltas_clamps_negatives():
+    """Cumulative values should never decrease, but if they do (bad data),
+    deltas clamp to zero rather than produce negative costs."""
+    events = [
+        {"date_str": "2026-04-14", "session_id": "sA", "cost_usd": 5.0,
+         "input_tokens": 100, "output_tokens": 50, "cache_read": 0, "cache_write": 0, "model": "m"},
+        {"date_str": "2026-04-15", "session_id": "sA", "cost_usd": 3.0,
+         "input_tokens": 80, "output_tokens": 40, "cache_read": 0, "cache_write": 0, "model": "m"},
+    ]
+    deltas = cctrack.compute_hook_deltas(events)
+    # Day 2 delta is negative, so it's dropped entirely (all-zero after clamping).
+    assert len(deltas) == 1
+    assert deltas[0]["date_str"] == "2026-04-14"
+
+
+def test_compute_hook_deltas_multiple_sessions_independent():
+    """Each session's deltas are computed independently."""
+    events = [
+        {"date_str": "2026-04-14", "session_id": "sA", "cost_usd": 10.0,
+         "input_tokens": 0, "output_tokens": 0, "cache_read": 0, "cache_write": 0, "model": "m"},
+        {"date_str": "2026-04-15", "session_id": "sA", "cost_usd": 15.0,
+         "input_tokens": 0, "output_tokens": 0, "cache_read": 0, "cache_write": 0, "model": "m"},
+        {"date_str": "2026-04-15", "session_id": "sB", "cost_usd": 4.0,
+         "input_tokens": 0, "output_tokens": 0, "cache_read": 0, "cache_write": 0, "model": "m"},
+    ]
+    deltas = cctrack.compute_hook_deltas(events)
+    by_key = {(d["date_str"], d["session_id"]): d["cost_usd"] for d in deltas}
+    assert by_key[("2026-04-14", "sA")] == 10.0
+    assert by_key[("2026-04-15", "sA")] == 5.0
+    assert by_key[("2026-04-15", "sB")] == 4.0
+
+
+# ── Remote hook data tests ─────────────────────────────────────────────
+
+def _remote_raw(hook_chunks: list[list[dict]], monthly_chunks: list[dict]) -> str:
+    """Build a raw SSH output stream the way fetch_remote_hook_data returns it."""
+    parts = []
+    for entries in hook_chunks:
+        parts.append(cctrack.HOOK_FILE_SEP)
+        for e in entries:
+            parts.append(json.dumps(e))
+    for doc in monthly_chunks:
+        parts.append(cctrack.MONTHLY_FILE_SEP)
+        parts.append(json.dumps(doc))
+    return "\n".join(parts) + "\n"
+
+
+def test_parse_remote_hook_data_basic():
+    """Statusline entries and monthly summaries are both parsed from the stream."""
+    raw = _remote_raw(
+        hook_chunks=[
+            [make_hook_entry(session_id="r1", ts="2026-04-15T10:00:00+00:00", cost_usd=3.00)],
+            [make_hook_entry(session_id="r2", ts="2026-04-16T10:00:00+00:00", cost_usd=7.00)],
+        ],
+        monthly_chunks=[
+            {"month": "2026-01", "days": {"2026-01-10": {"cost": 10.0, "sessions": 2}}},
+        ],
+    )
+    with patch("cctrack.fetch_remote_hook_data", return_value=raw):
+        events, monthly = cctrack.parse_remote_hook_data("fakehost")
+    assert len(events) == 2
+    by_session = {e["session_id"]: e for e in events}
+    assert by_session["r1"]["cost_usd"] == 3.00
+    assert by_session["r2"]["date_str"] == "2026-04-16"
+    assert monthly == {"2026-01-10": {"cost": 10.0, "sessions": 2}}
+
+
+def test_parse_remote_hook_data_dedup_across_files():
+    """Same (session, date) in multiple remote files: latest _ts wins."""
