Internal Knowledge Base
Q&A over your entire company knowledge
Build a Q&A system over your internal wikis, Notion pages, Confluence docs, and Slack archives so employees get instant, sourced answers instead of spending hours searching or waiting for a colleague's reply.
GPT-6 Luna
for 3M tokens a month · 88% input / 12% output
Why this model
For this workload — 3M tokens a month, 88% of them input — it comes to about $0.44 a month, the lowest of the models suggested here. Its 1M-token context window holds long inputs in a single call, which matters for this kind of work.
Alternative models
Implementation tips
- 1
Implement hybrid search (BM25 + vector embeddings) for retrieval — keyword search catches exact policy names and version numbers that semantic search misses, while vector search handles paraphrased questions.
- 2
Always include the source document title, URL, and last-updated date alongside each retrieved chunk in the prompt — the model can include citations in its answer, and users can verify freshness instantly.
- 3
Set up automated re-indexing on document change events via webhooks — a knowledge base where 20% of answers are stale erodes trust faster than having no system at all.