AI Automation Platform
· 3 min · sphinxagent.ai
Architecture
React Frontend → API Gateway → Lambda
↓
DynamoDB + SQS Queues
↓
OpenAI / Claude APIs
Sphinx Agent (formerly SnapIt Agent) lets users build custom AI workflows without coding. A visual workflow builder connects to a serverless execution engine that orchestrates API calls, data transforms, and AI prompts. Now live at sphinxagent.ai.
Workflow Execution
Each workflow is a series of steps - AI prompts, data transforms, webhooks, and conditionals. The execution engine processes steps sequentially, passing context between them.
// Step types in the workflow engine
switch (step.type) {
case 'ai_prompt':
context = await executeAIStep(step, context);
break;
case 'data_transform':
context = await transformData(step, context);
break;
case 'webhook':
context = await callWebhook(step, context);
break;
case 'condition':
// Branch based on context evaluation
break;
}
Multi-Provider AI
The platform supports both OpenAI and Anthropic Claude models. API keys are stored encrypted in SSM Parameter Store, never in code or environment variables.
Async Processing
Long-running workflows use SQS queues with dead letter queues for reliability. If a step fails, the workflow can be resumed from the last successful checkpoint.
Performance
- Execution: 500ms average per step
- Concurrency: 1,000+ agents simultaneously
- Reliability: 99.95% with DLQ handling
- Real-time monitoring: WebSocket-based execution tracking