+    raw = _remote_raw(
+        hook_chunks=[
+            [make_hook_entry(session_id="rA", ts="2026-04-15T10:00:00+00:00", cost_usd=2.00)],
+            [make_hook_entry(session_id="rA", ts="2026-04-15T18:00:00+00:00", cost_usd=6.00)],
+        ],
+        monthly_chunks=[],
+    )
+    with patch("cctrack.fetch_remote_hook_data", return_value=raw):
+        events, monthly = cctrack.parse_remote_hook_data("fakehost")
+    assert len(events) == 1
+    assert events[0]["cost_usd"] == 6.00
+
+
+def test_parse_remote_hook_data_empty():
+    with patch("cctrack.fetch_remote_hook_data", return_value=""):
+        events, monthly = cctrack.parse_remote_hook_data("fakehost")
+    assert events == []
+    assert monthly == {}
+
+
+def test_parse_remote_hook_data_malformed_monthly_skipped():
+    raw = (cctrack.HOOK_FILE_SEP + "\n"
+           + json.dumps(make_hook_entry(session_id="ok", cost_usd=1.0)) + "\n"
+           + cctrack.MONTHLY_FILE_SEP + "\n"
+           + "not json\n")
+    with patch("cctrack.fetch_remote_hook_data", return_value=raw):
+        events, monthly = cctrack.parse_remote_hook_data("fakehost")
+    assert len(events) == 1
+    assert monthly == {}
+
+
+def test_fetch_remote_hook_data_timeout():
+    with patch("subprocess.run", side_effect=cctrack.subprocess.TimeoutExpired("ssh", 60)):
+        result = cctrack.fetch_remote_hook_data("fakehost")
+    assert result == ""
+
+
+def test_merge_monthly_summaries_sums_overlapping_dates():
+    """Same date on two hosts sums cost and sessions."""
+    target = {"2026-01-10": {"cost": 10.0, "sessions": 2}}
+    cctrack.merge_monthly_summaries(target, {
+        "2026-01-10": {"cost": 5.0, "sessions": 1},
+        "2026-01-11": {"cost": 3.0, "sessions": 1},
+    })
+    assert target["2026-01-10"] == {"cost": 15.0, "sessions": 3}
+    assert target["2026-01-11"] == {"cost": 3.0, "sessions": 1}
+
+
+def test_remote_and_local_hook_costs_sum_per_date():
+    """Regression: a date covered by hooks on two hosts must sum both hosts'
+    costs, not keep just one host's value."""
+    local_events = cctrack.compute_hook_deltas([
+        {"date_str": "2026-06-11", "session_id": "local_s", "cost_usd": 9.05,
+         "input_tokens": 0, "output_tokens": 0, "cache_read": 0, "cache_write": 0, "model": "m"},
+    ])
+    remote_events = cctrack.compute_hook_deltas([
+        {"date_str": "2026-06-11", "session_id": "air_s", "cost_usd": 340.90,
+         "input_tokens": 0, "output_tokens": 0, "cache_read": 0, "cache_write": 0, "model": "m"},
+    ])
+    hook_daily = cctrack.aggregate_hook_data(local_events + remote_events)
+    assert abs(hook_daily["2026-06-11"]["cost"] - 349.95) < 0.001
+    assert hook_daily["2026-06-11"]["sessions"] == 2
+
+
 # ── aggregate_hook_data tests ──────────────────────────────────────────
 
 def test_aggregate_hook_data_single_day():
@@ -316,6 +517,35 @@ def test_rollup_appends_to_existing_monthly():
         assert data["days"]["2025-12-02"]["cost"] == 7.00
 
 
+def test_rollup_cross_day_session_attributes_deltas():
+    """A session active across multiple rolled-up days should have its
+    cumulative cost attributed as per-day deltas, not double-counted."""
+    with tempfile.TemporaryDirectory() as d:
+        # Session sA: cumulative rises 0 → 5 on day 1, 5 → 12 on day 2.
+        # Session sB: only active on day 2, cumulative rises 0 → 3.
+        write_hook_jsonl(d, [
+            make_hook_entry(session_id="sA", ts="2025-12-01T10:00:00+00:00", cost_usd=2.00),
+            make_hook_entry(session_id="sA", ts="2025-12-01T18:00:00+00:00", cost_usd=5.00),
+        ], date="2025-12-01")
+        write_hook_jsonl(d, [
+            make_hook_entry(session_id="sA", ts="2025-12-02T10:00:00+00:00", cost_usd=12.00),
+            make_hook_entry(session_id="sB", ts="2025-12-02T11:00:00+00:00", cost_usd=3.00),
+        ], date="2025-12-02")
+
+        cctrack.rollup_old_hook_files(cctrack_dir=Path(d), retention_days=0)
+
+        data = json.loads(Path(d, "monthly-2025-12.json").read_text())
+        # Day 1: sA delta = 5.00. Total = 5.00, 1 session.
+        assert abs(data["days"]["2025-12-01"]["cost"] - 5.00) < 0.001
+        assert data["days"]["2025-12-01"]["sessions"] == 1
+        # Day 2: sA delta = 12 - 5 = 7.00, sB delta = 3.00. Total = 10.00, 2 sessions.
+        assert abs(data["days"]["2025-12-02"]["cost"] - 10.00) < 0.001
+        assert data["days"]["2025-12-02"]["sessions"] == 2
+        # Sum equals actual session totals: sA $12 + sB $3 = $15.
+        total = sum(day["cost"] for day in data["days"].values())
+        assert abs(total - 15.00) < 0.001
+
+
 def test_rollup_deletes_empty_daily():
     """Daily file with no valid sessions is just deleted."""
     with tempfile.TemporaryDirectory() as d: