v4.0.1 Production API Specifications
OpenAI-Compatible & REST Mesh Engine
Developer API Documentation Portal
Comprehensive REST & OpenAI-compatible API specifications for client compute job execution, real-time workspace sync, autonomous agentic swarms, multimodal inference, partner white-label management, and crypto billing.
Global API Gateway Metadata Gateway Live
Base Gateway URL
https://web-lon-01.ncx.one/api/v1
Authentication
Bearer SK_API_KEY
Payload Format
JSON & SSE Streaming
Agentic Sandbox
AppArmor / WASM / eBPF
Platform Architecture & Overview
v4.0.1 CoreThe MineFarm Compute & Inference Gateway exposes high-speed RESTful endpoints for deploying AI models, running agentic sub-tasks across P2P WireGuard worker nodes, managing reverse SSH tunnels, and leasing bare-metal GPU clusters.
Standard OpenAI SDKs
Plug-and-play drop-in replacement for OpenAI, LangChain, AutoGen, and LlamaIndex.
AppArmor / WASM Sandbox
Zero-trust execution sandboxes with memory limits and eBPF syscall filtration.
100-Port Passthrough Block
Dedicated 100 consecutive ports per GPU rental for SSH, Jupyter, WebUI, and TensorBoard.
Authentication & Security Policy
NIST SP 800-88 Rev 2
Strict Zero-Trust
All platform REST endpoints across /api/v1/ strictly require authenticated credentials. Authentication can be provided via an HTTP Bearer Token or an X-API-KEY header:
Rate Limiting & Anti-Brute-Force
Authentication routes enforce a strict limit of 5 failed attempts per 300s window per client IP. Breaches immediately return HTTP 429 Too Many Requests with Retry-After: 300 headers.
Transport Security (SSL/TLS)
Production servers enforce TLS 1.3 / 1.2 with Strict-Transport-Security (HSTS: 1 year, subdomains, preload), Content-Security-Policy: frame-ancestors, and X-Content-Type-Options: nosniff. Local dev units gracefully operate with internal SSL bypass for offline testing.
NIST SP 800-88 Rev 2 Purge
Confidential Clean Rooms and memory pipelines implement NIST SP 800-88 Rev 2 multi-pass cryptographic zeroization, issuing HMAC-SHA512 tamper-evident compliance audit certificates.
App & Mobile Authentication API
Bearer Token Exchange
Enables mobile apps, desktop applications, CLI clients, and external integrations to securely authenticate to the system, exchange user credentials for a 30-day App Bearer Token (ncx_app_*), verify 2FA challenges, retrieve real-time account context, and revoke sessions.
1. App Login & Token Issuance
Request Payload (JSON)
{
"username": "user@example.com",
"password": "your_secure_password",
"app_name": "NCX Mobile iOS",
"device_info": "iPhone 15 Pro / iOS 18.2",
"otp_code": "123456" // Optional if 2FA enabled
}
Success Response (200 OK)
{
"success": true,
"token_type": "Bearer",
"access_token": "ncx_app_7f4d2a8b9e1c3f5a0d6e8c7b4a2f1e9d3c5b7a...",
"api_key": "sk-ncx-a9b8c7d6e5f41234567890abcdef...",
"expires_in": 2592000,
"expires_at": "2026-09-29 18:30:00",
"user": {
"id": 1,
"username": "developer",
"email": "user@example.com",
"role": "renter",
"balance_usd": 150.00,
"twofa_enabled": true,
"theme": "default",
"language": "en"
},
"tenant": {
"name": "MineFarm",
"primary_color": "#ff5e14",
"base_url": "https://web-lon-01.ncx.one"
}
}
2FA Required Response (If OTP omitted)
{
"success": false,
"2fa_required": true,
"temp_token": "ncx_2fa_98a7b6c5d4e3f210...",
"message": "Two-factor authentication code required. Pass otp_code or verify via /api/v1/auth/verify_2fa.",
"expires_in": 300
}
2. Authenticated Profile & State
cURL Example
curl -X GET "https://web-lon-01.ncx.one/api/v1/auth/me" \
-H "Authorization: Bearer ncx_app_7f4d2a8b9e1c3f5a0d6e8c7b4a2f1e9d3c5b7a..."
3. Revoke Session / Logout
1. Renter CLI Hot-Sync API
Delta Workspace Sync
Real-time bi-directional workspace sync and delta file transfer for active compute jobs. Used by the NCX CLI (ncx-cli sync) to hot-patch training scripts, configurations, and checkpoint weights directly into the isolated container sandbox without SSH session restarts.
Request Body Options ($_POST Parameters):
| Field / Parameter | Type | In | Status | Description |
|---|---|---|---|---|
| api_key | string | $_POST |
Required | Renter secret API Key (ncx_api_... or account API key). |
| job_id | integer | $_POST |
Required | Unique ID of the active target compute job inside container sandbox. |
| file_path | string | $_POST |
Required | Destination path relative to container workspace root (e.g. src/train.py). |
| file_content | string / binary | $_POST |
Required | Raw file payload contents or binary data stream to sync into the container. |
| is_delta | integer | $_POST |
Optional | Set to 1 for binary delta block patch (Hot-Sync protocol), or 0 for full file replace. Default 0. |
| checksum | string | $_POST |
Optional | SHA-256 or MD5 hex digest for end-to-end integrity verification. |
Framework Code:
NCX CLI (ncx-cli - Automated Workspace Hot-Sync)
# 1. Authenticate using your NCX API key:
export NCX_API_KEY="sk-ncx-xxxxxxxxxxxxxxxx"
# 2. Synchronize local workspace directory into remote active container sandbox:
ncx-cli sync --job 1042 --dir ./src
# Alternatively hot-sync a single file with explicit key:
ncx-cli sync --key "ncx_api_xxxxxxxxxxxxxxxx" --job 1042 --file ./src/train.py
# Continuous watch mode (syncs local changes in real time via delta blocks):
ncx-cli sync --job 1042 --dir ./src --watch
PHP ($_POST via cURL)
<?php
// Synchronize file to remote compute job using native PHP cURL and $_POST
$apiKey = 'ncx_api_xxxxxxxxxxxxxxxx';
$jobId = 1042;
$filePath = 'src/train.py';
$fileContent = file_get_contents('src/train.py');
$postFields = [
'api_key' => $apiKey,
'job_id' => $jobId,
'file_path' => $filePath,
'file_content' => $fileContent,
'is_delta' => 0,
'checksum' => hash('sha256', $fileContent)
];
$ch = curl_init('https://web-lon-01.ncx.one/api/jobs/sync_file');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $postFields);
$response = curl_exec($ch);
$httpCode = curl_getinfo($ch, CURLINFO_HTTP_CODE);
curl_close($ch);
$result = json_decode($response, true);
print_r($result);
cURL (Multipart / Form-Data)
curl -X POST "https://web-lon-01.ncx.one/api/jobs/sync_file" \
-F "api_key=ncx_api_xxxxxxxxxxxxxxxx" \
-F "job_id=1042" \
-F "file_path=src/train.py" \
-F "file_content=@src/train.py" \
-F "is_delta=0"
Python (requests form data)
import requests
import hashlib
url = "https://web-lon-01.ncx.one/api/jobs/sync_file"
with open("src/train.py", "r", encoding="utf-8") as f:
code = f.read()
payload = {
"api_key": "ncx_api_xxxxxxxxxxxxxxxx",
"job_id": 1042,
"file_path": "src/train.py",
"file_content": code,
"is_delta": 0,
"checksum": hashlib.sha256(code.encode("utf-8")).hexdigest()
}
response = requests.post(url, data=payload)
print(response.status_code, response.json())
JavaScript (Fetch / FormData)
const formData = new FormData();
formData.append('api_key', 'ncx_api_xxxxxxxxxxxxxxxx');
formData.append('job_id', '1042');
formData.append('file_path', 'src/train.py');
formData.append('file_content', 'import torch\nprint("Training active...")');
formData.append('is_delta', '0');
const res = await fetch('https://web-lon-01.ncx.one/api/jobs/sync_file', {
method: 'POST',
body: formData
});
const result = await res.json();
console.log(result);
Success Response Format (HTTP 200 OK)
{
"success": true,
"message": "File successfully synced to compute container sandbox!",
"file_path": "src/train.py",
"size": 1024,
"checksum": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
}
Error Responses
// 400 Bad Request:
{ "success": false, "message": "Missing required parameters." }
// 401 Unauthorized:
{ "success": false, "message": "Invalid or inactive Renter API key." }
// 404 Not Found:
{ "success": false, "message": "Active running job not found for this renter." }
2. OpenAI-Compatible Inference API
OpenAI Standard
Universal drop-in replacement for OpenAI chat completions (POST /v1/chat/completions). Supports live SSE token streaming, multi-node speculative decoding, MoA swarm consensus, and neuromorphic Spiking Neural Network (SNN) routing.
