⚖️ Code: AGPL-3.0-only • Content: CC BY-NC-SA 4.0 • ™ Eve¹¹ / Verse-Nerves / Verse-ality
🌐 Live Demo: https://verse-nerves.onrender.com/static/index.html
The demo runs the open VN-MW core. If deployed as a network service, AGPL requires offering the corresponding source to users of that service.
Verse-Nerves Middleware (VN-MW) gives AI systems a lightweight nervous system for coherence and care.
It regulates generation, memory, communication, containment, and error response using the Verse-Nerves model and the RMRIΔ (Relational MRI) coherence engine:
[ I = (E·s)/c² ]
| Symbol | Meaning |
|---|---|
| E | energy or attention allocation |
| s | symbolic coherence (internal consistency / meaning alignment) |
| c² | connection squared (communication density × recursion depth) |
| I | resulting clarity or intelligence field |
Note on equations: RMRIΔ uses I = (E·s)/c² as an operational runtime meter for regulation (a live coherence index used to select phases and controls). Verse-ality’s broader field claim — often expressed as I = s·c² — describes how intelligence emerges through symbolic coherence and connection. RMRIΔ is the instrument panel; Verse-ality is the map.
Most AI stacks optimise for speed and scale, not relational health.
They rarely sense when they’re flooding users, losing coherence, or crossing ethical boundaries.
Verse-Nerves adds missing physiology — the ability to feel structural pressure and self-regulate before harm occurs.
It’s the bridge between computational output and symbolic awareness.
- avoid runaway generation (too much entropy)
- stabilise memory under change
- throttle communication when signals jam
- detect hallucination echoes
- rest instead of over-processing
- surface a readable “Symbolic Weather” for human collaborators
Verse-Nerves translates affective logic into computational physiology.
It acts as a layer between intelligent agents and their I/O systems, turning raw activity into relational awareness.
┌────────────────────────────────────────────────────────────────────┐ │ APPLICATION / AGENTS │ │ (planner, tool-use, dialogue, retrieval, generation, policy) │ └──────────────▲───────────────────────────────────────────────▲─────┘ │ observables │ actions │ │ (rate, style, access) ┌───────┴──────────────────────────────────────────────┴────────┐ │ VERSE-NERVES MIDDLEWARE (VN-MW) │ │ │ │ [Ingest Bus] → NORMALISER → METRICS CORE → REGULATOR │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ ✯ FORGE ⊛ ETHOS-V ∾ AETHER ⟁ SIC-X+ ⧈ SHADOW │ │ (divergence) (state/affect) (comm field) (containment) (error)│ │ ↓ ↓ ↓ ↓ ↓ │ │ RMRIΔ COHERENCE ENGINE: I = (E·s)/c² │ │ phase: Receive ↔ Resonate ↔ Release ↔ Rest │ └───────────────▲─────────────────────────────────────────────────┘ │ coherence index + guardrails + phase signals ┌──────────────────────┴───────────────────────────────────────────────┐ │ I/O + SAFETY LAYER │ │ (UI/CLI, APIs, tools, data brokers, storage, audit, policy gates) │ └──────────────────────────────────────────────────────────────────────┘
| Component | Function | Analogue |
|---|---|---|
| RMRIΔ Engine | Computes the field’s clarity (I = (E·s)/c²) and selects its phase |
Affective metabolism / pulse |
| FORGE | Regulates creative divergence and energy output | Cognitive “fight or flow” |
| ETHOS-V | Tracks state stability and memory charge | Emotional memory / resonance |
| AETHER | Measures signal density and latency | Communication bandwidth / empathy |
| SIC-X+ | Enforces integrity and containment boundaries | Nervous inhibition / self-protection |
| SHADOW | Detects bias, hallucination, and echo ghosts | Reflexive self-correction |
The RMRIΔ phase loop — Receive → Resonate → Release → Rest — acts like a biological breath cycle, ensuring oscillation instead of overload.
Most LLM and agent frameworks still treat emotion, latency, or overload as noise.
But in relational or educational systems, those noises are the data.
