Project Pitch:
I built a smart n8n workflow that combines multiple AI agents to transform vague user inputs into highly structured prompts using the RISEN framework.
It distributes tasks across different specialized LLMs, aggregates the results, and generates one "Perfect Prompt" ready for complex AI tasks.
It shows how AI collaboration and workflow orchestration can create much stronger, business-ready outputs — not just single-model answers.
- Receives user input through a webhook
- Distributes the input across four specialized AI agents (each focused on a different skill set)
- Aggregates their outputs intelligently
- Synthesizes a final "Perfect Prompt" ready for high-quality AI tasks
The flow distributes user input to specialized AI agents using LangChain + OpenRouter:
- n8n (workflow orchestration)
- LangChain nodes for agent management
- OpenRouter LLMs (Gemini, DeepSeek, Phi-3, Dolphin)
- RISEN framework for prompt refinement
- JSON merging and aggregation logic
- perfect-prompts-n8n-workflow.json — the exported n8n workflow file
This project demonstrates how multiple AI models can collaborate to enhance user instructions, leading to more precise and effective AI interactions.
It highlights real-world skills in building smart, modular AI pipelines ready for production environments.
This diagram shows how a spoken user input is transcribed using OpenAI Whisper and converted into a structured RISEN prompt via n8n.
- Add automatic prompt evaluation scoring (clarity, completeness, and specificity)
- Expand agent specialization (e.g., create agents focused on creative writing, technical documentation, or customer support)
- Integrate optional feedback loop where users rate the generated prompts for continuous model improvement
- Deploy as a public-facing API endpoint to allow external apps to generate structured prompts dynamically
- Enable user feedback collection for looped learning
Demo built for AI Agent Implementation Manager portfolio presentation.