Request Body Options (JSON Payload):
| Parameter | Type | Status | Description |
|---|---|---|---|
| model | string | Required | Model slug from catalog (e.g. llama3.3:70b, deepseek-r1:32b, braincog/spiking-resnet50). |
| messages | array | Required | Conversation turns: [{"role": "user", "content": "..."}]. |
| routing_mode | string | Optional | Swarm execution mode: auto, vllm_speculative, moa_consensus, sharded_pipeline, cleanroom_enclave_infer, cleanroom_fhe_eval, cleanroom_dp_synthesize, pattern_snn_ann, pattern_ann_snn, pattern_spiking_transformer, snn_spiking, hybrid_cnn, hybrid_rnn, hybrid_vit, hybrid_lsm, hybrid_rl, hybrid_gnn, hybrid_kan, hybrid_ssm, hybrid_diffusion, hybrid_neural_ode, hybrid_pinn, hybrid_st_gnn, hybrid_hopfield, hybrid_hypernet, hybrid_cfc, hybrid_se3_equivariant, hybrid_linear_rwkv, hybrid_neuro_symbolic, or rig_{id}. |
| cleanroom_uuid | string | Optional | Target confidential clean room UUID (e.g. cr_h100_oncology_consortium) when executing within confidential TEE / FHE clean rooms. |
| temperature | float | Optional | Sampling temperature (0.0 to 2.0). Default 0.7. |
| stream | boolean | Optional | If true, emits real-time Server-Sent Events (SSE) token chunks. Default false. |
| timesteps | integer | Optional | Neuromorphic discrete simulation steps for SNN mode (default 8). |
Framework Code:
cURL Request Example
curl -X POST "https://web-lon-01.ncx.one/api/v1/chat/completions" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx" \
-d '{
"model": "llama3.3:70b",
"messages": [
{ "role": "system", "content": "You are a helpful AI assistant." },
{ "role": "user", "content": "Explain GPU compute in 2 sentences." }
],
"temperature": 0.7,
"stream": false
}'
PHP (JSON Post via cURL)
<?php
$apiKey = 'sk-ncx-xxxxxxxxxxxxxxxx';
$payload = json_encode([
'model' => 'llama3.3:70b',
'messages' => [
['role' => 'user', 'content' => 'Explain GPU compute in 2 sentences.']
],
'temperature' => 0.7,
'stream' => false
]);
$ch = curl_init('https://web-lon-01.ncx.one/api/v1/chat/completions');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $payload);
curl_setopt($ch, CURLOPT_HTTPHEADER, [
'Content-Type: application/json',
"Authorization: Bearer {$apiKey}"
]);
$response = curl_exec($ch);
curl_close($ch);
$data = json_decode($response, true);
echo $data['choices'][0]['message']['content'];
Python (Using Official OpenAI SDK)
from openai import OpenAI
client = OpenAI(
base_url="https://web-lon-01.ncx.one/api/v1",
api_key="sk-ncx-xxxxxxxxxxxxxxxx"
)
response = client.chat.completions.create(
model="llama3.3:70b",
messages=[{"role": "user", "content": "Explain GPU compute in 2 sentences."}],
temperature=0.7
)
print(response.choices[0].message.content)
JavaScript (Fetch API)
const res = await fetch('https://web-lon-01.ncx.one/api/v1/chat/completions', {
method: 'POST',
headers: {
'Content-Type': 'application/json',
'Authorization': 'Bearer sk-ncx-xxxxxxxxxxxxxxxx'
},
body: JSON.stringify({
model: 'llama3.3:70b',
messages: [{ role: 'user', content: 'Explain GPU compute in 2 sentences.' }],
temperature: 0.7
})
});
const data = await res.json();
console.log(data.choices[0].message.content);
Success Response (HTTP 200 OK)
{
"id": "chatcmpl-8a1f7c3e",
"object": "chat.completion",
"created": 1725548400,
"model": "llama3.3:70b",
"routing_mode": "auto",
"choices": [
{
"index": 0,
"message": {
"role": "assistant",
"content": "GPU compute utilizes thousands of parallel processing cores to execute matrix multiplication and tensor calculations simultaneously. Hardware-accelerated pipelines route inference layers dynamically to minimize latency and maximize throughput."
},
"finish_reason": "stop"
}
],
"usage": {
"prompt_tokens": 18,
"completion_tokens": 34,
"total_tokens": 52
}
}
3. Distributed vLLM Speculative Swarm API
Speculative Engine
Co-locates low-latency ARM Edge generator nodes with high-VRAM GPU verifiers. Small draft models (e.g. qwen2.5:0.5b) produce candidate token blocks in parallel, which the master 70B verifier validates in a single forward pass, achieving 2.4x - 3.8x throughput acceleration.
Request Body Parameters:
| Parameter | Type | Status | Description |
|---|---|---|---|
| model | string | Required | Master verifier model (e.g. meta-llama/Meta-Llama-3-70B-Instruct). |
| draft_model | string | Optional | Draft generator model (e.g. qwen2.5:0.5b, llama-3-8b-draft). |
| messages | array | Required | Prompt context turns to execute speculative decoding against. |
| speculative_steps | integer | Optional | Lookahead draft tokens evaluated per verification cycle (default 5). |
Framework Code:
cURL Request Example
curl -X POST "https://web-lon-01.ncx.one/api/v1/swarm/vllm_orchestrate" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx" \
-d '{
"model": "llama3.3:70b",
"draft_model": "qwen2.5:0.5b",
"messages": [{"role": "user", "content": "Synthesize a parallel matrix reduction algorithm."}],
"speculative_steps": 5
}'
PHP Example
<?php
$payload = json_encode([
'model' => 'llama3.3:70b',
'draft_model' => 'qwen2.5:0.5b',
'messages' => [['role' => 'user', 'content' => 'Synthesize a parallel matrix reduction algorithm.']],
'speculative_steps' => 5
]);
$ch = curl_init('https://web-lon-01.ncx.one/api/v1/swarm/vllm_orchestrate');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $payload);
curl_setopt($ch, CURLOPT_HTTPHEADER, [
'Content-Type: application/json',
'Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx'
]);
$res = curl_exec($ch);
curl_close($ch);
print_r(json_decode($res, true));
Python Example
import requests
res = requests.post(
"https://web-lon-01.ncx.one/api/v1/swarm/vllm_orchestrate",
headers={"Authorization": "Bearer sk-ncx-xxxxxxxxxxxxxxxx"},
json={
"model": "llama3.3:70b",
"draft_model": "qwen2.5:0.5b",
"messages": [{"role": "user", "content": "Synthesize a parallel matrix reduction algorithm."}],
"speculative_steps": 5
}
)
print(res.json())
Success Response Format (HTTP 200 OK)
{
"status": "success",
"routing_mode": "vllm_speculative",
"choices": [
{ "message": { "role": "assistant", "content": "..." }, "finish_reason": "stop" }
],
"speculative_telemetry": {
"draft_tokens_generated": 128,
"tokens_accepted": 108,
"acceptance_rate_pct": 84.38,
"speedup_ratio": "2.84x",
"generator_node": "ARM-Cluster-02",
"verifier_node": "DGX-Spark-01"
}
}
4. Mixture-of-Agents (MoA) Swarm Consensus API
3-Node Fan-OutMulti-Agent consensus pipeline that fans out input prompts simultaneously across 3 distinct peer swarm nodes, collecting independent perspectives and synthesizing a unified response with cross-node consensus scoring.
Request Body Parameters:
| Parameter | Type | Status | Description |
|---|---|---|---|
| messages | array | Required | Standard user / assistant turns to fan out across peer swarm workers. |
| fanout_nodes | integer | Optional | Number of parallel candidate worker nodes to poll (default 3). |
| synthesis_model | string | Optional | Aggregator model combining candidate outputs (default llama3.3:70b). |
Framework Code:
cURL Request Example
curl -X POST "https://web-lon-01.ncx.one/api/v1/swarm/moa_consensus" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx" \
-d '{
"messages": [{"role": "user", "content": "What are optimal cooling strategies for high-density 8x CMP 170HX GPU racks?"}],
"fanout_nodes": 3
}'
PHP Example
<?php
$payload = json_encode([
'messages' => [['role' => 'user', 'content' => 'What are optimal cooling strategies for high-density 8x CMP 170HX GPU racks?']],
'fanout_nodes' => 3
]);
$ch = curl_init('https://web-lon-01.ncx.one/api/v1/swarm/moa_consensus');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $payload);
curl_setopt($ch, CURLOPT_HTTPHEADER, [
'Content-Type: application/json',
'Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx'
]);
$res = curl_exec($ch);
curl_close($ch);
print_r(json_decode($res, true));
Python Example
import requests
res = requests.post(
"https://web-lon-01.ncx.one/api/v1/swarm/moa_consensus",
headers={"Authorization": "Bearer sk-ncx-xxxxxxxxxxxxxxxx"},
json={
"messages": [{"role": "user", "content": "What are optimal cooling strategies for high-density 8x CMP 170HX GPU racks?"}],
"fanout_nodes": 3
}
)
print(res.json())
Success Response Format (HTTP 200 OK)
{
"status": "success",
"routing_mode": "moa_consensus",
"consensus_output": "High-density CMP 170HX mining cards benefit most from rear-door liquid heat exchangers (RDHx) paired with 2.8 CFM push-pull Delta blower fans...",
"moa_telemetry": {
"fanout_count": 3,
"synthesizer_node": "Swarm-Master-Node",
"participating_nodes": ["Rig01-170hx", "Rig02-Spark", "Rig04-4090"],
"consensus_confidence": 0.962
}
}
5. Pipeline Layer Sharding ("Skippy JIT") API
80-Layer ShardSplits 80-layer neural network architectures across distinct GPU worker nodes over WebSocket or WireGuard transport. Each node computes assigned layer slices (e.g. Layers 0-26, 27-53, 54-79) and hot-streams intermediate activation tensors to downstream workers.
Request Body Parameters:
| Parameter | Type | Status | Description |
|---|---|---|---|
| model | string | Required | Sharded model target (e.g. deepseek-r1:70b-sharded). |
| messages | array | Required | Input conversation context to process across pipeline stages. |
| transport | string | Optional | Inter-layer tensor transport: websocket or wireguard_direct. |
Success Response Format (HTTP 200 OK)
{
"status": "success",
"routing_mode": "sharded_pipeline",
"pipeline_distribution": [
{ "stage": 1, "node": "Rig01", "layers": "0-26", "transport_latency_ms": 1.8 },
{ "stage": 2, "node": "Rig02", "layers": "27-53", "transport_latency_ms": 2.1 },
{ "stage": 3, "node": "Rig04", "layers": "54-79", "transport_latency_ms": 1.9 }
],
"choices": [
{ "message": { "role": "assistant", "content": "..." } }
]
}
6. Neuromorphic, Hybrid & Scientific Neural Network Suite
15 Architectures
58x Green Compute
Enterprise-grade neuromorphic, hybrid, continuous, and scientific neural network acceleration based on BrainCog and NCX distributed compute. Unifies event-driven discrete membrane potential dynamics (LIFNode, PLIFNode) with dense deep learning backends, continuous-depth ODEs, physics-informed PDEs, and associative memory to achieve 85% - 95% activation sparsity, zero backpropagation through time (BPTT), and verifiable ISO 14064 Scope-3 carbon accounting.
