Feature · Knowledge bases

Your content. Their answers.

Upload PDFs, paste URLs, sync Notion or Drive. Your agent answers from your content, with citations.

The problem

RAG is harder than it looks.

Naive vector search misses context. Re-indexing is fragile. Big documents time out. And nobody wants to maintain another pipeline that breaks every time the source changes.

The Oaura way

A pipeline that just works.

Drop in a source, Oaura ingests, chunks, embeds, and keeps it in sync. Hybrid retrieval (vector + keyword) returns precise passages with citations. Big jobs run in a durable background worker so they never time out.

How it works

Step by step.

STEP 01

Connect a source

Upload a file, paste a URL, or connect Notion, Drive, Slack, or Confluence. Oaura watches for changes and re-indexes.

STEP 02

Tune retrieval

Pick chunk size, embedding model, and reranking strategy. Or trust the defaults — they've been tuned on real production data.

STEP 03

Ground responses

Your agent answers from retrieved chunks with inline citations. Hallucinations drop. Trust goes up.

TypeScript · ingest API
await oaura.knowledgeBases.attach({
  agentId: "agt_a1b2c3",
  source: { type: "notion", databaseId: "..." },
  reindex: "incremental",
});
What's included

Out of the box.

  • PDF, DOCX, TXT, HTML, CSV ingestion
  • Notion, Google Drive, Slack, Confluence sync
  • Hybrid retrieval (vector + keyword)
  • Inline citations in widget responses
  • Durable background worker for big jobs
  • Schema upgrader for embedding migrations
  • Per-knowledge-base ACLs
  • Pluggable vector store (Qdrant default)
More

Behind the design.

Knowledge bases are the foundation of any production agent. The pipeline above ships out of the box — but if you have a stranger source, the same ingestion API accepts custom adapters.

Stand up your first agent today.

Five-minute setup. Bring your own LLM. No credit card.