Making RAG work for enterprise knowledge bases
Chunking
Split by meaning rather than fixed length and keep heading context.
Retrieval
Combine vector and keyword search for long-tail questions.
Re-ranking
Use a light re-ranker to drop noisy passages.
Citations
Attach sources to every answer for review and correction.
Related articles
NewsSeptember 19, 2026
Case study: a retail chain cut support response time by 70% with Nova Assistant
After connecting its knowledge base, first-contact resolution reached 68%.
Anonymous
InsightsOctober 3, 2026
Edge caching in practice: pushing API hit rates above 90%
Cache key design, invalidation and staged purges: lessons from our edge network.
Anonymous
InsightsSeptember 13, 2026
Designing observability for fleets of 10,000+ IoT devices
Heartbeats, offline detection and alert aggregation in practice.
Anonymous