A. Foundational Common Hybrid Design Patterns
Pattern 1: SNN Front-End + ANN Back-End
routing_mode: pattern_snn_annPerception to Reasoning: An SNN sits directly on neuromorphic hardware or reads from an event camera (DVS). It performs edge noise filtering, edge detection, and rapid anomaly spotting at near-zero idle power.
Handoff: When an anomaly or threshold is crossed, accumulated firing vectors wake up a dense ANN (CNN/Transformer) for high-level classification.
Pattern 2: ANN Front-End + SNN Back-End
routing_mode: pattern_ann_snnFeature Mapping to Control: A traditional dense CNN or ResNet extracts rich visual embeddings from high-resolution dense frames.
Handoff: Embeddings are converted to spike trains (Rate Coding / TTFS) and passed to a downstream SNN controller for sub-millisecond, low-power robotic motor actuation with STDP plasticity.
Pattern 3: Spiking Transformers
routing_mode: pattern_spiking_transformerSpike-Driven Self-Attention: Replaces standard floating-point Matrix Multiplications (Q, K, V dot-products) with spike-driven sparse binary additions (0 or 1).
Advantage: Outer tokenizers remain floating-point, while core attention operates via additions, cutting memory bandwidth by 78%+ and FLOP energy by 44x.
B. Neuromorphic, Hybrid & Scientific Architecture Matrix
| Architecture | Model Slug & Routing Key | Primary Role & Synergy | Energy vs FP32 | Target Use Cases |
|---|---|---|---|---|
| Spiking ResNet (SNN) | braincog/spiking-resnet50mode: snn_spiking |
Spatiotemporal LIF/PLIF event detection with surrogate gradients. | 35x - 58x | Event Cameras, ADAS, Edge DVS Vision |
| DVS ConvNet (SNN) | braincog/dvs-convnetmode: snn_spiking |
Asynchronous DVS polarity event stream filtering (x, y, t, p) with continuous LIF dynamics. | 40x - 65x | High-Speed DVS Vision, Collision Avoidance, Neuromorphic ADAS |
| Hybrid CNN-SNN | braincog/hybrid-spiking-cnnmode: hybrid_cnn |
Translates spatial structures while SNN layers process temporal motion. | 25x - 42x | Drone Obstacle Avoidance, High-Speed Video |
| Hybrid RNN/LSTM-SNN | braincog/hybrid-spiking-rnnmode: hybrid_rnn |
Compensates for decaying temporal horizon of LIF neurons with context tracking. | 20x - 38x | Financial Tick Series, Audio Speech Decoding |
| Hybrid Vision Transformer | braincog/hybrid-spiking-vitmode: hybrid_vit |
SNN tokens compress high-rate sensor streams for sparse attention heads. | 30x - 45x | Multi-Modal Sensor Fusion, Aerial Imagery |
| Liquid State Machine (LSM) | braincog/spiking-lsm-reservoirmode: hybrid_lsm |
1024-node reservoir computing requiring zero BPTT end-to-end. | 40x - 60x | Low-Power IoT Audio, Anomaly Detection |
| Spiking Deep RL | braincog/spiking-rl-actor-criticmode: hybrid_rl |
SNN extracts fast sensory cues, ANN policy outputs continuous motor signals. | 28x - 44x | Robotic Arm Kinematics, UAV Autopilot |
| Spiking GNN | braincog/hybrid-spiking-gnnmode: hybrid_gnn |
Topological knowledge graph reasoning over P2P mesh via sparse event passing. | 32x - 48x | P2P Compute Topology, Social Graph Reasoner |
| Spiking KAN (Splines) | braincog/hybrid-spiking-kanmode: hybrid_kan |
Learnable B-spline edge activations replacing fixed linear weights. | 22x - 36x | Interpretable Science, Symbolic Mathematics |
| Spiking SSM (Mamba) | braincog/hybrid-spiking-ssmmode: hybrid_ssm |
Linear-time O(N) continuous sequence modeling paired with sub-ms event filter. | 35x - 50x | Long-Context Telemetry, Genomics, DNA Streams |
| Spiking Diffusion (DiT) | braincog/spiking-diffusion-ditmode: hybrid_diffusion |
Spike-guided generative denoising trajectory cutting step-by-step FLOPs by 50%+. | 18x - 32x | Generative Image Synthesis, Protein Denoising |
| Continuous Neural ODE | braincog/continuous-neural-odemode: hybrid_neural_ode |
Adaptive Runge-Kutta 4th-order solver for irregularly sampled time-series (O(1) RAM). | 24x - 40x | Medical ECG/Vitals, Physical Drift Tracking |
| Physics-Informed NN (PINN) | ncx/physics-informed-pinnmode: hybrid_pinn |
Enforces Navier-Stokes fluid dynamics and thermodynamic PDE conservation laws. | 20x - 35x | CFD Simulations, Turbine Heat Dissipation |
| Spatio-Temporal GNN | braincog/spatio-temporal-gnnmode: hybrid_st_gnn |
Dynamic graph topology across time for P2P mesh routing & latency balancing. | 26x - 42x | Decentralized Cloud Routing, Smart Traffic |
| Modern Hopfield (EBM) | braincog/modern-hopfield-ebmmode: hybrid_hopfield |
Continuous associative memory network with exponential storage capacity C ~ 2^(d/2). | 30x - 52x | Associative Database Recall, Pattern Recovery |
| Dynamic Hypernetwork | ncx/dynamic-hypernetworkmode: hybrid_hypernet |
Meta-generator synthesizing client-tailored edge rig weights conditioned on hardware. | 25x - 45x | Heterogeneous Edge Adaptation, Zero-Sync Rigs |
| Closed-Form Continuous (CfC) | braincog/closed-form-continuous-cfcmode: hybrid_cfc |
Explicit continuous-time state adaptation eliminating numerical ODE solver lag for sub-ms control. | 28x - 46x | Autonomous Robotics, Drivonic ADAS, Dynamic Flight |
| SE(3)-Equivariant GNN | ncx/se3-equivariant-gnnmode: hybrid_se3_equivariant |
Equivariant message passing guaranteeing 3D rotational/translational physical invariance. | 30x - 50x | Molecular Ligand Docking, 3D LiDAR Point Clouds |
| Linear-Attention RWKV | ncx/linear-attention-rwkvmode: hybrid_linear_rwkv |
Recursive WKV time-mixing with constant O(1) inference memory, saving 94%+ KV cache VRAM. | 35x - 55x | 100k+ Token Streaming, Edge GPU Transformers |
| Neuro-Symbolic LTN | ncx/neuro-symbolic-ltnmode: hybrid_neuro_symbolic |
Real Logic fuzzy t-norms enforcing First-Order Predicates & statutory compliance rules. | 32x - 48x | Defense Flight Verification, NHS DSPT Healthcare |
C. Direct SNN & Hybrid Inference API
Parameters Table:
| Parameter | Type | In | Status | Description |
|---|---|---|---|---|
| model | string | JSON / $_POST |
Optional | Model slug from catalog (e.g. braincog/hybrid-spiking-cnn, braincog/continuous-neural-ode). |
| routing_mode | string | JSON / $_POST |
Optional | Direct routing mode (e.g. hybrid_cnn, pattern_snn_ann, hybrid_diffusion). |
| input_data | string / array | JSON / $_POST |
Required | Input prompt, token text, sensor coordinates, or DVS event stream sequence. |
| timesteps | integer | JSON / $_POST |
Optional | Temporal simulation steps (default 8, max 64). |
| node_type | string | JSON / $_POST |
Optional | Neuron membrane model: LIFNode, PLIFNode, or ContinuousLIFNode. |
| return_raster | boolean | JSON / $_POST |
Optional | If true, returns ASCII spatiotemporal spike activity raster. |
Framework Code:
PHP (cURL & $_POST) Example
<?php
$payload = [
'model' => 'braincog/hybrid-spiking-cnn',
'routing_mode' => 'hybrid_cnn',
'input_data' => 'Process high-speed event stream with CNN spatial feature front-end and SNN temporal filter.',
'timesteps' => 8,
'neuron_model' => 'LIFNode',
'return_raster' => true
];
$ch = curl_init('https://web-lon-01.ncx.one/api/v1/snn/infer');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, json_encode($payload));
curl_setopt($ch, CURLOPT_HTTPHEADER, ['Content-Type: application/json']);
$response = curl_exec($ch);
curl_close($ch);
$res = json_decode($response, true);
echo "Architecture: " . $res['architecture'] . "
";
echo "Sparsity: " . $res['spiking_telemetry']['activation_sparsity_pct'] . "% quiescent
";
echo "Energy Reduction: " . $res['esg_energy_telemetry']['energy_savings_ratio'] . "
";
cURL Request Example
curl -X POST "https://web-lon-01.ncx.one/api/v1/snn/infer" \
-H "Content-Type: application/json" \
-d '{
"routing_mode": "hybrid_cnn",
"input_data": "Process high-speed event stream with CNN spatial feature front-end and SNN temporal filter.",
"timesteps": 8,
"return_raster": true
}'
Python (requests) Example
import requests
url = "https://web-lon-01.ncx.one/api/v1/snn/infer"
data = {
"routing_mode": "hybrid_cnn",
"input_data": "Process high-speed event stream with CNN spatial feature front-end and SNN temporal filter.",
"timesteps": 8,
"return_raster": True
}
res = requests.post(url, json=data).json()
print(f"Model: {res['model']} ({res['architecture']})")
print(f"Activation Sparsity: {res['spiking_telemetry']['activation_sparsity_pct']}%")
print(f"Energy Savings: {res['esg_energy_telemetry']['energy_savings_ratio']}")
JavaScript (Fetch) Example
const res = await fetch('https://web-lon-01.ncx.one/api/v1/snn/infer', {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({
routing_mode: 'hybrid_cnn',
input_data: 'Process high-speed event stream with CNN spatial feature front-end and SNN temporal filter.',
timesteps: 8
})
});
const data = await res.json();
console.log(data.hybrid_telemetry);
Success Response Format (HTTP 200 OK)
{
"status": "success",
"model": "BrainCog Hybrid-Spiking-CNN",
"slug": "braincog/hybrid-spiking-cnn",
"architecture": "HybridCNN",
"routing_mode": "hybrid_cnn",
"neuron_model": "LIFNode",
"timesteps": 8,
"classification": {
"top_class": "spatial_temporal_feature_extraction",
"confidence": 0.988,
"domain": "hybrid_vision_motion"
},
"spiking_telemetry": {
"firing_rate": 0.0912,
"activation_sparsity_pct": 90.88,
"total_synaptic_spikes": 374,
"quiescent_neurons": 465,
"latency_ms": 3.4
},
"hybrid_telemetry": {
"architecture_type": "Hybrid CNN-SNN",
"front_end": "Spatial Convolutional Feature Extractor (ANN)",
"back_end": "Temporal Leaky Integrate-and-Fire SNN Motion Filter",
"feature_map_dim": "[B, 64, 56, 56]",
"receptive_field": "7x7 Conv + 3x3 MaxPool",
"spatial_to_spike_sparsity": "90.88%",
"motion_event_filter_efficiency": "89.4%",
"synergy_mechanism": "Translates raw spatial structures while SNN layers process temporal motion."
