HN Showcase
Back to browse
GitHub Repository

A sophisticated RAG intelligence engine for Singaporean laws, policies, and history. Comes with a triple-AI failover backend (Gemini/Llama/Groq), semantic embeddings using FAISS, and an Apple-inspired interactive UI. Designed with precision and high availability in mind.

76 starsPython

I built an webpage to showcase Singapore's infra and laws

by curiousbatman·Feb 12, 2026·1 point·0 comments
Visit ProjectView on HN

AI Analysis

○PassNiche GemWizardrySlick
The Take

Someone actually built uptime into the AI stack: a documented triple-failover for inference (Gemini Flash → Llama 3.3 via OpenRouter → Llama 3.3 via Groq) so demos don't die when one model is slow. The combo of 33k+ curated Singapore PDFs, local BGE-M3 embeddings and FAISS retrieval gives the project real credibility as an auditable knowledge engine, while the Framer glassmorphism UI shows attention to interaction. That said, reliance on proprietary inference endpoints and notable operational complexity could make reproduction and long-term hosting tricky.

Category
AI/ML
Target Audience

Legal researchers, policy analysts, journalists, civic tech developers and Singapore-focused information seekers

Post Description

Star the repo if you learned something new

Similar Projects

AI/ML●Mid

Built an webpage to show Singaporean infra and laws

Triple-LLM failover (Gemini → Llama 3.3 via OpenRouter → Groq), local BGE‑M3 embeddings and FAISS-backed retrieval show someone thought about latency and uptime, not just model demos. The README brags about 33k pages and 'non-hallucination' claims but stops short of evaluation details or realistic ops guidance — running 70B models and local embedding stacks is impressive on paper but a heavy lift in practice.

Niche GemWizardry
Goodguy27
323mo ago
AI/ML●●Solid

An beautiful webpage I made

There are real engineering moves here: 33k+ pages ingested, 1024-dim BGE-M3 embeddings served locally for privacy/latency, FAISS for millisecond retrieval and a clever 'triple‑AI' failover chain (Gemini → Llama via OpenRouter → Groq) to keep demos responsive. The frontend leans into Apple-style glassmorphism with Framer Motion interactions, so it actually feels like a thought-through product rather than a hack — biggest caveat is reliance on proprietary LLMs and infrastructure complexity for anyone wanting to reproduce it.

Niche GemWizardry
gigachadai
203mo ago
AI/ML●●Solid

Deploy a RAG pipeline as a REST API using RAGLight

Modular RAG with MCP integration, but Langchain and LlamaIndex already dominate.

Ship It
bessouat40
313mo ago