Add kb-arena to LLMOps

[xmpuspus]

Adds kb-arena to the LLMOps section, placed alphabetically between Izlo and Keywords AI.

About kb-arena

Repo: https://github.com/xmpuspus/kb-arena

Open-source RAG evaluation arena that benchmarks 7 retrieval strategies on user-provided documents:

1. Naive vector retrieval
2. Contextual vector (Anthropic's contextual embeddings)
3. QnA-pair pre-indexing
4. Knowledge graph retrieval (Neo4j)
5. RAPTOR (recursive summarization)
6. PageIndex (hierarchical)
7. Hybrid (BM25 + semantic)

Why it fits LLMOps

• Evaluation and observability — RAGAS metrics (faithfulness, context precision/recall, answer relevancy)
• A/B testing — ELO-based blind Strategy Arena for head-to-head ranking
• Cost tracking — per-strategy token/call cost accounting
• Multi-provider — Anthropic / OpenAI / Ollama support
• Production-grade — 514 tests, FastAPI backend, Neo4j + ChromaDB, Next.js UI

Adjacent to entries like Evidently, Laminar, Langfuse, and Opik in the same LLMOps section — this fills a specific gap for RAG architecture benchmarking.

Checklist per contributing.md

• Individual PR for this suggestion
• Alphabetical placement (kb < Keywords)
• Format matches existing entries (Project / Details / Repository columns with star badge)
• Star badge URL follows `img.shields.io/github/stars//.svg?style=flat-square` convention
• Repo is active (last commit April 2026), not archived, not 404