Files

T

wehub-resource-sync d26aa360a1 chore: import upstream snapshot with attribution

2026-07-13 13:37:43 +08:00

9.7 KiB

Raw Permalink Blame History

FlowSentinel — AI Workflow Command Center

A Next.js dashboard that orchestrates AI agent workflows through n8n, powered by Nebius LLMs, with every human and AI action tracked in Velt's immutable activity log, all exposed securely through Tailscale Funnel.

How It Works

┌─────────────────────────────────────────────────────────────────┐
│                    FLOWSENTINEL ARCHITECTURE                     │
│                                                                 │
│   ┌───────────────────────────────────────────────────────┐     │
│   │              Next.js Dashboard (UI)                   │     │
│   │  ┌──────────┐  ┌────────────┐  ┌──────────────────┐  │     │
│   │  │  Chat    │  │  Workflow  │  │  Activity Feed   │  │     │
│   │  │  Panel   │  │  Panel    │  │  (Velt SDK)      │  │     │
│   │  └────┬─────┘  └─────┬─────┘  └────────┬─────────┘  │     │
│   └───────┼──────────────┼─────────────────┼─────────────┘     │
│           │              │                 │                    │
│   ┌───────▼──────┐ ┌────▼──────────┐ ┌────▼─────────────┐     │
│   │ Nebius LLM   │ │ n8n Webhook   │ │ Velt Activity    │     │
│   │ Nemotron     │ │ + API         │ │ Logs API         │     │
│   │ via Token    │ │ Orchestration │ │ Immutable Trail  │     │
│   │ Factory API  │ │               │ │                  │     │
│   └──────────────┘ └───────────────┘ └──────────────────┘     │
│                                                                 │
│   ┌─────────────────────────────────────────────────────────┐   │
│   │              Tailscale Funnel                           │   │
│   │   localhost:3000  ──▶  https://you.ts.net (public)     │   │
│   └─────────────────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────────────────┘

What It Does

Component	Tool	Role
AI Brain	Nebius Token Factory	Powers chat + workflow planning with Nemotron Super
Live Research	Built-in Web Research	Fetches time-sensitive updates (for example, latest vendor announcements) and grounds responses with sources
Orchestrator	n8n	Chains multi-step AI workflows via webhook + visual automation
Auditor	Velt Activity Logs	Immutable, real-time trail of every human & AI action
Shield	Tailscale Funnel	Secure HTTPS exposure without ports or cloud infra

Project Structure

flowsentinal_audittrail/
├── src/
│   ├── app/
│   │   ├── layout.tsx            # Root layout with Velt provider
│   │   ├── page.tsx              # Dashboard entry point
│   │   ├── globals.css           # Tailwind + custom design system
│   │   └── api/
│   │       ├── chat/route.ts     # Nebius LLM chat endpoint
│   │       ├── workflow/route.ts # n8n workflow trigger + status polling
│   │       └── activities/route.ts # Activity log CRUD
│   ├── components/
│   │   ├── Dashboard.tsx         # Three-panel layout
│   │   ├── ChatPanel.tsx         # AI chat with Nebius
│   │   ├── WorkflowPanel.tsx     # n8n workflow runner
│   │   ├── ActivityFeed.tsx      # Velt activity timeline
│   │   ├── Header.tsx            # Status bar + live indicators
│   │   └── providers/
│   │       └── VeltWrapper.tsx   # Velt SDK initialization
│   └── lib/
│       ├── nebius.ts             # Nebius client (OpenAI-compatible)
│       ├── webResearch.ts        # Live web research helpers for time-sensitive queries
│       ├── n8n.ts                # n8n workflow client
│       └── activity.ts           # Local + Velt cloud activity logging helpers
├── README.md
├── setup-tailscale.sh            # One-command Tailscale Funnel setup
├── package.json
├── .env.example
├── tailwind.config.ts
├── tsconfig.json
└── next.config.mjs

