"""Bluesky audience audit (#107). Public appview only, no login, read-only.
For every top-level post of mine (posts.json from list_posts.py): likes, reposts, quotes, replies (post_stats, no classify),
plus my followers. Each unique account is classified ONCE with classify_actor (cached in cache.json).
Writes audit.json. Pauses 0.5 s between calls."""
import importlib.util, json, time, urllib.parse, datetime as dt
from pathlib import Path
sp = importlib.util.spec_from_file_location("core", Path(__file__).with_name("bsky_public.py"))
core = importlib.util.module_from_spec(sp); sp.loader.exec_module(core)
H = Path(__file__).parent
posts = json.loads((H / "posts.json").read_text(encoding="utf-8"))
types = json.loads((H / "types.json").read_text(encoding="utf-8"))
cache_f = H / "cache.json"
cache = json.loads(cache_f.read_text(encoding="utf-8")) if cache_f.exists() else {}
assert all(p["rkey"] in types for p in posts), "unlabelled post"
accounts = {}  # did -> {"handle", "acts": [[rkey, kind], ...], "follower": bool}
per_post = []
for p in posts:
    s = core.post_stats(p["uri"], classify=False)
    time.sleep(0.5)
    dids = []
    for w in s["responders"]:
        a = accounts.setdefault(w["did"], {"handle": w["handle"], "acts": [], "follower": False})
        for k in w["did_what"]:
            a["acts"].append([p["rkey"], k])
        dids.append(w["did"])
    per_post.append({"rkey": p["rkey"], "type": types[p["rkey"]], "created": p["created"], "counts": s["counts"],
                     "responders": dids, "reply_texts": s["reply_texts"]})
cur, followers = "", []
while True:
    j = core._pub("app.bsky.graph.getFollowers", f"actor={core.MY_DID}&limit=100" + (f"&cursor={urllib.parse.quote(cur)}" if cur else ""))
    followers += j.get("followers", [])
    cur = j.get("cursor")
    if not cur or not j.get("followers"):
        break
for f in followers:
    accounts.setdefault(f["did"], {"handle": f["handle"], "acts": [], "follower": False})["follower"] = True
for did, a in accounts.items():
    if did not in cache:
        try:
            cache[did] = core.classify_actor(did)
        except Exception as e:  # noqa: BLE001
            cache[did] = {"verdict": "unclear", "error": str(e)[:120], "auto": [], "human": []}
        time.sleep(0.5)
        cache_f.write_text(json.dumps(cache, indent=1, ensure_ascii=False), encoding="utf-8")
    a["auto_verdict"] = cache[did]["verdict"]
    a["signals"] = {k: cache[did].get(k) for k in ("auto", "human", "followers", "follows", "posts", "age_days", "labels", "bio")}
out = {"run_at": dt.datetime.now(dt.timezone.utc).isoformat(timespec="seconds"), "n_posts": len(posts),
       "posts": per_post, "accounts": accounts, "n_followers": len(followers)}
(H / "audit.json").write_text(json.dumps(out, indent=1, ensure_ascii=False), encoding="utf-8")
print("posts", len(posts), "accounts", len(accounts), "followers", len(followers))
for did, a in sorted(accounts.items(), key=lambda kv: kv[1]["auto_verdict"]):
    s = a["signals"]
    print(f"{a['auto_verdict']:12} @{a['handle']:40} acts={[f'{r[:6]}:{k}' for r, k in a['acts']]} fol={a['follower']} "
          f"| f{s['followers']}/{s['follows']} p{s['posts']} d{s['age_days']} | A:{'; '.join(s['auto'] or [])} | H:{'; '.join(s['human'] or [])} | {(s['bio'] or '')[:70]!r}")