#!/usr/bin/env python3
"""collector.py · Qoder CN 用量采集器
扫会话日志聚合每模型调用量，输出 ~/.stellar-gauge/usage.json。
只读宿主日志，不写宿主任何东西。
"""
import json, os, re, glob, time
from datetime import datetime, timedelta

LOGS_ROOT = os.path.expanduser(
    "~/Library/Application Support/QoderCN/logs")
STATE_FILE = os.path.expanduser("~/.stellar-gauge/usage.json")
OFFSETS_FILE = os.path.expanduser("~/.stellar-gauge/offsets.json")

RE_CTX = re.compile(
    r"(\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2}\.\d+).*Context usage update: "
    r"\{.*?\"usedTokens\":(\d+)")
RE_MODEL = re.compile(r"applyModelRuntimeMeta: model=([A-Za-z0-9_.-]+)")

# Qoder 内部别名 → 展示名
ALIAS = {
    "qmodel_38max": "Qwen3.8-Max",
    "qmodel_37max": "Qwen3.7-Max",
    "dsmodel_v4pro": "DeepSeek-v4-Pro",
}


def load(path, default):
    try:
        with open(path) as f:
            return json.load(f)
    except Exception:
        return default


def scan():
    offsets = load(OFFSETS_FILE, {})
    usage = load(STATE_FILE, {"requests": {}, "tokens": {},
                              "daily": {}, "meta": {}})
    now = datetime.now()
    for f in glob.glob(os.path.join(LOGS_ROOT, "*", "questWindow", "agent.log")):
        try:
            size = os.path.getsize(f)
            off = offsets.get(f, 0)
            if size < off:  # 日志轮转
                off = 0
            if size == off:
                continue
            with open(f, errors="ignore") as fh:
                fh.seek(off)
                chunk = fh.read()
            offsets[f] = size
        except Exception:
            continue
        cur_model = "unknown"
        prev_used = None
        for line in chunk.splitlines():
            m = RE_MODEL.search(line)
            if m:
                cur_model = ALIAS.get(m.group(1), m.group(1))
                continue
            m = RE_CTX.search(line)
            if m:
                ts, used = m.group(1), int(m.group(2))
                # 增量估算：两次上报差值近似为本轮新增 token
                delta = used - prev_used if prev_used else min(used, 8000)
                prev_used = used
                day = ts[:10]
                usage["requests"][cur_model] = \
                    usage["requests"].get(cur_model, 0) + 1
                usage["tokens"][cur_model] = \
                    usage["tokens"].get(cur_model, 0) + max(delta, 0)
                d = usage["daily"].setdefault(day, {})
                d[cur_model] = d.get(cur_model, 0) + max(delta, 0)
    # 只留 14 天
    keep = (now - timedelta(days=14)).strftime("%Y-%m-%d")
    usage["daily"] = {k: v for k, v in usage["daily"].items() if k >= keep}
    usage["meta"] = {"updated_at": now.isoformat(timespec="seconds"),
                     "note": "token 为本地增量估算，非计费 Credits"}
    tmp = STATE_FILE + ".tmp"
    with open(tmp, "w") as f:
        json.dump(usage, f, ensure_ascii=False, indent=1)
    os.replace(tmp, STATE_FILE)
    with open(OFFSETS_FILE, "w") as f:
        json.dump(offsets, f)
    return usage


if __name__ == "__main__":
    u = scan()
    print(json.dumps({"requests": u["requests"],
                      "meta": u["meta"]}, ensure_ascii=False, indent=1))