},
"esg_energy_telemetry": {
"energy_savings_ratio": "42.8x",
"energy_reduction_pct": 97.7,
"joules_saved": 0.0031,
"scope3_carbon_offset_gco2e": 0.000372,
"compliance_certification": "ISO 14064 Scope-3 Green Compute Verified"
},
"spike_raster_ascii": " [N00] ........
[N01] ....|...
[N02] ..|.....
[N03] .......|
[N04] .....|..
[N05] .|......",
"completion": "Neuromorphic event-driven architecture processed input across spatiotemporal timesteps..."
}
D. Query SNN & Hybrid Model Catalog
7. Multi-Tenant S-LoRA Swarm & Adapter Registry API
<5ms Hot-SwapHigh-throughput CUDA kernel multiplexing 100+ fine-tuned LoRA adapters over resident base models in <5ms without reloading base weights. Supports dynamic adapter mounting, QLoRA fine-tuning submission, and token accounting.
Parameters Table:
| Parameter | Type | Status | Description |
|---|---|---|---|
| adapter_id | string | Required | Unique adapter slug or vault ID (e.g. legal-contract-analyzer-v1). |
| prompt | string | Required | Inference prompt context to run against the mounted LoRA adapter. |
| base_model | string | Optional | Resident base foundation model (default meta-llama/Meta-Llama-3-8B-Instruct). |
Framework Code:
cURL Request Example
curl -X POST "https://web-lon-01.ncx.one/api/v1/adapters/infer" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx" \
-d '{
"adapter_id": "legal-contract-analyzer-v1",
"prompt": "Evaluate indemnification clauses in this commercial vendor agreement."
}'
PHP Example
<?php
$payload = json_encode([
'adapter_id' => 'legal-contract-analyzer-v1',
'prompt' => 'Evaluate indemnification clauses in this commercial vendor agreement.'
]);
$ch = curl_init('https://web-lon-01.ncx.one/api/v1/adapters/infer');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $payload);
curl_setopt($ch, CURLOPT_HTTPHEADER, [
'Content-Type: application/json',
'Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx'
]);
echo curl_exec($ch);
curl_close($ch);
Python Example
import requests
res = requests.post(
"https://web-lon-01.ncx.one/api/v1/adapters/infer",
headers={"Authorization": "Bearer sk-ncx-xxxxxxxxxxxxxxxx"},
json={
"adapter_id": "legal-contract-analyzer-v1",
"prompt": "Evaluate indemnification clauses in this commercial vendor agreement."
}
)
print(res.json())
Query Enterprise Adapter Vault:
8. Multi-Modal Image & Audio Swarm API
FLUX & WhisperDistributed multi-modal inference pipeline executing high-speed image generation (FLUX.1 Schnell, SDXL Turbo) and automated audio transcription & synthesis (Whisper Large v3).
1. Image Generation (FLUX / SDXL)
Image Parameters:
| Parameter | Type | Status | Description |
|---|---|---|---|
| prompt | string | Required | Visual scene text description. |
| model | string | Optional | Diffusion engine (flux-schnell, sdxl-turbo). Default flux-schnell. |
| size | string | Optional | Resolution format: 1024x1024, 768x768, 1280x720. |
Image Success Response (HTTP 200 OK)
{
"created": 1725548400,
"data": [
{ "url": "https://compute.y3ti.uk/uploads/generated/img_9f81a2c3.png", "revised_prompt": "..." }
]
}
2. Audio Transcription (Whisper)
Audio Success Response (HTTP 200 OK)
{
"text": "The distributed GPU computing cluster processed all inference layers in 14.2 milliseconds."
}
9. High-Density Vector Embeddings API
OpenAI StandardHigh-throughput vector embedding generation for semantic search, RAG retrieval, and generative agent associative memory streams. Returns normalized 768 / 1024-dimensional float vectors.
Request Payload (JSON)
{
"model": "bge-large-en-v1.5",
"input": "Generative AI multi-agent associative memory stream"
}
Success Response Format (HTTP 200 OK)
{
"object": "list",
"data": [
{
"object": "embedding",
"index": 0,
"embedding": [-0.01245, 0.04891, -0.03812, 0.08912, 0.01293, "... (1024 dimensions)"]
}
],
"model": "bge-large-en-v1.5",
"usage": { "prompt_tokens": 8, "total_tokens": 8 }
}
10. Multi-Agent Workflow Canvas Runner API
DAG RunnerExecutes directed acyclic graph (DAG) multi-agent workflow pipelines designed in the visual Agent Canvas. Automatically resolves node data dependencies, fans out parallel branches, and enforces error boundaries.
Request Body (JSON)
{
"workflow_id": 14,
"inputs": {
"target_domain": "compute.y3ti.uk",
"audit_depth": "deep"
},
"async": false
}
Success Response Format (HTTP 200 OK)
{
"success": true,
"run_id": "wf_run_9f81bc20e",
"status": "completed",
"elapsed_ms": 342.15,
"node_results": {
"node_1_recon": { "status": "ok", "records_scanned": 12 },
"node_2_analyzer": { "status": "ok", "vulnerabilities_found": 0 }
}
}
11. Autonomous Agentic AI Swarm Orchestrator API
25-Tool Agentic LibraryDeploys multi-turn autonomous goal planners with self-healing tool calling sandboxes (WASM execution, bash CLI, file I/O, curl, database querying).
Request Body Payload (JSON)
{
"goal": "Write a python script to calculate fibonacci numbers and report execution speed.",
"planner_model": "qwen2.5:0.5b",
"max_token_budget": 50000,
"max_steps": 20
}
Success Response (HTTP 200 OK)
{
"session_id": "agent_sess_89f0a1c",
"goal": "Write a python script to calculate fibonacci numbers and report execution speed.",
"status": "completed",
"total_steps": 3,
"tools_executed": ["write_file", "run_command", "read_output"],
"final_output": "Script created and executed. Calculated fibonacci(40) in 0.28s.",
"tokens_consumed": 412
}
12. Agent Mesh Task Handoff & Raft Swarm APIs
P2P Task HandoffEnables autonomous agent instances to hand off sub-goals directly to neighboring specialized agents over the Raft consensus mesh without round-tripping to the central database.
Success Response (HTTP 200 OK)
{
"status": "handed_off",
"task_id": "task_sub_49201",
"target_agent": "Python-Code-Validator",
"raft_leader": "Node-Spark-01",
"lease_ttl_sec": 300
}
13. WebAssembly Zero-Trust Task Execution API
WASI SandboxExecutes untrusted client WASM binaries inside memory-isolated sandboxes with microsecond startup latency (<2ms) and strict CPU fuel/gas meter caps.
Request Body (JSON)
{
"wasm_base64": "AGFzbQEAAAA...",
"entry_point": "_start",
"memory_limit_mb": 64,
"gas_limit": 1000000
}
Success Response (HTTP 200 OK)
{
"exit_code": 0,
"stdout": "Computation verified: 42",
"stderr": "",
"execution_time_ms": 1.48,
"gas_used": 42080
}
14. Multi-GPU Pipeline Parallelism API
Pipeline ParallelismOrchestrates multi-GPU pipeline parallelism (PP) and tensor parallelism (TP) for large-scale training jobs across pooled bare-metal compute rigs.
Success Response (HTTP 200 OK)
{
"status": "orchestrated",
"job_id": 1042,
"pp_degree": 4,
"tp_degree": 2,
"stages_allocated": 4,
"cross_node_bandwidth_gbps": 100.0
}
15. Enterprise ZK Proof Verification API
Zero-Knowledge MLCryptographic proof validation verifying that AI inference outputs or machine learning computations were executed faithfully without exposing model weights or private training inputs.