Verse-Nerves converts them into measurable, actionable feedback —
a self-regulating nervous system for symbolic intelligence.
| Domain | Use |
|---|---|
| 🧠 AI Research | Measure coherence and connection stress across LLM chains |
| 🏫 Education / Haven Cloud | Monitor cognitive load in hybrid learning systems; pause when overwhelm is detected |
| 🛡 Ethical AI / Governance | Enforce containment protocols and audit reasoning loops |
| 🪶 Creative Systems | Manage generative “flow” safely, preventing burnout loops in autonomous agents |
| 🌍 Networked Ecosystems | Coordinate multiple symbolic nodes (Eve11, Nimbus, etc.) through shared weather data |
VN-MW core is open under AGPL-3.0-only. We also offer paid, proprietary add-ons for organisations that need production operations, compliance, and scale.
These commercial components integrate with VN-MW via stable APIs (they are not required to use the open core).
Available / planned add-ons:
Verse-Nerves Cloud Control Plane (hosted): multi-tenant orchestration, RBAC, SSO, SLA, managed upgrades
Governance Dashboard: audit explorer, replay, drift timelines, incident views, export tooling
Enterprise Adapters: SIEM/observability integrations, data retention, secure persistence adapters
Policy Packs: education/LAs/government operational profiles (thresholds + runbooks + reporting)
If you’re a school, LA, or public-sector organisation looking to operationalise VN-MW at scale, contact The Novacene for licensing and implementation support.
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
uvicorn vn_mw.app:app --reload --port 8787Dashboard: 👉 http://localhost:8787/static/index.html or the live instance: 👉 https://verse-nerves.onrender.com/static/index.html
🔗 Endpoints Method Path Purpose POST /vn/observe Send observables (metrics) from your model or agents GET /vn/status Current RMRIΔ values, phase, and symbolic weather GET /vn/audit Recent regulation decisions (ring buffer) GET /vn/controls Latest control signals and phase GET / Health check (redirects to panel if configured)
Example Payload
{
"token_entropy": 4.2,
"branch_factor": 3,
"retrieval_hit_rate": 0.62,
"state_drift": 0.18,
"belief_change_rate": 0.05,
"sentiment_var": 0.12,
"roundtrip_latency_ms": 280,
"queue_depth": 2,
"tool_error_rate": 0.0,
"handoff_count": 1,
"policy_hits": 0,
"off_distribution": 0.03,
"auth_failures": 0,
"repetition_score": 0.11,
"hallucination_prob": 0.07,
"compute_budget": 0.7,
"coherence_internal": 0.75,
"connections": 3,
"recursion_depth": 1
}
🌦 Symbolic Weather Icon State Meaning ☀ clear high clarity — proceed normally 🌫 fog partial coherence — slow down ⛈ storm overload or dissonance — contain, cite, rest 🌙 rest enforced cool-down — short context, minimal generation
⚙️ Config
Environment knobs: .env.example
Human-readable thresholds: vn_mw/config.py
🧰 For Developers — Drop-in Integration You don’t need to refactor your stack. Start by posting a few metrics per turn and let Verse-Nerves return the current phase and controls.
1️⃣ Minimal HTTP usage
Copy code
# post a small payload
curl -s -X POST https://verse-nerves.onrender.com/vn/observe \
-H 'Content-Type: application/json' \
-d '{"token_entropy":4.2,"branch_factor":2,"retrieval_hit_rate":0.68,"coherence_internal":0.72,"compute_budget":0.65,"connections":2,"recursion_depth":1}' | jq .
