TQNN Python SDK

Official Python SDK for the TQNN Fault-Tolerant Inference Platform.

The TQNN Python SDK provides a lightweight interface for interacting with the managed TQNN cloud runtime.

Build confidence-aware applications that continue operating when real-world data is noisy, incomplete, uncertain, or partially unavailable.

Features

• Official Python SDK for TQNN
• Managed cloud inference
• Confidence-aware predictions
• Input-integrity reporting
• Controlled decision states
• Runtime diagnostics
• Domain-agnostic inference
• Zero endpoint configuration

Installation

pip install -U tqnn

Requirements

• Python 3.9+
• Active TQNN API key

Don't have one yet?

Create a free Explorer account and receive 100 free API calls at:

https://tqnnlabs.com

Quick Start

from tqnn import TQNNClient

client = TQNNClient(
    api_key="TQNN_xxxxxxxxxxxxxxxxx"
)

response = client.run_any(
    data=[1.2, 1.4, 1.6, 2.1],
    mode="TABULAR",
    task="fault_diagnosis"
)

print(response)

The SDK automatically connects to the official TQNN cloud runtime.

No endpoint configuration is required.

Creating a Client

from tqnn import TQNNClient

client = TQNNClient(
    api_key="YOUR_API_KEY"
)

Running Inference

response = client.run_any(
    data=data,
    mode="TABULAR",
    task="fault_diagnosis",
    label="pump_sensor_window",
    metadata={
        "class_labels": [
            "healthy",
            "fault"
        ]
    }
)

Parameters

Example Response

{
  "platform": "TQNN",
  "engine": "TQNN Fault-Tolerant Inference Engine",
  "engine_version": "2.0.0",
  "mode": "TABULAR",

  "result": {
    "prediction_index": 1,
    "prediction_label": "fault",
    "confidence": 0.94,
    "decision": "ACCEPT"
  },

  "tqnn_report": {
    "primary_capability": "fault-tolerant inference",

    "confidence": {
      "score": 0.94,
      "label": "high",
      "margin": 0.28,
      "entropy": 0.17
    },

    "data_integrity": {
      "score": 0.98,
      "label": "valid",
      "feature_count": 4,
      "finite_fraction": 1.0,
      "missing_fraction": 0.0,
      "constant_input": false
    },

    "decision": {
      "status": "ACCEPT",
      "threshold_met": true
    },

    "warnings": []
  },

  "diagnostics": {
    "probabilities": [0.06, 0.94],
    "confidence_margin": 0.28,
    "normalized_entropy": 0.17,
    "acceptance_threshold": 0.61,
    "threshold_met": true,
    "qualia": {},
    "intent": [],
    "meta": {}
  }
}

Convenience Methods

result = client.result(response)

report = client.report(response)

diagnostics = client.diagnostics(response)

Decision States

Every inference returns a standardized decision.

• ACCEPT
• REVIEW
• REJECT

Applications can use these states to determine whether an inference should proceed automatically or be reviewed before action.

Supported Modes

Example Applications

• Fault Detection
• Fault Diagnosis
• Industrial Monitoring
• Sensor Health
• Predictive Maintenance
• Biosignal Analysis
• Financial Analytics
• Scientific Data
• General Structured Data

Error Handling

try:
    response = client.run_any(
        data=data,
        mode="TABULAR"
    )

except Exception as e:
    print(e)

Best Practices

Store API keys using environment variables.

export TQNN_API_KEY="TQNN_xxxxxxxxxxxxxxxxx"

import os
from tqnn import TQNNClient

client = TQNNClient(
    api_key=os.environ["TQNN_API_KEY"]
)

Never commit API keys to public repositories.

Version Compatibility

This SDK targets the public TQNN Fault-Tolerant Inference API v2.

Future SDK releases will remain compatible with the evolving public API contract.

Philosophy

The SDK is intentionally lightweight.

Rather than abstracting platform responses, it exposes the complete standardized output returned by the managed runtime.

Applications receive:

• Predictions
• Confidence scores
• Decision states
• Input-integrity metrics
• Runtime diagnostics
• Structured metadata

This allows developers to build reliable systems while retaining full visibility into every inference.

License

The TQNN Python SDK is released under the MIT License.

Open Source

• Python SDK
• Client utilities
• Integration helpers
• Example code

Proprietary

• Managed inference runtime
• Execution substrate
• Platform orchestration
• Production infrastructure

About TQNN Labs

TQNN Labs develops cloud-hosted fault-tolerant inference infrastructure for applications operating on noisy, incomplete, uncertain, or unreliable data.

The platform combines:

• Hybrid quantum-classical execution
• Confidence-aware inference
• Input-integrity reporting
• Controlled decision logic
• Managed cloud deployment

Website:

https://tqnnlabs.com

Built in Canada 🇨🇦

Vision

TQNN Labs is building infrastructure that helps applications understand not only what was predicted, but how confidently that prediction can be trusted.

Reliable inference should continue working even when real-world data does not.

Build once. Infer through uncertainty.