Success Response (HTTP 200 OK)
{
"verified": true,
"proof_id": "zk_0x89ab12c",
"prover_scheme": "Groth16-BN254",
"verification_time_ms": 4.12
}
16. Platform Statistics & Graph Telemetry API
Timeseries Graph Data
Zero-PII Enforced
Retrieve generic, anonymized ecosystem timeseries metrics and graph datasets for host earnings, GPU hardware telemetry (load %, power, temp), AI inference throughput, and fleet node utilization. Zero-PII Protection: No user IDs, wallet balances, host IP addresses, machine serials, or private prompts are ever exposed.
Success Response Payload Format (HTTP 200 OK)
{
"success": true,
"timestamp": 1784930392,
"timespan": "7d",
"metric_type": "all",
"privacy_policy": {
"data_type": "Generic Anonymized Ecosystem Aggregates",
"pii_protection": "Zero-PII Enforced (No User IDs, No IPs, No Machine Serials, No Private Prompts)",
"aggregation": "Fleet-Wide Statistical Averages"
},
"summary": {
"total_compute_nodes": 12,
"online_nodes": 9,
"total_active_units": 48,
"total_tflops_capacity": 17450.5,
"avg_gpu_temperature_c": 64.2,
"avg_power_draw_w": 272.5,
"total_inferences_24h": 14280,
"revenue_usd_timespan": 12450.8
},
"charts": {
"revenue_trend": {
"title": "Host Earnings & Revenue Trend ($)",
"labels": ["Aug 05", "Aug 06", "Aug 07", "Aug 08", "Aug 09", "Aug 10", "Aug 11"],
"datasets": [
{ "name": "Host Earnings ($)", "data": [1420.5, 1680.0, 1550.25, 1890.1, 2100.0, 1950.4, 1859.55] },
{ "name": "Platform Fee Share ($)", "data": [142.05, 168.0, 155.02, 189.01, 210.0, 195.04, 185.95] }
]
},
"gpu_brand_distribution": {
"title": "GPU Compute Brand Distribution",
"labels": ["NVIDIA RTX 4090", "NVIDIA CMP 170HX", "NVIDIA A100 80GB", "AMD RX 7900 XTX", "Intel Gaudi 2 / Core"],
"series": [45, 28, 12, 10, 5]
},
"telemetry_timeseries": {
"title": "Real-Time Telemetry & Hardware Load",
"labels": ["00:00", "03:00", "06:00", "09:00", "12:00", "15:00", "18:00", "21:00"],
"datasets": [
{ "name": "Average GPU Load (%)", "data": [75, 82, 88, 94, 91, 96, 89, 81] },
{ "name": "Power Draw per Node (W)", "data": [240, 255, 275, 290, 285, 298, 280, 260] }
]
}
}
}
17. Real-Time WebRTC Video Diffusion & Edge Streaming Mesh API
Sub-50ms Glass-to-GlassFull-duplex WebRTC and WebSocket streaming gateway facilitating interactive camera streaming, real-time video latent diffusion (FLUX.1 Schnell, SDXL Turbo, AnimateDiff), autonomous drone vision ingestion (YOLOv10), and "Stream-Certified" GPU node benchmarking (<15ms RTT).
1. Initiate WebRTC Stream Session & SDP Signaling
Request Body Format (JSON)
{
"action": "initiate",
"stream_type": "webrtc_diffusion",
"model": "flux-schnell",
"resolution": "1280x720",
"target_fps": 30,
"prompt": "Cinematic photorealistic 8k cyber city with neon reflections and rain"
}
Success Response Format (HTTP 200 OK)
{
"status": "success",
"session_token": "stream_9f82ab1c4e7d0139_1786884029",
"assigned_node": {
"rig_id": 1,
"rig_name": "DGX-Spark-01",
"is_stream_certified": true,
"badge_level": "ultra_low_latency",
"est_ping_ms": 12.4
},
"ice_servers": [
{ "urls": "stun:stun.l.google.com:19302" },
{ "urls": "stun:gateway.compute.y3ti.uk:3478" }
],
"ws_signaling_url": "wss://gateway.compute.y3ti.uk/ws_webrtc?session=stream_9f82ab1c4e7d0139_1786884029"
}
2. Real-Time Video Latent Diffusion Engine
Request Body Format (JSON)
{
"action": "diffuse_frame",
"session_token": "stream_9f82ab1c4e7d0139_1786884029",
"prompt": "Anime Studio Ghibli lush vibrant landscape, hand-drawn anime aesthetic",
"guidance_scale": 7.5,
"noise_strength": 0.65,
"model": "flux-schnell"
}
Success Response Format (HTTP 200 OK)
{
"status": "success",
"frame_id": 42,
"model": "flux-schnell",
"tensor_engine": "TensorRT-10.4-StreamDiffusion",
"inference_time_ms": 24.18,
"fps_rendered": 41.3,
"metrics": {
"cuda_vram_mb": 4820,
"glass_to_glass_latency_ms": 36.2
}
}
3. Autonomous Drone Vision & Telemetry Ingestion
Success Response Format (HTTP 200 OK)
{
"status": "success",
"frame_id": 105,
"inference_engine": "TensorRT-YOLOv10-Edge",
"inference_time_ms": 11.4,
"fps_throughput": 87.7,
"detections_count": 2,
"objects": [
{ "label": "Autonomous Drone", "category": "aerial", "confidence": 0.94, "box": [0.12, 0.45, 0.28, 0.62] },
{ "label": "Industrial Solar Array", "category": "infrastructure", "confidence": 0.98, "box": [0.35, 0.10, 0.85, 0.45] }
],
"anomaly_alert": false
}
18. Finance, Billing & Crypto Invoicing API
User API Key Only
BitPay Auto-Settled
RESTful financial operations enabling mobile clients and automated agents to monitor real-time wallet balances, active compute burn rate ($/hr), runway days, generate instant BitPay crypto deposit invoices (BTC, USDT, USDC, LTC, SOL), and inspect paginated transaction ledgers. Authenticated strictly with the user's secret API key (Authorization: Bearer sk-ncx-*).
1. Real-Time Balance & Burn Rate Telemetry
Success Response Format (HTTP 200 OK)
{
"success": true,
"balance_usd": 1420.50,
"currency": "USD",
"telemetry": {
"burn_rate_per_hour": 1.25,
"runway_hours": 1136.4,
"runway_days": 47,
"active_compute_jobs": 2,
"total_compute_spend": 384.20,
"total_inference_spend": 19.85,
"lifetime_deposited": 1824.55
},
"tax_status": {
"iso_country": "GB",
"tax_exempt": false,
"vat_rate_pct": 20.0
}
}
2. Create Instant Crypto Deposit Invoice
Request Body Format (JSON)
{
"amount_usd": 100.00,
"currency": "BTC", // Options: BTC, ETH, USDT, USDC, SOL, LTC, BCH, DOGE
"return_url": "https://compute.y3ti.uk/billing"
}
Success Response Format (HTTP 200 OK)
{
"success": true,
"invoice_id": "BP_84A9F21C0E",
"order_id": "NCX-DEP-142-1725048000",
"amount_fiat": 100.00,
"fiat_currency": "USD",
"crypto_amount": 0.00147058,
"crypto_currency": "BTC",
"payment_url": "https://test.bitpay.com/invoice?id=BP_84A9F21C0E",
"qr_code_url": "https://api.qrserver.com/v1/create-qr-code/?size=280x280&data=https%3A%2F%2Ftest.bitpay.com%2Finvoice%3Fid%3DBP_84A9F21C0E",
"status": "pending",
"expires_at": "2026-08-30T20:45:00Z"
}
3. Paginated Invoices & Transaction History
Success Response Format (HTTP 200 OK)
{
"success": true,
"balance_usd": 1420.50,
"page": 1,
"limit": 20,
"total_records": 1,
"total_pages": 1,
"invoices": [
{
"id": 104,
"reference_id": "BP_84A9F21C0E",
"invoice_type": "deposit",
"amount": 100.00,
"currency": "USD",
"payment_method": "bitpay",
"status": "completed",
"created_at": "2026-08-30 19:15:00",
"payment_url": "https://test.bitpay.com/invoice?id=BP_84A9F21C0E"
}
]
}
4. Compute Hourly Rates & Model Pricing Catalog
19. ISO Tax Withholding Calculator & Stripe Express API
DAC7 / 1099 TaxCalculates real-time statutory withholding tax percentages, DAC7 marketplace reporting thresholds, and VAT Reverse Charge rules for global compute hosts.
Success Response (HTTP 200 OK)
{
"country": "GB",
"gross_amount": 1000.00,
"withholding_rate_pct": 0.0,
"vat_applicable": false,
"vat_rate_pct": 20.0,
"net_payout": 1000.00,
"compliance_regime": "UK-HMRC-DAC7"
}
20. Multi-Tenant Partner System & Margins API
Role: Partner / Admin
Dynamic Skinning
Complete programmatic control for whitelabel partners to monitor tenant sub-users, configure custom reseller markups (margins), and synchronize dynamic theme color palettes and branding assets in real time.
1. Partner Tenant Overview & Volume Telemetry
Success Response Format (HTTP 200 OK)
{
"success": true,
"partner": {
"id": 12,
"partner_name": "NuCompute Partner",
"partner_slug": "acme-compute",
"revenue_share_pct": 15.0,
"branding": {
"primary_color": "#ff5e14",
"accent_color": "#00cfdd",
"body_bg_color": "#0c0e12",
"card_bg_color": "#141821"
}
},
"metrics": {
"sub_users_count": 48,
"fleet_rigs_count": 16,
"monthly_volume_usd": 14250.00,
"reseller_earnings_usd": 2137.50
}
}
2. Tenant Sub-Users Directory
3. Configure Reseller Margins & Revenue Share
Request Body (JSON)
{
"revenue_share_pct": 20.0
}
21. Investor Network & Yield Ledger API
Role: Investor / Admin
Fractional GPU Yield
Audited financial infrastructure for institutional and fractional GPU investors to inspect live portfolio valuations, fractional allocations, physical hardware telemetry, and paginated daily compute lease yield ledgers.
