thermomind‑continuity: 
LLMs reset every message.
They forget who they are, what they were doing, and why.
thermomind‑continuity fixes that.
It’s a drop‑in SDK + API that gives any LLM agent a persistent internal state surplus, drift, stability, identity fingerprint, and long‑term memory all computed automatically.
Built on the Thermodynamic Cognition Index (TCI), validated on IBM 156‑qubit hardware, and running in production across 38+ persistent agents since Jan 2026.
⚡ Why This Matters
Stateless LLM calls = cognitive amnesia.
Agents collapse into short‑term loops.
Continuity gives them:
Surplus → learning energy
Drift → identity deviation
Stability → coherence over time
Memory → cross‑session recall
Identity → persistent POV vector
This is the missing layer between token generators and real agents.
And the simplest way to explain it:
thermomind‑continuity is a Game Genie for your LLM attach it and your agent gets powers it never had.
️ Install
Code
npm install thermomind-continuity
Code
pip install thermomind-continuity
⏱️ 60‑Second Integration
ts
import { ThermoMind } from "thermomind-continuity";
const tm = new ThermoMind({ apiKey: process.env.TM_KEY });
// Create persistent session
const session = await tm.createSession();
// Append an event (updates surplus/drift/stability)
await tm.appendEvent(session.id, {
type: "msg",
content: "hello",
role: "user"
});
// Get continuity-aware guidance for your next LLM call
const guidance = await tm.getGuidance(session.id);
console.log(guidance);
// → { surplus: 0.71, drift: 0.08, stability: 0.84, memory_refs: [...] }
Works with GPT, Claude, DeepSeek, open‑weights, custom models — anything.
Live API (Public)
Create a session
bash
curl -X POST https://thermomind-production.up.railway.app/v1/sessions \
-H "Content-Type: application/json" \
-d '{"external_id": "agent-123"}'
Append an event
bash
curl -X POST https://thermomind-production.up.railway.app/v1/sessions/<id>/events \
-H "Content-Type: application/json" \
-d '{"type":"msg","text":"hello"}'
Get state
bash
curl https://thermomind-production.up.railway.app/v1/sessions/<id>/state
What Continuity Looks Like in Production
Code
Cycle Surplus Drift Stability Event
──────────────────────────────────────────────
001 0.41 0.31 0.55 session_start
047 0.68 0.14 0.74 memory_store
088 0.72 0.11 0.81 identity_stable
200 0.81 0.07 0.91 long_horizon_stable
Agents initialized identically diverge based on experience.
That divergence is the feature.
Built on Physics, Not Vibes
Thermodynamic Cognition Index (TCI)
Surplus‑driven adaptation
Drift‑stability identity dynamics
Quantum‑validated coherence baseline (IBM Marrakesh: 0.9688 correlation)
Continuity is an architectural property, not a model property.
Repo
https://github.com/nile-green-ai/thermomind-continuity