"Turn Natural Language into Executable Infrastructure." Automating the robotic drudgery of the 99% using Agentic AI.
In the modern enterprise, automation is gated by complexity.
- Zapier/n8n require you to understand "Webhooks", "JSON parsing", and logic loops.
- Camunda requires you to be a BPMN engineer.
- Custom Code requires a development team and weeks of time.
This leaves 90% of the workforce—Operations Managers, HR Leads, Small Business Owners—stuck doing repetitive manual work because they cannot code.
Orvexia is an Agentic AI Platform that democratizes software engineering. It allows users to build enterprise-grade automation workflows simply by speaking.
We don't just "chat." We build. Our AI acts as a Senior Workflow Architect that translates intent into infrastructure, connects APIs, handles error logic, and deploys live workflows in seconds.
- Framework: React.js (Vite)
- Visualization:
React Flow(for rendering the interactive node graph) - Styling: Tailwind CSS + Framer Motion
- State Management: Zustand / React Context
- API: FastAPI (Python)
- AI Engine: LangChain + Groq (Llama 3 70B)
- Validation: Pydantic (Strict Schema Enforcement)
- Database: PostgreSQL / SQLite (for persistent workflow state)
- Orchestration: LangGraph (Stateful multi-actor applications)
- Tooling: Custom "Tool Sandbox" (Gmail, GitHub, Notion, Slack, Vercel APIs)
- Memory: Context-aware history (The AI remembers previous modifications)
The core of Orvexia. Users type a command like:
"When I get an email about a bug, summarize it with AI, log it to Notion, and alert the #dev-team on Slack." The Agent intelligently selects the right tools, draws the connections, and configures the payloads automatically.
We provide a "Glass Box" experience. Users can see exactly what the AI built using a drag-and-drop node graph (powered by React Flow). This ensures trust and allows for manual fine-tuning.
Track the ROI of automation.
- Time Saved: Calculated based on average manual execution time.
- Success Rate: Monitor workflow health.
- Execution Trends: See which workflows are driving value.
Our AI doesn't just execute; it advises. If a user builds a risky workflow, the AI suggests improvements:
"I noticed you are creating a GitHub issue but didn't add a notification step. Should I add a Slack alert for you?"
We solve the "AI Hallucination" problem with a 3-Layer Defense System:
- Strict Schema Validation (Pydantic): The AI is forced to output strict JSON matching our internal schemas. If it tries to invent a non-existent parameter, the backend rejects it before it reaches the UI.
- Deterministic Tool Sandboxing: The AI cannot execute arbitrary code. It can only call pre-defined, safe functions from our
TOOLS_DB. - Human-in-the-Loop: We favor "Co-piloting" over "Auto-piloting." The AI drafts the blueprint, but the human must verify the visual diagram before deployment.
| Feature | 🐢 Traditional (Zapier/Camunda) | ⚡ Orvexia (Agentic AI) |
|---|---|---|
| Creation Speed | Hours (Manual Drag & Drop) | Seconds (Natural Language) |
| User Requirement | Technical (Logic/APIs) | Non-Technical (Intent only) |
| Flexibility | Rigid (Static Workflows) | Fluid (Self-Correcting AI) |
| Cost | High Engineering Overhead | Zero Engineering Overhead |
- Zero-Friction Accessibility: Unlocks automation for the 500M+ non-technical knowledge workers.
- Instant ROI: Reduces "Time-to-Automation" by 99%.
- Shadow IT Reduction: Provides a governed, safe environment for employees to build tools without using insecure, unapproved apps.
- Python 3.9+
- Node.js 16+
- Groq API Key
- Clone the Repo
git clone [https://github.com/2232def/orvexia.git](https://github.com/2232def/orvexia.git) cd orvexia - Backend setup
cd backend pip install -r requirements.txt # Create .env file with GROQ_API_KEY uvicorn app.main:app --reload
- Frontend setup
cd frontend npm install npm run dev
👥 Contributors [Anay Mishra] [Dev Kumar Singh] [Niloy Mallik] [Abhijit Mondal]