Problem
Building a ranking product for a new niche normally means rebuilding the same architecture from scratch. New frontend, new data pipeline, new LLM prompts, new UI components. Each iteration is weeks of work even when the underlying pattern is identical.
Solution
A parameterized Next.js template + niche config schema + orchestration script that generates an entire multi-LLM ranking app from a single JSON config file. Point it at a new niche, run new-niche.sh, and get a working frontend with 4-LLM ensemble scoring.
Results
- 4-LLM ensemble (GPT, Gemini, Claude, Grok) per niche
- Powers Nash Satoshi, Glow Index, and college rankings
Niche config schema (JSON) defines all niche-specific parameters. Parameterized Next.js template consumes config at build time via next.config.js environment injection. Prompt generator creates niche-specific LLM scoring prompts from config templates. new-niche.sh orchestrates: create config, scaffold files, register LaunchAgent, assign port, generate health check runbook. 4-LLM ensemble (GPT-4o, Gemini 2.5, Claude Sonnet, Grok) scores in parallel, consensus verdict drives rankings.
Tech Stack
- Next.js
- TypeScript
- Multi-LLM Ensemble
- OpenAI GPT-4o
- Google Gemini
- Anthropic Claude
- xAI Grok
- n8n
- Python FastAPI
- Shell Scripting