这一页直接向公开节点提问，把链上的原始数据取回来显示。它不经过 scan.botchain.ai，
也不经过任何会改写数据的后端索引。请求的是
https://rpc.botchain.ai
顶栏徽章写着这一页当前的数据来自哪一层：① 直连主网 677 → ② 仓库自带的只读代理（原样中转，不改数据）→ ③ 测试网 968（历史留痕，非主网）→ ④ 离线快照（非实时）。
降级会换颜色，绝不会在读代理、读测试网或读快照时显示成「已连主网」。
/ The badge states which layer answered: direct mainnet, our read-only relay (verbatim pass-through), testnet 968, or the offline snapshot. A downgrade changes colour and never reads as "connected to mainnet" when it is not.
This page asks a public node directly and prints the raw on-chain answer, payload untouched. No scan.botchain.ai, no backend indexer.
✅ 上面那笔「判定交易」是保单 #2：链上 inputHash / outputHash 与仓库留痕 09-AI判定留痕/snapshot-policy2.json、judgement-policy2.json 重算出来的哈希逐字节相同，modelVersion 是 ai-judge-v1 —— 这一笔可以当场互证：拿仓库里的 JSON 重算，得到的就该是链上这两个哈希。
⚠️ 同族六笔里只有 #2 / #4 / #5 是这样：#0、#1、#3 三笔的 modelVersion 是 bot-rehearsal-v1（排练脚本提交，两个哈希是 keccak256("input/N") 占位值），本来就没有 AI 留痕，不能与 AI 留痕互证。仓库里 #0 / #1 另有两份 ai-judge-v1 留痕，但它们属于第一阶段 Sepolia 链（chainId 11155111），与 968 无关。见 09-AI判定留痕/README.md。
The judgement tx above is policy #2: its on-chain inputHash / outputHash match, byte for byte, the hashes recomputed from the repo traces 09-AI判定留痕/{snapshot,judgement}-policy2.json (modelVersion = ai-judge-v1) — it can be cross-checked on the spot. ⚠️ Only #2/#4/#5 of the six are like this; #0, #1 and #3 were submitted by the rehearsal script (modelVersion = bot-rehearsal-v1, placeholder hashes) and never had an AI trace. The #0/#1 traces in the repo belong to the phase-1 Sepolia chain (chainId 11155111), not 968.