import json, hashlib, random
random.seed(14)

def fp(r):
    s = r["state"] if isinstance(r["state"], str) else json.dumps(r["state"], sort_keys=True)
    q = json.dumps(r["questions"], sort_keys=True)
    return hashlib.md5((s + q).encode()).hexdigest()

spec = [("data/adjud_train.jsonl", None),
        ("data/adjud_rendered.jsonl", None),
        ("data/jb7_train.jsonl", 6000), ("data/jb6_train.jsonl", 5000),
        ("data/sft_v6.jsonl", 5000), ("data/sft_v8b.jsonl", 4000),
        ("data/hf4.jsonl", 4000), ("data/hf3b.jsonl", 3000),
        ("data/hf2b.jsonl", 3000), ("data/hf5_bb.jsonl", 2000),
        ("data/rl_sft_v9.jsonl", 3000), ("data/sft_v9.jsonl", 2500)]

seen = set()
out = []
for path, cap in spec:
    try:
        rows = [json.loads(l) for l in open(path)]
    except FileNotFoundError:
        print("missing", path)
        continue
    random.shuffle(rows)
    take = rows if cap is None else rows[:cap]
    kept = 0
    for r in take:
        h = fp(r)
        if h in seen:
            continue
        seen.add(h)
        tag = path.split("/")[-1].replace(".jsonl", "")
        r["id"] = "v14:" + tag + ":" + str(kept) + ":" + h[:8]
        out.append(r)
        kept += 1
    print(path, kept)
random.shuffle(out)
with open("data/v14_mix.jsonl", "w") as f:
    for r in out:
        f.write(json.dumps(r, ensure_ascii=False) + "\n")
print("total", len(out))