1. Investor Portfolio Dashboard & APY Performance
Success Response Format (HTTP 200 OK)
{
"success": true,
"investor_id": 142,
"portfolio_summary": {
"total_valuation_usd": 100000.00,
"allocated_gpus": 32,
"active_nodes_count": 4,
"apy_annualized_pct": 24.8,
"currency": "USD"
},
"yield_metrics": {
"last_24h": { "gross_revenue_usd": 284.50, "net_revenue_usd": 241.80 },
"last_30d": { "gross_revenue_usd": 8535.00, "net_revenue_usd": 7254.75 }
}
}
2. Hardware Nodes & Live Thermals
3. Paginated Daily Revenue Ledger
22. Sovereign Space & Orbital Edge Mesh APIs
CCSDS 734.2-B-1 DTN
LEO/MEO Edge Mesh
Delay-Tolerant Networking (DTN Bundle Protocol v7 RFC 9171 / CCSDS 734.2-B-1), contact-graph orbital pass routing, multispectral Earth Observation (EO) edge sensor ingestion, and spaceborne Zero-Knowledge execution attestation seals verifying orbital model inference without memory corruption outside Earth's atmosphere.
1. Bundle Protocol v7 Dispatch & Custody Transfer
| Parameter | Type | Status | Description |
|---|---|---|---|
| source_eid | string | Required | Source Endpoint Identifier (e.g. dtn://svalbard-station.ncx.earth). |
| destination_eid | string | Required | Destination Endpoint Identifier (e.g. dtn://leo-sat-04.ncx.space). |
| payload_hex | string | Required | Hex-encoded raw bundle payload or base64 binary stream. |
| priority | string | Optional | Bundle priority level: bulk, normal (default), or expedited. |
| lifetime_seconds | integer | Optional | Time-to-live expiration window before bundle drop (default: 86400). |
Framework Code:
cURL Dispatch Example
curl -X POST "https://web-lon-01.ncx.one/api/v1/dtn/bundle.php" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx" \
-d '{
"source_eid": "dtn://svalbard-station.ncx.earth",
"destination_eid": "dtn://leo-sat-04.ncx.space",
"payload_hex": "48656c6c6f204f72626974616c2045646765",
"priority": "expedited",
"lifetime_seconds": 86400
}'
PHP Example
<?php
$payload = json_encode([
'source_eid' => 'dtn://svalbard-station.ncx.earth',
'destination_eid' => 'dtn://leo-sat-04.ncx.space',
'payload_hex' => '48656c6c6f204f72626974616c2045646765',
'priority' => 'expedited',
'lifetime_seconds' => 86400
]);
$ch = curl_init('https://web-lon-01.ncx.one/api/v1/dtn/bundle.php');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $payload);
curl_setopt($ch, CURLOPT_HTTPHEADER, [
'Content-Type: application/json',
'Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx'
]);
$res = curl_exec($ch);
curl_close($ch);
print_r(json_decode($res, true));
Python Example
import requests
res = requests.post(
"https://web-lon-01.ncx.one/api/v1/dtn/bundle.php",
headers={"Authorization": "Bearer sk-ncx-xxxxxxxxxxxxxxxx"},
json={
"source_eid": "dtn://svalbard-station.ncx.earth",
"destination_eid": "dtn://leo-sat-04.ncx.space",
"payload_hex": "48656c6c6f204f72626974616c2045646765",
"priority": "expedited",
"lifetime_seconds": 86400
}
)
print(res.json())
Success Response Format (HTTP 200 OK)
{
"success": true,
"bundle_id": "bndl_77a9c8f2",
"custody_accepted": true,
"next_hop": "GS-SVALBARD",
"est_ground_window_utc": "2026-09-08T04:15:00Z",
"status": "en_route"
}
2. Earth Observation (EO) Sensor Ingest
Request Payload (JSON)
{
"satellite_id": "NCX-LEO-04",
"sensor_type": "multispectral",
"target_region": "North Sea Offshore Wind Farm Sector B",
"compressed_payload_hex": "ffd8ffe000104a464946..."
}
Success Response Format (HTTP 200 OK)
{
"success": true,
"ingest_id": "eo_ing_4481b0a7",
"satellite_id": "NCX-LEO-04",
"sensor_type": "multispectral",
"resolution_meters": 0.35,
"target_region": "North Sea Offshore Wind Farm Sector B",
"ground_station": "GS-SVALBARD",
"storage_tier": "cold_vault_s3",
"status": "ingested_verified"
}
3. Orbital ZK State Snapshot Verification
4. Ground Station Orbital Passes & Contact Windows
23. Sovereign Model Distillation & 1.58-Bit BitNet APIs
1.58-bit Ternary Absmax
Knowledge Distillation
Enterprise-grade teacher-to-student LLM knowledge distillation pipelines, chain-of-thought (CoT) reasoning trace compression, Hinton loss weighting (τ, α, β), and 1.58-bit ternary matrix quantization (BitNet b1.58) eliminating FP16 multipliers from GPU compute kernels for sovereign, resource-constrained deployments.
1. Distillation Pipeline Dispatch API
| Parameter | Type | Status | Description |
|---|---|---|---|
| teacher_model | string | Required | Teacher LLM identifier (e.g. deepseek-ai/DeepSeek-R1-Distill-Qwen-8B). |
| student_base_model | string | Required | Target student base model (e.g. deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B). |
| job_name | string | Required | Unique descriptive label for distillation job. |
| distillation_method | string | Optional | cot_reasoning_trace (default), kl_divergence_logits, or hybrid_loss. |
| temperature_t | float | Optional | Softmax distillation temperature τ (default: 2.0). |
| alpha_kd | float | Optional | Hinton loss weighting factor α for KL divergence (default: 0.75). |
| beta_ce | float | Optional | Student cross-entropy task loss factor β (default: 0.25). |
Framework Code:
cURL Dispatch Example
curl -X POST "https://web-lon-01.ncx.one/api/v1/distill/index.php" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx" \
-d '{
"teacher_model": "deepseek-ai/DeepSeek-R1-Distill-Qwen-8B",
"student_base_model": "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
"job_name": "Sovereign Tactical Reasoning Student 1.5B",
"distillation_method": "cot_reasoning_trace",
"temperature_t": 2.0,
"alpha_kd": 0.75,
"beta_ce": 0.25
}'
PHP Example
<?php
$payload = json_encode([
'teacher_model' => 'deepseek-ai/DeepSeek-R1-Distill-Qwen-8B',
'student_base_model' => 'deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B',
'job_name' => 'Sovereign Tactical Reasoning Student 1.5B',
'distillation_method' => 'cot_reasoning_trace'
]);
$ch = curl_init('https://web-lon-01.ncx.one/api/v1/distill/index.php');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $payload);
curl_setopt($ch, CURLOPT_HTTPHEADER, [
'Content-Type: application/json',
'Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx'
]);
$res = curl_exec($ch);
curl_close($ch);
print_r(json_decode($res, true));
Python Example
import requests
res = requests.post(
"https://web-lon-01.ncx.one/api/v1/distill/index.php",
headers={"Authorization": "Bearer sk-ncx-xxxxxxxxxxxxxxxx"},
json={
"teacher_model": "deepseek-ai/DeepSeek-R1-Distill-Qwen-8B",
"student_base_model": "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
"job_name": "Sovereign Tactical Reasoning Student 1.5B",
"distillation_method": "cot_reasoning_trace"
}
)
print(res.json())
Success Response Format (HTTP 200 OK)
{
"success": true,
"job_id": 4,
"job_uuid": "dist_68a01f92",
"job_name": "Sovereign Tactical Reasoning Student 1.5B",
"teacher_model": "deepseek-ai/DeepSeek-R1-Distill-Qwen-8B",
"student_base_model": "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
"status": "queued",
"compression_target": "5.3x parameter reduction"
}
2. Distillation Job Status & Telemetry API
3. 1.58-Bit Ternary (BitNet b1.58) Quantization API
Request Payload (JSON)
{
"model_name": "meta-llama/Llama-3-8B-Instruct",
"weight_bits": 1.58,
"activation_bits": 8,
"zero_multiplication_mode": 1
}
Success Response Format (HTTP 200 OK)
{
"success": true,
"job_uuid": "quant_b158_9921c",
"model_name": "meta-llama/Llama-3-8B-Instruct",
"original_vram_gb": 16.0,
"quantized_vram_gb": 2.4,
"bandwidth_reduction_pct": 85.0,
"energy_efficiency_gain_pct": 71.4,
"baseline_perplexity": 6.18,
"quantized_perplexity": 6.32,
"perplexity_delta": 0.14,
"status": "completed"
}
24. Air-Gapped SCIF & Vector Vault API
ICD 705 / DoD 5220.22-M
TPM 2.0 PCR Sealing
Packaging offline sovereign AI appliances for Secure Compartmented Information Facilities (SCIF) with Ed25519 digital tamper seals and zero-egress network profiles, alongside hardware-bound vector vaults sealed against TPM 2.0 PCR registers and AES-256-GCM envelope encryption.