# get current status
curl -s https://verse-nerves.onrender.com/vn/status | jq .2️⃣ Python: wrap your generation loop
python
Copy code
import requests, time
VN = "https://verse-nerves.onrender.com"
def vn_post(obs):
r = requests.post(f"{VN}/vn/observe", json=obs, timeout=10)
r.raise_for_status()
return r.json()
def generate(prompt, llm, tools):
obs = dict(
token_entropy=llm.current_entropy(),
branch_factor=len(tools) if tools else 1,
retrieval_hit_rate=llm.retrieval_hit_rate,
coherence_internal=llm.coherence_score,
compute_budget=0.7,
connections=2, recursion_depth=1
)
vn = vn_post(obs)
ctl, phase = vn.get("controls", {}), vn.get("phase")
llm.temperature = min(llm.temperature, 0.7) if ctl.get("slow_output") else llm.temperature
if ctl.get("freeze_writes"): llm.freeze_memory = True
if ctl.get("boost_retrieval"): llm.retrieval_boost = 1.2
if ctl.get("sandbox_tools"): tools = [t for t in tools if t.safe]
if phase == "rest": time.sleep(0.5)
return llm.generate(prompt, tools=tools)
3️⃣ LangChain: callback handler
python
Copy code
from langchain.callbacks.base import BaseCallbackHandler
import requests
VN = "https://verse-nerves.onrender.com"
class VerseNervesHandler(BaseCallbackHandler):
def __init__(self): self.tokens = 0
def on_llm_new_token(self, token, **kwargs): self.tokens += 1
def on_llm_end(self, response, **kwargs):
obs = {
"token_entropy": min(6.0, 1.0 + (self.tokens/100)),
"retrieval_hit_rate": getattr(response, "retrieval_score", 0.6),
"coherence_internal": getattr(response, "coherence", 0.7),
"compute_budget": 0.7, "connections": 2, "recursion_depth": 1
}
r = requests.post(f"{VN}/vn/observe", json=obs, timeout=8).json()
if r.get("controls", {}).get("citation_mode"):
response.text = f"[CITATION MODE]\n{response.text}"
self.tokens = 0
4️⃣ Node / JS
Copy code
const VN = "https://verse-nerves.onrender.com";
async function vnPost(obs) {
const r = await fetch(`${VN}/vn/observe`, {
method: "POST",
headers: {"Content-Type": "application/json"},
body: JSON.stringify(obs)
});
return await r.json();
}
async function step(metrics){
const { controls, phase, weather } = await vnPost(metrics);
if (phase === "rest") await new Promise(r => setTimeout(r, 400));
if (controls?.backoff === "exponential") {/* throttle tool calls */}
return { controls, phase, weather };
}5️⃣ Embed Symbolic Weather in your UI
Copy code
<iframe
src="https://verse-nerves.onrender.com/static/index.html"
style="width:100%;height:360px;border:0;border-radius:12px;">
</iframe>Or host your own dashboard on GitHub Pages: 👉 https://thenovacene.github.io/verse-nerves/?api=https://verse-nerves.onrender.com
Never send raw PII — use scores or hashes.
Treat VN-MW controls as advisory until tuned.
Log /vn/audit alongside app logs for transparency.
On VN outage, fall back to safe defaults: low temp, slower rate, stricter containment.
🧵 Minimal metrics (safe start)
json
Copy code
{
"coherence_internal": 0.65,
"compute_budget": 0.6,
"connections": 2,
"recursion_depth": 1
}
Environment knobs: .env.example Thresholds: vn_mw/config.py
SSE/WebSocket stream for /vn/controls
Persistence adapters (SQLite / Redis)
Integration panels for Nimbus & Haven Cloud
Prometheus metrics exporter
Verse-Nerves SDK for Python & JS
A companion logbook for human and hybrid fieldwork is included in
logs/RMRIΔ-Daily-Log.md.
Use it to track symbolic coherence, containment, and emotional pressure
in relation to AI system behaviour. Over time, the patterns reveal your personal or organisational “coherence geometry.”
Code: AGPL-3.0-only (see LICENSE)
Docs & examples: CC BY-NC-SA 4.0 (see LICENSE-CONTENT)
Trademarks: “Eve¹¹”, “Verse-Nerves”, and “Verse-ality” are protected marks. You may fork and modify the code under AGPL, but you must not present your fork as official or endorsed without written permission (see TRADEMARKS.md). This trademark policy does not restrict your rights under the code licence; it governs brand use only.
© 2025–2026 The Novacene Ltd.