Quick Start

1. Clone and install

git clone https://github.com/Arindam200/awesome-ai-apps.git
cd awesome-ai-apps/advance_ai_agents/flowsentinal_audittrail
npm install
cp .env.example .env.local

2. Get your API keys

Service	Where to get it	Env variable
Nebius	Nebius	`NEBIUS_API_KEY`
n8n	n8n.io/download or `npx n8n`	`N8N_WEBHOOK_URL`, `N8N_API_KEY` (optional), `N8N_BASE_URL`
Velt	console.velt.dev	`NEXT_PUBLIC_VELT_API_KEY` (UI), `VELT_API_KEY` (backend ingestion)

cp .env.example .env.local
# Edit .env.local with your keys and n8n webhook URL
# Optional: add VELT_API_KEY for backend activity ingestion to Velt Cloud

3. Run the dashboard

npm run dev
# Open http://localhost:3000

4. Set up your n8n webhook workflow

The app sends workflow trigger requests to n8n automatically when you click Plan in the chat. You need to create a matching webhook workflow in n8n:

Open n8n at http://localhost:5678 (or run npx n8n to start it).
Create a new workflow and add a Webhook node as the trigger.
Set the Webhook node to HTTP Method: POST and note the generated URL.
Set Respond: to When Last Node Finishes (so the app gets a synchronous response).
Add any downstream nodes you want (e.g. HTTP Request, Set, Code, etc.).
The app sends this JSON body to your webhook:

{
  "query": "user's original question",
  "analysis": "Nebius AI analysis summary",
  "steps": "Step 1 | Step 2 | Step 3"
}

Your n8n workflow can return any JSON — the app will display it in the Workflow panel log.
Copy the webhook URL (use the Test URL while developing, Production URL when live) into N8N_WEBHOOK_URL in .env.local.
Activate the workflow (toggle top-right) before using Production URL.

Tip: Start simple — a Webhook → Set node that echoes the inputs back is enough to confirm the integration works end-to-end.

5. (Optional) Expose via Tailscale Funnel

# Requires Tailscale v1.52+ installed and logged in
chmod +x setup-tailscale.sh
./setup-tailscale.sh

# Or directly:
tailscale funnel 3000

Your dashboard is now accessible at https://your-machine.your-tailnet.ts.net/ — share this URL with your team.

The Dashboard

Panel	What it does
Chat Panel	Talk to Nemotron Super via Nebius — ask questions, get analysis, plan workflows
Workflow Panel	Trigger and monitor n8n executions in real-time
Activity Feed	Immutable timeline of every action — human, AI, and system — powered by Velt

API Endpoints

Method	Path	Description
`POST`	`/api/chat`	Send messages to Nebius LLM, supports chat and workflow planning modes
`POST`	`/api/workflow`	Trigger an n8n workflow via webhook
`GET`	`/api/workflow?run_id=X`	Check workflow status
`GET`	`/api/activities`	Fetch activity log with optional type filter
`POST`	`/api/activities`	Log a custom activity event
`GET`	`/api/diagnostics/velt`	View backend Velt sync status and last error
`POST`	`/api/diagnostics/velt/ping`	Emit a diagnostic event to test Velt ingestion

Assistant Response Pattern

FlowSentinel chat is designed to return:

Direct answer first (## Answer)
Sources for time-sensitive queries (## Sources Checked)
Optional automation follow-up (## Optional Automation (n8n))

Use the Plan button when you want a full workflow plan staged for execution.

Models

Uses Nemotron Super (configurable via NEBIUS_MODEL) via Nebius Token Factory API. The integration stays OpenAI-compatible (no new SDK required) and can be extended to Nebius fine-tuned/custom models for broader production use cases.

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.tokenfactory.nebius.com/v1/",
  apiKey: process.env.NEBIUS_API_KEY,
});

9.7 KiB Raw Permalink Blame History