Prove which model made a decision - years after it's gone

A birth certificate for every model setup

• Seal the model, version, provider, the fingerprint of the system prompt and policy in force, temperature and limits, who put it live and when.
• Every decision the engine seals carries that lineage id, so each one points back to the exact setup.
• Retire the model and the certificate gets a provable end date: it was live from this day to that day.
• Years later, one decision receipt walks all the way back - decision → setup → sealed moment → Bitcoin anchor → the prompt and policy fingerprint. The model being gone changes nothing.

import requests, hashlib
H={"Authorization":"Bearer YOUR_KEY"}
# once, when you put a model live
c=requests.post("https://sebbi.pro/x/modellineage/commit",headers=H,json={
  "model":"gpt-4o","version":"2024-08-06","provider":"OpenAI",
  "system_prompt":open("prompt.txt").read(),"policy":open("policy.md").read(),
  "temperature":0.2,"put_live_by":"risk-team"}).json()
lineage=c["lineage_id"]

# on every decision: seal it, then bind it to the lineage
d=seal_my_decision(...)   # your existing sebbi decision
requests.post("https://sebbi.pro/x/modellineage/stamp",headers=H,json={"code":lineage,"decision_hash":d["audit_hash"]})

# when you retire the model
requests.post("https://sebbi.pro/x/modellineage/retire",headers=H,json={"code":lineage})