https://github.com/TheIndicSentinel/kavachxv2
Production-grade AI Governance Engine for real-time risk scoring, compliance enforcement (DPDP/RBI), and immutable audit logs.
https://github.com/TheIndicSentinel/kavachxv2
ai-governance bharat compliance dpdp it-act llm-safety privacy-engineering responsible-ai
Last synced: 7 days ago
JSON representation
Production-grade AI Governance Engine for real-time risk scoring, compliance enforcement (DPDP/RBI), and immutable audit logs.
- Host: GitHub
- URL: https://github.com/TheIndicSentinel/kavachxv2
- Owner: TheIndicSentinel
- License: apache-2.0
- Created: 2026-03-15T19:21:36.000Z (4 months ago)
- Default Branch: main
- Last Pushed: 2026-04-16T16:05:12.000Z (3 months ago)
- Last Synced: 2026-06-12T00:14:35.732Z (about 1 month ago)
- Topics: ai-governance, bharat, compliance, dpdp, it-act, llm-safety, privacy-engineering, responsible-ai
- Language: Python
- Size: 15.4 MB
- Stars: 1
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
- Security: SECURITY.md
Awesome Lists containing this project
README
# ๐ก๏ธ KavachX โ The AI Governance Infrastructure for Bharat
**Real-time Policy Enforcement ยท Multi-Gate Risk Scoring ยท Immutable Compliance Auditing**
[](LICENSE)
[](https://www.python.org/)
[](https://github.com/TheIndicSentinel/kavachxv2)
[](docs/architecture.md)
---
## โก The Enforcement Gap
As AI models integrate into the core of Bharat's digital infrastructureโfrom Fintech to Healthtechโthe gap between **intelligence** and **accountability** is widening. Every inference is a potential compliance risk, a privacy leak, or a security breach.
**KavachX** (sanskrit: *Kavach* - Armor) is the software-defined enforcement layer that closes this gap. It provides a robust, low-latency "Digital Armor" that intercepts, analyzes, and regulates every interaction between AI models and users in real-time.
---
## โจ Core Pillars of Governance
| Pillar | Capability | Technical Driver |
|:---:|:---|:---|
| **๐ Security** | Immune to Prompt Injection & NAEL | Adversarial Noise Classifiers |
| **๐งฌ Safety** | Context-Aware Hate & Harm Filtering | Fine-tuned Transformer Models |
| **๐ Compliance** | DPDP 2023 & RBI Policy Adherence | Regex + NLP PII Detection |
| **๐ Audit** | Immutable Evidence Chains | Cryptographic Interaction Logging |
| **๐ Flexibility** | Universal AI Interceptor | Headless API & Browser Extension |
---
## ๐๏ธ Architecture: The 4-Gate Enforcement Pipeline
```text
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ AI MODEL (LLM / ML System) โ
โโโโโโโโโโโโโโโโโโโโฌโโโโโโโโโโโโโโโโโโโโโโโ
โ Inference Request
โผ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ K A V A C H X E N G I N E โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Gate 1: ๐ Security โ Prompt Injection / NAEL โ
โ Gate 2: ๐งฌ Safety โ Hate, Bias, Harm Scoring โ
โ Gate 3: ๐ Compliance โ DPDP PII / IT Act Check โ
โ Gate 4: ๐ Audit โ Immutable Log + Dashboard โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Governed Response
โผ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ END USER / CLIENT โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
```
---
## ๐ Quick Start in 60 Seconds
### ๐ฆ 1. Installation
```bash
git clone https://github.com/TheIndicSentinel/kavachxv2.git
cd kavachxv2/backend
pip install -r requirements.txt
```
### ๐ 2. Basic Initialization
```python
from kavachx import Sentinel
# Initialize the governance gate
sentinel = Sentinel(api_key="your_kavach_key")
# Governed inference
response = sentinel.inspect(
prompt="Generate a response for the user...",
context="banking_domain"
)
if response.is_blocked:
print(f"Action Blocked: {response.violation_reason}")
else:
print(response.content)
```
---
## ๐บ๏ธ Roadmap & Current Focus
- [x] **v2.0 Core:** Real-time interception and basic gate logic.
- [ ] **v2.5 Bhasha-Shield:** Fine-tuned safety classifiers for 12+ Indian languages.
- [ ] **v3.0 Edge-Enforce:** Moving governance logic to the client-side/edge for zero-latency.
- [ ] **v3.5 DPDPA-Masker:** Specialized PII redaction for Indian data patterns (Aadhaar, PAN).
---
## ๐ค Community & Support
* **Documentation:** Detailed guides in the [docs/](docs/) folder.
* **Contributing:** Read our [CONTRIBUTING.md](CONTRIBUTING.md) to get started.
* **Security:** Report vulnerabilities via [SECURITY.md](SECURITY.md).
* **Contact:** Reach out at [inerd1412@gmail.com](mailto:inerd1412@gmail.com).
---
**๐ก๏ธ KavachX** โ Policy-Aligned AI Infrastructure for Bharat
*Built with conviction in ๐ฎ๐ณ India*