Skip to content

Releases: threatvec/SentinelAI

SentinelAI v1.1.0

31 May 14:39
@threatvec threatvec
v1.1.0
4aa7ce0

Choose a tag to compare

View all tags
SentinelAI v1.1.0 Latest
Latest
SentinelAI

SentinelAI v1.1.0

AI Security & Monitoring Platform

SentinelAI is a security platform built for the AI era. It combines three defenses in one toolkit:

Module What it does
🔍 Code Scanner Static analysis for hard-coded secrets, OWASP Top 10 (SQLi, XSS, command/path injection), dependency CVEs, and AI-generated code smells
🛡️ LLM Firewall Blocks prompt-injection attempts, detects and redacts PII, tracks token usage and cost budgets, and logs every interaction
👁️ Agent Monitor Tracks runtime agent behavior (file, network, API access), flags anomalies, and offers a kill switch for rogue agents

It ships with a local-first web dashboard, JSON/HTML/SARIF reports, a GitHub Action, and a pre-commit hook — so security travels with your code from editor to CI.

✨ What's New in v1.1.0

This is a hardening and reliability release focused on closing security gaps and making the integrations actually work end-to-end.

🔒 Security fixes

Area Fix
HTML reports Patched a stored XSS — scan findings (file paths, code snippets, messages) are now properly escaped before being rendered into the HTML report, so attacker-controlled content in a scanned repo can no longer execute in the viewer's browser
Dashboard Added a path-traversal guard so dashboard file routes can no longer be tricked into reading files outside the intended directory

🐛 Bug fixes

Area Fix
GitHub Action Fixed the Action's summary JSON keys so the fail-on severity gate is read correctly — CI now actually fails the build when findings exceed the configured threshold instead of silently passing
PII detector Fixed an overlap-removal bug that could drop or mangle adjacent/overlapping PII matches during redaction; overlapping spans are now resolved deterministically
Dependency scanner Hardened version comparison so vulnerable-range matching no longer misfires on pre-release, build-metadata, or non-standard version strings

🚀 Improvements

Area Change
Custom rules Wired up custom rule loading from rules/*.yaml — drop your own detection rules into the rules/ directory and the engine picks them up automatically
Examples Rewrote the three broken examples (basic_scan.py, llm_firewall_example.py, agent_monitor_example.py) against the real, current APIs so they run as-is
Branding Added the official SentinelAI brand logo to the project and documentation

📦 Install

From PyPI:

pip install sentinelai

Docker (also published to GitHub Container Registry):

# Scan the current directory
docker run --rm -v "$(pwd):/scan" ghcr.io/threatvec/sentinelai:latest scan /scan

From source:

git clone https://github.com/threatvec/SentinelAI.git
cd SentinelAI
pip install -e ".[dev]"

Quick start

# Scan a project and write an HTML report
sentinelai scan ./my-project --rules secrets,owasp --output report.html --format html

# Launch the local dashboard
sentinelai dashboard --port 8000

Full Changelog: https://github.com/threatvec/SentinelAI/commits/v1.1.0

SentinelAI
Built by threatvec & talkdedsec — because AI needs a security guard too.
Assets 2
Loading

🛡️ SentinelAI v1.0.0 — Initial Release

06 Apr 05:56
@threatvec threatvec
v1.0.0
a7025e9

Choose a tag to compare

View all tags
🛡️ SentinelAI v1.0.0 — Initial Release

🛡️ SentinelAI v1.0.0 — Initial Release

AI Security & Monitoring Platform.

Features

  • 🔍 Secret Scanner — 50+ patterns for API keys, tokens, passwords, certificates
  • 🛡️ Code Scanner — OWASP Top 10: SQL injection, XSS, command injection, path traversal
  • 🔥 LLM Firewall — Prompt injection detection & PII protection
  • 🤖 Agent Monitor — Runtime behavior tracking with LangChain & CrewAI plugins
  • 📊 Web Dashboard — Local-first, dark theme, real-time monitoring
  • 📄 Report Generator — JSON, HTML, SARIF output formats
  • ⚡ CLI — 7 commands with Rich terminal output
  • 🐳 Docker — Dockerfile + docker-compose ready
  • ⚙️ GitHub Action — CI/CD integration
  • 🔗 Pre-commit hooks — Automated security scanning

Quick Start

pip install sentinelai
sentinelai scan .
sentinelai dashboard --port 8000

Docker

docker compose up scan
docker compose up dashboard

CI/CD

- uses: threatvec/SentinelAI@v1
  with:
    rules: "secrets,code,owasp"
    fail-on: "high"

Download

Download the .whl or .tar.gz below:

  • sentinelai-1.0.0-py3-none-any.whl — Python wheel (pip install)
  • sentinelai-1.0.0.tar.gz — Source distribution

Security — VirusTotal Scan

✅ 0/65 — Clean — No security vendors flagged this file as malicious.

SHA256: de1e152deacc6e8d2fa6ede7808953e54d88f507ce24b81c20ae607391dcf023

🔍 View Full VirusTotal Report

Stats

  • 90 files | 10,657 lines | 37 tests (all passing)
  • Python 3.11+ | FastAPI | Click | Rich

Made by talkdedsec & threatvec

Loading