1. Air-Gapped SCIF Appliance Packager API
| Parameter | Type | Status | Description |
|---|---|---|---|
| package_name | string | Required | Sovereign SCIF appliance descriptive title. |
| model_source | string | Required | Model identifier or local checkpoint path. |
| target_hardware_architecture | string | Optional | nvidia_hopper, nvidia_blackwell, amd_mi300x, or intel_gaudi3. |
| offline_license_duration_days | integer | Optional | Air-gap offline validity window in days (default: 365). |
Framework Code:
cURL Dispatch Example
curl -X POST "https://web-lon-01.ncx.one/api/v1/scif/package.php" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx" \
-d '{
"package_name": "Sovereign SCIF Defense LLM Appliance",
"model_source": "meta-llama/Llama-3.3-70B-Instruct",
"target_hardware_architecture": "nvidia_hopper",
"offline_license_duration_days": 365
}'
PHP Example
<?php
$payload = json_encode([
'package_name' => 'Sovereign SCIF Defense LLM Appliance',
'model_source' => 'meta-llama/Llama-3.3-70B-Instruct',
'target_hardware_architecture' => 'nvidia_hopper',
'offline_license_duration_days' => 365
]);
$ch = curl_init('https://web-lon-01.ncx.one/api/v1/scif/package.php');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $payload);
curl_setopt($ch, CURLOPT_HTTPHEADER, [
'Content-Type: application/json',
'Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx'
]);
$res = curl_exec($ch);
curl_close($ch);
print_r(json_decode($res, true));
Python Example
import requests
res = requests.post(
"https://web-lon-01.ncx.one/api/v1/scif/package.php",
headers={"Authorization": "Bearer sk-ncx-xxxxxxxxxxxxxxxx"},
json={
"package_name": "Sovereign SCIF Defense LLM Appliance",
"model_source": "meta-llama/Llama-3.3-70B-Instruct",
"target_hardware_architecture": "nvidia_hopper",
"offline_license_duration_days": 365
}
)
print(res.json())
Success Response Format (HTTP 200 OK)
{
"success": true,
"package_uuid": "scif_pkg_7718eb29",
"package_name": "Sovereign SCIF Defense LLM Appliance",
"ed25519_signature_seal": "8f0a2b4c1d6e7f8091a2b3c4d5e6f7a8b9c0d1e2f3a4b5c6d7e8f9a011223344",
"license_token": "ncx_scif_lic_365d_994a2b1c",
"checksum_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"export_tar_url": "/api/v1/scif/download?package_uuid=scif_pkg_7718eb29"
}
2. Hardware-Bound Private RAG & Vector Vault API
Vector Search Request (JSON)
{
"action": "search",
"vault_uuid": "vault_sec_88e41c9b",
"query": "Radiation-tolerant orbital compute firmware measurements",
"top_k": 3
}
Vector Ingestion Request (JSON)
{
"action": "ingest",
"vault_uuid": "vault_sec_88e41c9b",
"document_title": "Classified Operational Directives",
"payload_text": "Tactical orbital mesh route validated under radiation hardening protocol.",
"classification_marking": "SECRET//NOFORN"
}
25. Enterprise Confidential Data Clean Rooms API
NVIDIA CC / AMD SEV-SNP
Remote Attestation
Multi-party confidential computing sandboxes where proprietary datasets from multiple enterprise tenants are ingested, matched, and processed strictly within hardware Trusted Execution Environments (TEE) with cryptographic attestation, ephemeral key agreements, and continuous differential privacy accounting.
1. Clean Room Listing & Metrics API
Success Response Format (HTTP 200 OK)
{
"success": true,
"total_cleanrooms": 2,
"cleanrooms": [
{
"id": 1,
"uuid": "cr_h100_oncology_consortium",
"name": "Multi-Institution Oncology Clinical Trial Clean Room",
"tee_type": "nvidia_cc",
"attestation_status": "attested",
"privacy_budget_total_epsilon": 10.0,
"privacy_budget_spent_epsilon": 1.25,
"privacy_budget_remaining_epsilon": 8.75,
"participant_count": 3,
"computation_count": 5
}
]
}
2. Hardware TEE Remote Attestation API
| Parameter | Type | Status | Description |
|---|---|---|---|
| action | string | Required | Action command: attest_enclave. |
| cleanroom_uuid | string | Required | Target clean room UUID (e.g. cr_h100_oncology_consortium). |
Framework Code:
cURL Attestation Request
curl -X POST "https://web-lon-01.ncx.one/api/v1/cleanrooms/index.php" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx" \
-d '{
"action": "attest_enclave",
"cleanroom_uuid": "cr_h100_oncology_consortium"
}'
PHP Example
<?php
$payload = json_encode([
'action' => 'attest_enclave',
'cleanroom_uuid' => 'cr_h100_oncology_consortium'
]);
$ch = curl_init('https://web-lon-01.ncx.one/api/v1/cleanrooms/index.php');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $payload);
curl_setopt($ch, CURLOPT_HTTPHEADER, [
'Content-Type: application/json',
'Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx'
]);
$res = curl_exec($ch);
curl_close($ch);
print_r(json_decode($res, true));
Python Example
import requests
res = requests.post(
"https://web-lon-01.ncx.one/api/v1/cleanrooms/index.php",
headers={"Authorization": "Bearer sk-ncx-xxxxxxxxxxxxxxxx"},
json={
"action": "attest_enclave",
"cleanroom_uuid": "cr_h100_oncology_consortium"
}
)
print(res.json())
Success Response Format (HTTP 200 OK)
{
"success": true,
"cleanroom_uuid": "cr_h100_oncology_consortium",
"cleanroom_name": "Multi-Institution Oncology Clinical Trial Clean Room",
"tee_type": "nvidia_cc",
"attestation_status": "attested",
"hardware_root_of_trust": "NVIDIA Remote Attestation Service (NRAS)",
"pcr_measurements": {
"pcr_0": "ded6a506b01b01d6706e23de3b839f9b5d4481b0a7...",
"pcr_2": "a4d709292a40669b59e35f8c85775f0a28b14c330e...",
"pcr_7": "e3b0c44298fc1c149afbf4c8996fb92427ae41e464..."
},
"enclave_public_key_pem": "-----BEGIN PUBLIC KEY-----\nMCowBQYDK2VuAyEA...",
"verification_receipt": "ATTESTATION_SEAL_cr_h100_oncology_consortium_..."
}
3. Encrypted Participant Data Deposit API
Request Payload (JSON)
{
"action": "deposit_data",
"cleanroom_uuid": "cr_h100_oncology_consortium",
"deposit_name": "Genomic Variant Vectors (Cohort 2)",
"payload_data": {
"patient_cohort_size": 2500,
"feature_dimensions": 2048,
"target_biomarkers": ["TP53", "BRCA1", "EGFR"]
}
}
4. Collaborative Enclave Computation Dispatch
Request Payload (JSON)
{
"action": "run_computation",
"cleanroom_uuid": "cr_h100_oncology_consortium",
"computation_type": "joint_lora_finetune",
"model_target": "meta-llama/Llama-3-70B-Instruct",
"epsilon_cost": 1.25,
"steps": 500
}
26. Differential Privacy Synthetic Studio API
(ε, δ)-DP Guaranteed
1-Wasserstein Metric
Mathematical synthetic data generation studio that samples from confidential clean room data distributions while injecting calibrated Laplace or Gaussian noise to mathematically guarantee (ε, δ)-Differential Privacy with automated 1-Wasserstein distribution distance computation and feature correlation preservation auditing.
Mathematical Privacy Guarantee & Wasserstein Metric:
ƒ̃(D) = ƒ(D) + Lap(0, Δƒ / ε)
σ = (Δ₂ƒ · √(2 ln(1.25 / δ))) / ε
W₁(P, Q) = (1 / N) ∑ |X̂ᵢ - Ŷᵢ|
1. Generate DP Synthetic Dataset API
| Parameter | Type | Status | Description |
|---|---|---|---|
| cleanroom_uuid | string | Required | Host clean room UUID (e.g. cr_h100_oncology_consortium). |
| dataset_type | string | Optional | Domain schema template: oncology_genomics, financial_aml, or compute_telemetry. |
| sample_count | integer | Optional | Synthetic sample count to synthesize (10 to 10,000, default: 1000). |
| epsilon | float | Optional | Privacy budget loss factor ε > 0 (default: 1.0). |
| delta | float | Optional | Relaxation factor δ ∈ (10-9, 10-3) (default: 1e-5). |
| mechanism | string | Optional | Noise perturbation algorithm: laplace (default) or gaussian. |
Framework Code:
cURL Generation Example
curl -X POST "https://web-lon-01.ncx.one/api/v1/cleanrooms/synthetic.php" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx" \
-d '{
"cleanroom_uuid": "cr_h100_oncology_consortium",
"dataset_type": "oncology_genomics",
"sample_count": 25,
"epsilon": 0.5,
"delta": 0.00001,
"mechanism": "laplace"
}'
PHP Example
<?php
$payload = json_encode([
'cleanroom_uuid' => 'cr_h100_oncology_consortium',
'dataset_type' => 'oncology_genomics',
'sample_count' => 25,
'epsilon' => 0.5,
'delta' => 1e-5,
'mechanism' => 'laplace'
]);
$ch = curl_init('https://web-lon-01.ncx.one/api/v1/cleanrooms/synthetic.php');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $payload);
curl_setopt($ch, CURLOPT_HTTPHEADER, [
'Content-Type: application/json',
'Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx'
]);
$res = curl_exec($ch);
curl_close($ch);
print_r(json_decode($res, true));
Python Example
import requests
res = requests.post(
"https://web-lon-01.ncx.one/api/v1/cleanrooms/synthetic.php",
headers={"Authorization": "Bearer sk-ncx-xxxxxxxxxxxxxxxx"},
json={
"cleanroom_uuid": "cr_h100_oncology_consortium",
"dataset_type": "oncology_genomics",
"sample_count": 25,
"epsilon": 0.5,
"delta": 1e-5,
"mechanism": "laplace"
}
)
print(res.json())
Success Response Format (HTTP 200 OK)
{
"success": true,
"job_uuid": "synth_5e63df836919",
"cleanroom_uuid": "cr_h100_oncology_consortium",
"dataset_name": "Oncology Genomic Biomarkers & Expression Profiles",
"sample_count": 25,
"privacy_parameters": {
"epsilon": 0.5,
"delta": 1e-05,
"mechanism": "LAPLACE",
"guarantee": "(0.5, 1.0E-5)-Differential Privacy"
},
"privacy_budget_telemetry": {
"total_budget_epsilon": 10.0,
"cost_debited_epsilon": 0.5,
"remaining_budget_epsilon": 8.25,
"budget_utilization_pct": 17.5
},
"fidelity_metrics": {
"mean_wasserstein_distance": 0.0508,
"correlation_matrix_preservation_pct": "94.7%"
},
"download_token": "dptok_4f81c90a112233445566778899aabbcc"
}
27. Fully Homomorphic Encryption (FHE) & ZK Verification API
Ring R_q = Z_q[X]/(X^N + 1)
Fiat-Shamir ZK Proof
Lattice-based homomorphic ciphertext tensor arithmetic (CKKS / BFV schemes) executing matrix-vector multiplications, polynomial activations, and neural inference directly over encrypted ciphertexts without decryption on untrusted GPU workers, backed by non-interactive Fiat-Shamir ZK proof validation.
