Knowledge sources (RAG)

Knowledge lets an agent answer from your own content instead of guessing. Add a source, assign it to a bot, and the agent retrieves the relevant passages while it talks.

Source types

• Text — paste content directly (up to 50,000 characters per source).
• URL — point Hania at a page to fetch and index.
• File — upload documents (.txt, .pdf, .docx, .md, .html, .htm; up to 5 files at a time, ≤ 5 MB each).

How it's processed

When you add a source, Hania chunks it, generates embeddings, and indexes it for semantic retrieval — all automatically. Each source carries a processing_status (pending → processing → completed, or failed); wait for completed before relying on it. You can check counts by type and status from the knowledge stats.

Retrieval at conversation time

Assign a source to a bot, and the agent searches it during conversations, pulling in relevant passages to ground its answers. Keep sources focused and well-titled — clean, scoped content retrieves better than one giant dump.

Add & manage

Add and manage sources from the Knowledge area in the dashboard, or over the Knowledge API (JSON for text/URL, multipart for files). Delete is soft by default (recoverable); a hard delete removes the source and its embeddings permanently.