AI PLATFORM · MULTI-LLM
Vote AI Gateway
Multi-LLM AI Platform — 3 AI Providers, 6 AI Modules, 5-Step Chatbot Safety Flow
Vote AI Gateway — 20s Overview
Project Video
Overview
A multi-LLM gateway integrating Claude, GPT, and Gemini with 6 AI modules — chatbot, text improvement, board summarization, report generation, content filtering, and more. The chatbot never calls an LLM: each question passes a 5-step safety flow (injection detection → risky-question block → numeric-request block → FAQ answer → safety notice) and is answered from the FAQ only.
Our Role
System Architecture · Multi-LLM Abstraction · 5-Step Chatbot Safety Flow · Admin Dashboard · Cost Optimization
Tech Stack
The Challenge
- 1Political neutrality — election law violations from AI bias
- 2Misinformation prevention — AI hallucinations on voting data
- 3Prompt injection attacks targeting safety-critical systems
- 4Cost control across multiple AI providers and modules
Our Solution
- ✓Provider-agnostic architecture with 3 LLM providers and automatic failover
- ✓5-step chatbot safety flow: injection detection → risky-question block → numeric-request block → FAQ answer → safety notice
- ✓API-key authentication and an integration guide for the client's Java/Spring system
- ✓Per-request cost tracking with budget warnings at 70%/90% thresholds
Key Features
Multi-LLM Router
3 providers, automatic switch to another provider when a call fails
5-Step Chatbot Safety Flow
Injection detection → risky-question block → numeric-request block → FAQ answer → safety notice
6 AI Modules
Chatbot, text improvement, summarization, reports, filtering
Admin Dashboard
Usage stats, monthly budget warnings, banned-word & FAQ management, model connection tests
Live Metrics
6
AI Modules in Production
30+
REST APIs
5
Chatbot Safety Steps
450ms
Avg. Chatbot Response