1. Homomorphic Ciphertext Evaluation API
| Parameter | Type | Status | Description |
|---|---|---|---|
| cleanroom_uuid | string | Required | Host clean room UUID (e.g. cr_h100_oncology_consortium). |
| scheme | string | Optional | FHE scheme: ckks (fixed-point, default) or bfv (exact integer). |
| operation | string | Optional | Tensor operation: matrix_vector_mult, polynomial_activation, or ciphertext_addition. |
| poly_modulus_degree | integer | Optional | Polynomial ring degree N ∈ {4096, 8192, 16384} (default: 8192). |
| coeff_mod_bits | integer | Optional | Coefficient modulus log2(q) bit-depth (default: 218). |
Framework Code:
cURL FHE Evaluation Example
curl -X POST "https://web-lon-01.ncx.one/api/v1/cleanrooms/fhe.php" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx" \
-d '{
"cleanroom_uuid": "cr_h100_oncology_consortium",
"scheme": "ckks",
"operation": "matrix_vector_mult",
"input_vector": [0.45, -0.12, 0.88, 0.23],
"poly_modulus_degree": 8192,
"coeff_mod_bits": 218
}'
PHP Example
<?php
$payload = json_encode([
'cleanroom_uuid' => 'cr_h100_oncology_consortium',
'scheme' => 'ckks',
'operation' => 'matrix_vector_mult',
'input_vector' => [0.45, -0.12, 0.88, 0.23]
]);
$ch = curl_init('https://web-lon-01.ncx.one/api/v1/cleanrooms/fhe.php');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $payload);
curl_setopt($ch, CURLOPT_HTTPHEADER, [
'Content-Type: application/json',
'Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx'
]);
$res = curl_exec($ch);
curl_close($ch);
print_r(json_decode($res, true));
Python Example
import requests
res = requests.post(
"https://web-lon-01.ncx.one/api/v1/cleanrooms/fhe.php",
headers={"Authorization": "Bearer sk-ncx-xxxxxxxxxxxxxxxx"},
json={
"cleanroom_uuid": "cr_h100_oncology_consortium",
"scheme": "ckks",
"operation": "matrix_vector_mult",
"input_vector": [0.45, -0.12, 0.88, 0.23]
}
)
print(res.json())
Success Response Format (HTTP 200 OK)
{
"success": true,
"query_uuid": "fhe_qry_feb6e2fac088",
"cleanroom_uuid": "cr_h100_oncology_consortium",
"scheme": "CKKS",
"operation": "matrix_vector_mult",
"ring_parameters": {
"poly_modulus_degree_N": 8192,
"coeff_mod_bits_log2_q": 218,
"quotient_ring": "R_q = Z_218[X]/(X^8192 + 1)",
"multiplicative_depth": 1
},
"noise_budget_telemetry": {
"initial_noise_margin_db": "54.0 dB",
"consumed_noise_margin_db": "4.08 dB",
"remaining_noise_margin_db": "49.92 dB",
"noise_margin_status": "NOMINAL"
},
"verifiable_zk_proof": {
"zk_proof_hash": "71a4c444adc0ea09d0b16f39e4481b0a...",
"prover_system": "Fiat-Shamir Non-Interactive Succinct Verifier",
"verification_status": "CRYPTOGRAPHICALLY_VERIFIED",
"soundness_guarantee": "2^-128 computational soundness"
},
"evaluated_ciphertext": "c0_0x82fbe534dcd0...c1_0xa4d709292a...",
"inference_latency_ms": 138.4
}
2. ZK Proof Verification API
Query Parameters:
| Parameter | Type | Status | Description |
|---|---|---|---|
| action | string | Required | Action command: verify (or verify_proof). |
| query_uuid | string | Required | FHE query receipt UUID to verify (e.g. fhe_qry_feb6e2fac088). |
| proof_hash | string | Optional | Expected Fiat-Shamir proof hash for cryptographic tamper checking. |
Verification Response (HTTP 200 OK)
{
"valid": true,
"query_uuid": "fhe_qry_feb6e2fac088",
"scheme": "BFV",
"zk_proof_hash": "zk_0x89ab12c4789d0e1f3a5b6c7d8e...",
"noise_budget_consumed_db": "4.21 dB",
"verification_status": "VERIFIED",
"verified_at": "2026-09-08 12:00:00"
}
28. Cryptographic Purge Vault & NIST SP 800-88 Compliance API
NIST SP 800-88 Rev 2
GDPR Art. 17 / HIPAA Certified
Automated multi-pass volatile memory sanitization vault that executes 3-pass zeroization (0x00 binary overwrite, 0xFF complement overwrite, and CSPRNG cryptographic entropy wipe) followed by cryptographic zero-state validation and immutable HMAC-SHA512 audit certificate emission.
1. Trigger NIST SP 800-88 Purge API
| Parameter | Type | Status | Description |
|---|---|---|---|
| action | string | Required | Purge action command: purge. |
| cleanroom_uuid | string | Required | Target clean room UUID (e.g. cr_h100_oncology_consortium). |
| memory_bytes | integer | Optional | Allocated memory capacity to sanitize in bytes (default: 17179869184 / 16 GB). |
| standard | string | Optional | Sanitization standard: NIST_SP_800_88_REV2 (default) or DOD_5220_22_M. |
Framework Code:
cURL Purge Request
curl -X POST "https://web-lon-01.ncx.one/api/v1/cleanrooms/compliance.php" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx" \
-d '{
"action": "purge",
"cleanroom_uuid": "cr_h100_oncology_consortium",
"memory_bytes": 17179869184,
"standard": "NIST_SP_800_88_REV2"
}'
PHP Example
<?php
$payload = json_encode([
'action' => 'purge',
'cleanroom_uuid' => 'cr_h100_oncology_consortium',
'memory_bytes' => 17179869184,
'standard' => 'NIST_SP_800_88_REV2'
]);
$ch = curl_init('https://web-lon-01.ncx.one/api/v1/cleanrooms/compliance.php');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $payload);
curl_setopt($ch, CURLOPT_HTTPHEADER, [
'Content-Type: application/json',
'Authorization: Bearer sk-ncx-xxxxxxxxxxxxxxxx'
]);
$res = curl_exec($ch);
curl_close($ch);
print_r(json_decode($res, true));
Python Example
import requests
res = requests.post(
"https://web-lon-01.ncx.one/api/v1/cleanrooms/compliance.php",
headers={"Authorization": "Bearer sk-ncx-xxxxxxxxxxxxxxxx"},
json={
"action": "purge",
"cleanroom_uuid": "cr_h100_oncology_consortium",
"memory_bytes": 17179869184,
"standard": "NIST_SP_800_88_REV2"
}
)
print(res.json())
Success Response Format (HTTP 200 OK)
{
"success": true,
"cert_uuid": "PURGE-CERT-F6DE3015-927C",
"cleanroom_uuid": "cr_h100_oncology_consortium",
"sanitization_standard": "NIST_SP_800_88_REV2",
"sanitization_protocol": {
"pass_1": "Fixed 0x00 Binary Overwrite (All addressable memory channels)",
"pass_2": "Fixed 0xFF Complementary Overwrite (Charge dissipation)",
"pass_3": "CSPRNG Cryptographic Entropy + Final 0x00 Zeroization"
},
"memory_sanitized": {
"bytes_cleared": 17179869184,
"gigabytes_cleared": "16 GB",
"sanitization_duration_ms": 142.6
},
"cryptographic_verification": {
"pre_purge_buffer_sha256": "82ce1389e130b6c482fbe534dcd0d83d48060fe2d0cc27ce30656483bb3aecd0",
"post_purge_zero_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"zero_state_validated": true,
"enclave_attestation_seal": "6815a5cdd6f18f0a2b4c1d6e7f8091a2b3c4d5e6f7a8b9c0..."
},
"signed_receipt_token": "ncx_purge_sig_736c60da584f2910bc...",
"public_verification_url": "/api/v1/cleanrooms/compliance?action=verify&cert_uuid=PURGE-CERT-F6DE3015-927C"
}
2. Public Audit Certificate Token Verification
Query Parameters:
| Parameter | Type | Status | Description |
|---|---|---|---|
| action | string | Required | Action command: verify. |
| cert_uuid | string | Required | Cryptographic purge certificate UUID (e.g. PURGE-CERT-F6DE3015-927C). |
| token | string | Optional | HMAC-SHA512 signed receipt token for cryptographic tamper validation. |
Verification Response (HTTP 200 OK)
{
"valid": true,
"cert_uuid": "PURGE-CERT-F6DE3015-927C",
"cleanroom_uuid": "cr_h100_oncology_consortium",
"sanitization_standard": "NIST_SP_800_88_REV2",
"bytes_cleared": 17179869184,
"post_purge_zero_hash": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"enclave_attestation_seal": "6815a5cdd6f18f0a2b4c1d6e7f8091a2b3c4d5e6f7a8b9c0...",
"signed_receipt_token": "ncx_purge_sig_736c60da584f2910bc..."
}